WO2004079314A1 - Colorimeter, colorimeter sensor unit and colour determination process - Google Patents

Abstract

A hand-held colorimeter (101) suitable for use by blind or colour-blind individuals to determine the colour of a surface, for example of a fabric, has an aperture (110) which, in use, is covered by the surface (113) whose colour is to be determined. Three LEDs (115, 116, 117) emitting orange, green, and blue light illuminate the surface and diffuse reflections therefrom containing orange, green, and blue spectrum sample values are used to determine the luminous reflectivity and chromaticity values for the colour of the surface. The measured values are compared with colorimetric values of reference surfaces and the best fit used to determine the reference surface whose colour is closest to the colour of the surface. The colorimeter may output the name of the colour aurally.

Description

COLORIMETER, COLORIMETER SENSOR UNIT AND COLOUR DETERM-QVATION PROCESS

BACKGROUND OF THE INVENTION

FIELD OF THE INVENTION

This invention relates to colorimeters, colorimeter sensor units and a colour determination process for use therewith.

DESCRIPTION OF THE PRIOR ART

People who are blind or colour-blind have difficulties when choosing their clothes from their wardrobes. They would be greatly assisted by a device that would measure the colour of items such as socks, shoes, shirts, pants etc., and announce the results in an audible or other non- isual format. To meet the requirements of this market, such a device preferably should be easily portable, preferably handheld, rugged and inexpensive.

There are many US patents which disclose devices for determining colour, including US Patent 3,060,790 (Ward); US 3,512,893 (Faulhaber et at); US 4,917,500 (Lugos); US 5,021,645 (Satula et al); US 5,303,037 (Taranowski); US 5,838,451 (McCarthy); US 5,963,333, US 6,020,583, and US 6,147,761 by (Walowit ei a ); US 6,157,454 (Wagner et al); and US 6,323,481 (Ueki). A disadvantage of these devices, however, is that they require_, the use of optical elements such as optical filters, light pipes/guides, lenses, mirrors, and reflector cones, which increases cost and reduces ruggedness.

It is possible to dispense with such intervening optical components in certain situations. For example, US 3,910,701 (Henderson et al.) discloses photometric instruments having a plurality of LEDs and at least one photodetector and one of which has no intervening optics. Henderson et al. were not concerned with determining colour, however, but primarily with the detection of diseases in humans and plants. Their focus was upon determining reflectivity and absorption at different wavelengths rather than the colour of a surface, i.e., spectrometric rather than colorimetric. Another example, disclosed in US 5,671,059 (Vincent), is a colorimeter for a desktop printer which uses electroluminescent emitters without intervening lenses or other optical components to measure the colours being printed and which compares their characteristics with the digital image data used to generate them and so allow for correction of errors. When suggesting that this colorimeter could be used in a hand-held colour probe, however, Vincent _, states that a lens, optical reflector or other optical components positioned between the colorimeter and a colour sample may be employed to optimize the optical performance in a desired application. This would, of course, increase expense and reduce ruggedness.

US 5, 137,364, also by McCarthy, discloses a colorimeter which uses a plurality of light emitting diodes surrounding a set of photodetectors with a shield for preventing direct irradiation of the detectors by the emitters, but no intervening optics. A disadvantage of the devices disclosed by Henderson et al, Vincent and McCarthy

'364, however, is that they do not address the problem of specular reflection affecting colour determination.

Reflection from an incident ray or pencil of light may be categorized into two parts, a diffuse reflection part and a specular reflection part. Specular reflection is characteristic of a smooth, glossy surface, the reflection from a good mirror being entirely specular. Diffuse reflection is characteristic of a rough or matte surface and, in the Lambertian model of diffuse reflection, is scattered into a hemisphere, i.e., in all directions.

The colour of a surface is determined by the spectral variation of the reflectivity over the visible range. However, the spectral variation of the specular component may be (and usually is) not the same as the spectral variation of the diffuse component. In fact, for many common surfaces, the specular reflectivity is substantially independent of colour and therefore the spectrum of the reflected light is substantially the same as that of the incident light.

For almost all purposes, the colour of a surface is deemed to be determined by the spectral characteristics of the diffuse reflectivity. If specularly reflected components of the reflected light are collected for measurement, they will usually result in errors in the diffuse colour determination of the surface.

While materials such as cloth and fabric generally have low levels of specular reflectivity and the measurement of their colours would not be greatly affected by including the specular components, measurements of more glossy surfaces characteristic of leather or vinyl for shoes or jackets would be severely affected. These limitations are addressed by the present invention, according to its first and second aspects.

A third aspect of the invention is concerned with the process of determining the colour of a surface once the reflected light has been captured and detected. Colour is the subjective human interpretation of a spectrum of electromagnetic radiation having wavelengths in the range between 400 and 700 nm. During normal daylight conditions, the eye has a photopic response and senses perceived light in three overlapping spectral regions roughly approximated by red, green and blue. The relative balance and total of the power in these three channels determine the chromaticity and intensity of the perception. The transformation between spectrum and colour is non-reciprocal; whereas a spectrum can always be uniquely transformed into a colour, the converse is not true; many different spectra (called metamers) can result in the same colour perception.

Moreover, although many colours, exemplified by those of the rainbow, can be represented by light of a single wavelength, there is also a range of non-spectral colours that can only be produced by blending different spectral components. For example, the colours mauve and purple are obtained by combining red and blue components, while the colour white must be produced by one of many component combinations, typically red, green and blue.

In order to measure the colour of a passive material such as clothing, the surface under test must be illuminated. The apparent colour of the surface under illumination is determined by the convolution of the spectrum of the illumination and the spectral reflectivity of the surface. Usually, people wish to appreciate the colour of a surface when it is illuminated by white light similar to sunlight, having a relatively even distribution of power over the visible spectrum.

The transformation of spectrum into colour co-ordinates as defined by the Commission Internationale de I'Eclairage (CIE) in 1931 is based on the average response measured over a large population. However, people have minor variations in response and will typically perceive the same scene slightly differently in terms of colour. This implies that, for this application, there is a level of accuracy of transformation beyond which there is little to be gained. According to the CIE tristimulus model, the colour is determined from the spectrum

S(λ) of the reflected light by analysing the spectrum into three channels using CIE defined spectral functions X(λ), Y(λ) and Z(λ) approximately corresponding to the colours red, green and blue. This results in three signals X, Y and Z where:

^• λ=700 . χ = f^λ=700S(λ)X(λ)d(λ)

Jλ=400

The lightness or darkness of a surface colour is characterized by the luminous reflectivity, that is the reflectivity of a surface when measured with Ulumination having a spectral distribution defined by the Y channel of the 1931 CIE model and so is obtained directly from. the value of "Y". The CIE chromaticity is represented by the normalized parameters x, y and z where: x = X/(X+Y+Z) y = Y/(X+Y+Z) z = Z/(X+Y+Z) The first step of determining the colour is therefore to determine the spectrum S(λ) of the reflected light. There are several methods of measuring the spectrum. To obtain a high spectral resolution, a large number of samples would be required. This type of measurement is typified by the spectral scanning technique.

In one such approach, the surface is illuminated with white light as described above and the reflected light spectrum is analysed with a graded filter wheel that scans a narrow spectral channel across the visible spectrum, resulting in a time varying signal where the time is related to the wavelength. In a similar approach, the illumination from the white light source is directed through the narrow spectral channel prior to being reflected from the surface under test. With a scanning measurement, the number of independent samples is equal to the ratio of the total half-power visible spectrum width, e.g. 200 nm, to the sampling width of the scanner. In these methods, the illuminating light source may instead be non-white but nevertheless cover the entire visible spectrum. Similarly, the sensor response may vary across the spectrum. In these cases, the transformation must be weighted by the spectrum of the illumination and the sensor response. The above techniques are ideal inasmuch as the entire spectral region of reflectance is measured with a high resolution and the transformation to colour co-ordinates can be exact. However, there are issues of complexity, cost and robustness associated with graded circular filters and the associated rotating mechanisms. However, a spectrum measurement with high wavelength resolution is not necessary for the accurate determination of colour. The human eye, that constitutes the basis of colour, uses just three spectral samples. These samples overlap to cover the visible spectrum between 400 nm arid 700 nm but they differ in their spacing and shape. In particular, the X and Y channels are relatively close together at about 600 nm and 550 nm, respectively, whereas the Z channel is relatively distant at about 450 nm. The provision of physical filters replicating the CIE spectral functions is not fully practical and would be expensive even to the limited extent that it may be practical. However, the operation of the human eye clearly indicates the feasibility of determining colour with as little as three spectral samples. Clearly, a colorimeter requiring only a few spectral samples would be less complex and expensive than one requiring a high-resolution measurement of the spectrum as it is compatible with a static design with no moving parts.

One sampling technique is to use a broadband (white) light source in combination with a set of optical bandpass filters that define the location and width of the spectral samples. The filters may be situated before or after the light reflects from the surface under test. The main cost of this approach is the provision of the optical system that typically includes beam-splitters as well as filters.

Alternatively, and less expensively, the spectral sampling can be implemented by illuminating the surface under test with a set of Light Emitting Diodes (LEDs), each having a different central wavelength, and collecting the reflected light using a single broadband photodetector such as a silicon photodiode. The above-mentioned US 3,910,701 (Henderson et al.) discloses a spectrometric instrument having a plurality of LEDs and at least one photodetector but which, in order to cover a relatively wide range, uses several interchangeable modules, each containing a different set of LEDs.

As the total half-power visible spectrum is about 200 nm in width and each LED based sample typically is about 40 nm in width, about five such spectral samples are required to cover the visible region. For example, US Patent 3,060,790 (Ward) discloses a colorimeter based on the use of five LED sourced sample wavelengths and suitable photosensors enabling chromaticity co-ordinates to be computed by simple electrical circuits. Disadvantageously, using multiple light sources and detectors increases complexity and cost.

Another disadvantage, identified in the discussion of prior art in US 5,838,451 (McCarthy), was the lack of availability of light sources with peak wavelengths in the region around 550 nm. According to McCarthy, prior art devices used multiple emitters and detectors with peak responses outside that region but whose response curves extended into it. McCarthy addressed this perceived deficiency by using newly-available LEDs with peak energies in the region of 530 nm. This enabled him to obtain coverage of the required spectrum with a set of only four LEDs, providing they had specific overlapping wavelength distributions. This is still not entirely satisfactory since LEDs that are readily available and inexpensive do not necessarily have the required wavelength distributions or values.

The second aspect of the present invention addresses this limitation by means of a sampling technique using light sources having relatively narrow wavelength distributions and which need not coincide with the peaks identified in the CIE model.

SUMMARY OF THE INVENTION

An object of the present invention in its various aspects is to at least mitigate the limitations of such known colorimeters, or at least provide an alternative.

According to a first aspect of the invention, there is provided a colorimeter for determining the colour of a surface which, when illuminated, will produce at least diffusely reflected light, comprising:

- a housing including a wall having an aperture therein, interior surfaces of the housing being adapted to absorb light impinging thereon, the aperture to be covered externally of the housing by the surface when the colorimeter is in use; - a plurality of light sources and a photodetector (PD) means disposed in the housing and generally facing the aperture, the light sources being each responsive to an electrical drive signal to emit light in a predetermined spectral segment of the visible spectrum,

- the light sources and the PD being mutually spaced apart and oriented so that substantially all of light from each light source that is specularly reflected by the surface is directed away from the PD and the PD will receive at least a portion of the diffusely reflected light from each light source and produce a corresponding electrical output signal having a plurality of values each representing the diffuse reflection characteristics of the surface for the spectral segment of the corresponding light source; the colorimeter further comprising: a drive unit for supplying said electrical drive signals to the light sources so as to cause them to emit said light and means for processing the corresponding electrical output signal from the photodetector means, the drive unit and processing means being so configured that the said value for any one of the light sources is distinct from the values for the other light sources, the processing means being configured to derive from the plurality of values of the electrical signal an indication as to the colour of the surface. Preferably, the processor is configured to select from a predetermined set of reference colours that which most closely approximates the colour of the surface.

The drive unit may be configured to supply said electrical drive signals sequentially so that only one light source emits light at any time, the processor means being configured to separate the corresponding sequential values. The sequence need not recur. Preferably, the light sources are arranged to provide light in predetermined spectral segments of the visible spectrum that are different and spaced apart across the visible spectrum such that the spacing at the red end of the visible spectrum is about 50 nm or less and the spacing at the blue end of the spectrum is about 100 nm or less. Where the minimum number of spectral samples is three, they should preferably be located al positions at or near to the spectral locations corresponding to the peak responses of the three types of eye cone, that is approximately 600, 550 and 450 nm, it being appreciated however that there is no great criticality of proximity to these optima and that deviations may degrade the accuracy of colour determination so little as to be of no significant consequence when used within the application for which the invention was generated. Similarly, increases in accuracy deriving from the use of additional spectral samples will likely result in a minor or insignificant consequence when used within the application for which the invention was generated

In preferred embodiments, therefore, at least three spectral samples are used that need not be overlapping or contiguous; but are taken at such different wavelengths that they permit accurate determination of the colour, as defined in the CIE 1931 model, of most common reflective object surfaces. Preferably the electrical drive signals are each modulated with a relatively low frequency signal differing from the typical power supply, frequencies of 50 Hz and 60 Hz, and their low order harmonics. A modulation frequency of 1 kHz is convenient.

According to a second aspect of the invention, there is provided a colorimeter sensor unit for use in a colorimeter for determining the colour of a surface which, when illuminated, will produce at least diffusely reflected light, the sensor unit comprising:

- a housing including a wall having an aperture therein, interior surfaces of the housing being adapted to absorb light impinging thereon, the aperture to be covered externally of the housing by the surface when the colorimeter is in use; - a plurality of light sources and a photodetector (PD) means disposed in the housing and generally facing the aperture, the light sources being each responsive to an electrical drive signal to emit light in a predetermined spectral segment of the visible spectrum,

- the light sources and the PD being mutually spaced apart and oriented so that substantially all of light from each light source that is specularly reflected by the surface is directed away from the PD and the PD will receive at least a portion of the diffusely reflected light from each light source and produce a corresponding electrical output signal having a plurality of values each representing the diffuse reflection characteristics of the surface for the spectral segment of the corresponding light source.

As .discussed hereinbefore, it is desirable, especially in a hand-held colorimeter, to process the colour sample values and determine the colour without involving considerable processing and without necessarily using light sources whose wavelengths coincide with the spectral peaks of the human eye response.

According to a third aspect of the invention, there is provided a process for determining luminous reflectivity and chromaticity values from sample values obtained by irradiating a surface with light, receiving at least a portion of such light reflected from the surface and producing a corresponding electrical signal having a plurality of sample values each representing the reflection characteristics of the surface for a corresponding spectral segment of the visible spectrum, the process comprising the steps of:

- calculating the luminous reflectivity value Y as a weighted average of the sample values as follows:

Y = Σ_j A_JR_j where Y is the luminous reflectivity value; _j , j = 1 to n are the sample values, n being the number of spectrum samples; and

A,-, j = 1 to n are weighting coefficients which depend on the location and width of the spectrum samples, and

- calculating the three CIE chromaticity values Pj , i = 1 to 3, as a matrix transform of normalized sample values as follows:

P_i = ∑_j A_ij r_j where Pj , i = 1 to 3 are the three CIE chromaticity values; r- = R_j/∑_j R_j, j = 1 to n are the normalized sample values; and Ay , i = 1 to 3, j = 1 to n, is a colour transform matrix having matrix elements that depend on the location and width of the spectrum samples. In the field of telecommunications, Shannon's Sampling Theorem [Communication in the Presence of Noise, C.E. Shannon, Proc. I.R.E., Vol. 37, No.l, Jan. 1949] is well known and states that a band-limited, time- varying signal can be accurately reconstructed from a series or set of short periodic samples. The minimum sampling rate, also known as the Nyquist rate, is equal to twice the maximum frequency of the signal spectrum. More recently, the theory of wavelets has advanced the art of reconstructing a function from samples.

In an analogous fashion, the optical spectrum can be accurately reconstructed from a set of narrow spectral samples. In the analogy, time corresponds to wavelength and signal level corresponds to the product of spectral density and spectral sample width. The condition for an accurate reconstruction is that the wavelength or spectral spacing between samples must be no more than half the spectral width encompassing a full change cycle of spectral reflectivity, e.g., from high to low and back to high. With this restriction, the samples can be very narrow leaving a preponderance of unsampled spectrum extent without prejudice to the accuracy of the reconstruction.

Because the human eye has only three spectral sample locations, it is similarly restricted in its ability to discern any spectral change cycles that are narrower than twice the sample spacing. Moreover, as the sample spacing is about 100 nm between blue and green but only 50 nm between green and red, it can discern shorter cyclic changes in the red end of the visible spectrum than in the blue end.

Because the spectral sample widths of the human eye are so broad, the eye can capture and discern light of any wavelength between about 400 and 700 nm. This is useful as some light sources have narrow spectral widths as exemplified by Hght from a laser or from a specific radius of a rainbow. If the eye spectral sample width were very narrow, some colours of the rainbow would be invisible. However, for the comprehension of the reflective color-metric properties of object surfaces commonly encountered in everyday life, the comprehensive spectral capability of the human eye is not essential. The spectrum of reflectivity of such surfaces does not normally comprehend narrow bands of high or low reflectivity but instead presents relatively broad spectral bands that form a spectrum showing relatively slow changes with respect to wavelength.

According to a fourth aspect of the invention, there is provided a process for use in a colorimeter having stored color-metric coordinates of a group of reference surfaces, the process using luminous reflectivity and chromaticity values measured for a colour to be determined, comprising the steps of:

- determining a best fit between:

- the measured luminous reflectivity and chromaticity values; and

- corresponding luminous reflectivity and chromaticity values stored for each of the known reference surfaces;

- the best fit being determined using a least mean square difference criterion as follows:

S_m ² = wΔY_m ²+ΔH_ra ²

- the least value of S_m ² for m = 1 to k, being deemed as the best fit. where S_m ² is a mean square difference criterion; w is a weighting factor 0.1 < w < 10;

Δ Y_ra ² = (Y - Y-_n)² is a reflectivity mean square difference criterion;

ΔH.-²

is a chromaticity mean square difference criterion; and k is the number of the known reference surfaces. Preferably, w is equal to unity. Advantageously, the transform process according to the third aspect and the selection process according to the fourth aspect may be employed together, one to determine the colorimetric coordinates for the unknown surface and the other to use those coordinates to select the reference colour which most closely approximates it. Moreover, either or both of them may be employed in the colorimeter of the first aspect.

It is possible that the user may be hard of hearing or use the colorimeter in a noisy environment, which may make an aural output substantially inaudible while the colorimeter remains upon the surface being measured. This situation is addressed by a fifth aspect of the invention according to which there is provided a colorimeter having light source means for illuminating a surface whose colour is to be determined, photodetector means for receiving light from the surface and producing a corresponding electrical signal, means for processing the electrical signal and providing an aural indication of the colour so determined, and user- operable switch means for delaying the outputting of such aural indication.

Such an arrangement allows the user to remove the colorimeter from the surface and hold it closer to one ear, say, before operating the switch means to initiate the colour indication, e.g., by aural announcement.

Various features, advantages and objects of the invention will become apparent from the following description of a preferred embodiment which is described by way of example only with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a simplified perspective view of a colorimeter; Figure 2 is a schematic block diagram of electronic circuitry in the colorimeter; Figure 3 illustrates sequential drive signals for driving LEDs in the colorimeter; Figure 4 is a block schematic diagram of analog processing circuitry for a photodetector of the colorimeter;

Figures 5A, 5B and 5C are partial cross-sectional views of a sensor unit of the colorimeter which incorporate ray diagrams;

Figure 6 illustrates a flowchart depicting a colour determination process carried out by a processor in the colorimeter; and Figure 7 is a flowchart depicting a "best fit" process for selecting a reference colour that is closest to the unknown colour.

DETAIL DESCRIPTION OF A PREFERRED EMBODIMENT Figure 1 is a perspective view of a colorimeter 101 comprising a palm-sized casing

102, shown with one end removed, and a base 103. A sensor unit 104 is mounted upon the base 103 at one end and a printed circuit board (PCB) 105 is mounted upon the base 103 at ^' its opposite end.

The sensor unit 102 comprises a cylindrical housing 106 closed at one end by an integral end wall 107 and at its opposite end by a second printed circuit board 108 which has a plinth portion 108 A extending beyond the housing 106 towards the first printed circuit board 105, to which it is connected by a cable 109. The second printed circuit board 108 seals that end of the housing 106 and mounts the sensor unit 102 upon the base 103.

The length of the housing 106 is such that the end wall 107 abuts the inner surface of the casing 102. A central aperture 110 in the end wall 107 is sealed by a transparent window element 111. The casing 102 has a hole 112 which is concentric with, but has a diameter much larger than, the aperture 110, so that, when a user places the colorimeter 101 onto a surface 113 (see Figure 5A), e.g. a piece of clothing, whose the colour is to be determined, the aperture 110 is covered by the surface 113. The sealing of the housing 106 by the window 111 and PCB 108 A protects against the ingress of contaminants, such as dust and moisture. A photodetector (PD) 114 is mounted on the PCB 108A in the centre of the sensor housing 106 and opposite the aperture 110. Thus, the aperture 110, hole 112 and PD 114 are aligned on the cylindrical axis CA (see Figure 5 A) of housing 106. Three light emitting diodes (LEDs) 115, 116 and 117 mounted within housing 106 upon the second PCB 108 are spaced from each other and equidistant from the PD 114. The LEDs 115, 116 and 117 and the PD 114 are of the surface mount device type without any lenses. The LEDs 115, 116, and 117, and the PD 114 are semiconductor devices of planar construction with wide angle emission and detection characteristics, respectively. A low cylindrical shield 118 surrounding the PD 114 prevents light from any of the LEDs from directly impinging upon the PD 114. The interior of the housing 106 is blackened to rninimize any stray reflections from impinging on the PD 114.

The LEDs 115, 116 and 117 emit orange, green and blue light, respectively, with spectral maxima of their emission wavelengths at approximately 610 nm (orange), 555nm (green), and 445nm (blue), respectively. The PD 114 is a sihcon photodiode capable of converting Hght having the emission wavelengths of the LEDs 115, 116 and 117 into an electrical signal, i.e. a photocurrent, A temperature sensor 119 for measuring ambient temperature is similarly mounted within the housing 106 upon PCB 108, the temperature information being used for making software-based temperature compensations as will be described hereinafter.

Analog electronic circuitry 120 for processing the electrical signal from the PD 114 is mounted upon the PCB 108. It could be provided on the underside providing that no through holes compromise the seal. A user-operable pushbutton on/off switch 121 is mounted to the base 103 with the button 121 A protruding through the base 103 for access by the user. A loudspeaker 122 is attached by its frame to the inner surface of the casing 102 at a position adjacent the first PCB 105 and is connected thereto by a cable 123.

Referring also to Figure 2, the first PCB 105 carries a battery 124, a voltage converter unit 125, a PMOS switch 126, an NMOS switch 127, a voice/speech synthesizer 128. a speaker drive amplifier 129, and a microcomputer 130. As shown in Figure 2, the analog circuitry 120 comprises three precision drive amplifiers 131, 132 and 133 for driving LEDs 115, 116 and 118, respectively, and signal processing circuit 134 for processing the electrical signal (current) from photodetector 114. While all of the analog processing circuit may be located on PCB 108, as shown in Figure 1, the LED precision drive amplifiers 131, 132 and 133 could be provided on PCB 108, if more convenient.

The microcomputer 130 includes a colorimetric signal processor 135 which receives signals from the PD analog processing circuit 134 and the temperature sensor 119 and, through a controller 136, supplies various control signals, such as clock signals and reference signals, to the other components in a conventional way. It also monitors the state of pushbutton 121, as wiU be described more specifically later. The colorimetric signal processor 135 processes signals from the PD analog signal processor 134 to determine the colour of the surface 113 and supplies a corresponding colour identification signal to voice synthesizer 128 which supplies a corresponding signal to speaker drive amplifier 129 causing loudspeaker 122 to enunciate the corresponding colour. The PMOS switch 126 connects the voltage converter unit 125 to the positive teπtiinal of battery 124, the negative terminal of which is grounded. When closed, switches 121 and 127, which are connected in parallel, connect the (control) gate of PMOS switch 126 to ground, causing switch 126 to close and connect the voltage converter 125 to battery 124. The voltage converter 125 then supplies the various other components of the colorimeter at whatever voltage is appropriate.

When the colorimeter 101 is not in use, the PMOS switch 126 is open and aU components are off. To take a reading, the user places the colorimeter 101 onto the surface 113 so that the aperture 110 is obscured and depresses the pushbutton switch 121, causing PMOS switch 126 to close. The colorimetric processor 135, by way of the controller 136 and amplifiers 131, 132 and 133, causes the LEDs 115, 116 and 117 to be energized sequentially, as shown in Figure 3, to illuminate the surface 113, via the aperture 110. As illustrated in Figures 5 A, 5B and 5C for LED 115 only, an incident ray L from each of the LEDs 115, 116 and 117 reflected by the surface 113 will generate specular and diffuse reflections SRL, DRL. Each of the three diffuse reflections DRL constitutes a spectral sample of the light emitted by the corresponding one of the LEDs 115, 116 and 117 and the three spectral samples will be used to determine the colour of the surface 113 as will be described hereinafter.

Although, for ease of depiction and description, Figures 5 A, 5B and 5C illustrate the ray diagram for only LED 115, it will be appreciated that it applies analogously to LEDs 116 and 117. Thus, Figure 5A illustrates the Lambertian emission profile or polar diagram of the LED 115. Referring to Figure 5B, PD 114 is in the diffuse reflection lobe DRL but out of the way of the specular reflections lobe SRL. To ensure that specular reflections are directed away from the PD 114, the aperture 110 has edges such that the surface 113 covering the aperture 110 is substantially parallel to the PCB 108, and its lateral dimension, specificaUy its diameter D, is less than the distance between each of the LEDs 115, 116 and 117 and the PD 114. Specular reflections from the two surfaces of the window element 111 are also directed away from the PD 114, so there is no need to provide anti-reflection coatings on the two surfaces of window 111.

Referring to Figure 5C, which illustrates the limiting case where a specular reflection just misses .the PD, an upper bound for the diameter D of the aperture 110 is: D < S - R_L - R_S (4) where: S is the centre-to-centre distance between each of the LEDs 115, 116, 117 and the PD 114;

R_L is the radius of each of the LEDs 115, 116, 117; and R_s is the radius of the active area of PD 114. In this boundary case, ray L_! from an extremity of the active area of the LED 115 causes specular reflection

to miss an extremity of the active area of the PD 114. Diffuse reflections DRLi still impinge on PD 114. It wUl be noted that the specular reflection SRL_X is shown as a narrow lobe since there will be some dispersion as a result of roughness of the surface 113. It should be appreciated that all of the lobe should avoid the PD 114. The angle of incidence of the light Lj from LED 115 on the surface 113 is nominally

45 degrees from the normal to the surface 113 (and hence the cylindrical axis CA) and the collection angle of the PD 114 to the surface 113 also is nominally normal. As indicated before, the same applies to the light from LEDs 116 and 117. These angular arrangements are consistent with the achievement of a transfer efficiency of the light power compatible with satisfactory operation of the equipment.

Typically, for a practical colorimeter 101, the circular aperture 110 would have a diameter of 7 mm and be spaced from the PD 114 by about 10 mm, and the distance S would also be about 10 mm. These dimensions are insensitive to any smaU variations, because of the wide angle emission characteristics of the LEDs 115, 116 and 117, and the wide field of view of the PD 114.

The relatively large size of the aperture 110, in conjunction with the wide angle emission characteristics of the LEDs 115, 116 and 117, causes substantially even illumination of the surface 113, thereby ensuring that the colour determination will not be unduly affected by locahzed concentrations or patterns. The diffuse reflections DRL can be considered to be samples of the orange, green, and blue light from the LEDs 115, 116 and 117, respectively, which impinges upon PD 114, each sample being over the band of wavelengths of the corresponding LED. The corresponding electrical signal generated by PD 114 contains a corresponding set of three sample values.

The driving of the LEDs 115, 116 and 117, and the processing of the corresponding diffuse light samples, will now be described in more detail with reference also to Figures.3 arid 4. In order to separate the orange, green and blue spectral samples, a sequence of three successive pulses, each modulated by a 1 kHz. subcarrier, is apphed to the three LEDs, one pulse to each, by LED drivers 131, 132 and 133 which drive the LEDs 115, 116, and 117 sequentially, each with a subcarrier modulated burst of current of the same duration, typicaUy

100 ms. The subcarrier modulation is a square wave signal having a nominal frequency of 1 kHz.

The corresponding electrical signal generated by PD 114 will constitute three similar sequential pulses, comprising samples of the diffuse reflection of the orange, green, and blue Hght, respectively.

The PD analog signal processor 134 demodulates the electrical signal from the PD 114 to produce corresponding orange, green, and blue sample values. As shown in Figure 4, the

PD analog signal processor 134 comprises a preamplifier 137, a phase synchronous demodulator 138, a low-pass filter 139, and a DC amplifier 140. Using a square wave reference signal received from the microprocessor controUer 136 via line 141, the phase synchronous demodulator 140 demodulates the subcarrier modulated samples and passes the demodulated samples to DC amphfier 140 via low-pass filter 139, which limits the noise. The low-pass filter bandwidth is less than 10 Hz and is just sufficient to aUow the signal to substantiaUy reach its steady-state value by the end of the burst or pulse. This enhances the signal-to-noise ratio, further facilitating the use of low-power LEDs. DC amplifier 140 conditions the signal before supplying it to the microprocessor 130 (Figure 2). Within the microprocessor 130, which includes an analog-to-digital converter, the three sample values are segregated and digitized, then processed by the colorimetric signal processor

135 to determine the CIE colorimetric coordinates corresponding to the colour of the surface

113.

The light outputs of the LEDs 115, 116 and 117 drop by about 1 per cent per degree Celsius at a fixed drive current, so the processor 135 apphes a temperature correction to the sample values prior to the calculation of the luminous reflectivity and chromaticity values. The processor 135 derives the temperature correction using a digital signal representing ambient temperature measured by the temperature sensor 119. Such temperature correction allows the LEDs 115, 116 and 117 to be driven with a simple fixed constant current source circuit.

The colorimetric processor 135 uses a transform algorithm to determine the CΪE colorimetric coordinates of the colour of surface 113 from the three sample values, as will be described in detail later, and then compares the measured luminous reflectivity and chromaticity values with those of a set of reference coloured surfaces, previously stored in the microprocessor 130, and determines the "best fit" as being the colour of the surface 113.

The stored colorimetric coordinates comprise a set of luminous reflectivity and chromaticity coordinates for each of the reference coloured surfaces of a suitable commercially available colour check card, each set of coordinates being associated in the store with a unique identifier. A suitable such card is the Gretag Macbeth Color Checker (GMCC) Color Rendition Chart, which has 24 squares each comprising one of the known reference coloured surfaces. Five of the squares are grey scale. (It will be appreciated that the user may scan black, white or grey articles.)

The speech synthesizer 128 stores the names of the 24 colours in association with their respective unique identifiers. When the microprocessor 130 has determined the best fit between the colorimetric coordinates measured for the surface 113 and the colorimetric coordinates of the reference coloured surfaces, it supplies the corresponding unique identifier to the speech synthesizer 128 which uses it to select and enunciate the name of the corresponding reference colour as being the colour of the surface 113.

In addition to controlling the phase synchronization of the subcarrier modulation and demodulation circuits, the controller unit 136 controls the time synchronization of the three modulated pulses and the sampling of the corresponding photocurrents. Typical operation of the colorimeter 101 is illustrated by the flowchart shown in Figure

6. When the user holds the colorimeter 101 on a surface 113 whose colour is to be determined and presses the pushbutton 121, the colorimeter 101 powers up. In step 6.01, the microprocessor 130 confirms that the pushbutton is depressed, and closes NMOS switch 127 to maintain the supply to the circuitry even if the pushbutton has been closed only momentarily. In step 6.02, the equipment powers up, and, in steps 6.03 to 6.08 the orange, green, and blue LEDs 115, 116 and 117 are energized sequentially and the corresponding measured orange, green, and blue spectrum sample values stored. In step 6.09, the microprocessor 130 applies the temperature correction to the values and, in step 6.10, calculates the luminous reflectivity and chromaticity values using the temperature corrected sample values. In step 6.11, it determines the best fit between the calculated luminous reflectivity and chromaticity values and the sets stored for the 24 reference colours, deeming that to be the colour of the surface 113.

The processor does not supply the result to the speech synthesizer 128 immediately, however. Rather, in step 6.12, the microprocessor 130 determines whether or not the pushbutton 121 has been released. If it has not, the microprocessor 130 "poUs" it at intervals of, for example, one quarter second until it has been released. Only then does the microprocessor 130 pass the selected identifier to the speech synthesizer 128 which, in step

6.13, announces the colour to the user. Finally, in step 6.15, the microprocessor 130 releases

NMOS switch 127 thereby opening PMOS switch 126, whereupon the equipment powers down.

In order to determine the colour of the surface 113 from the three sample values, the colorimetric processor 135 performs two processes, namely a transformation process to determine luminous reflectivity and chromaticity from the measured spectrum sample values at the wavelength bands of the three LEDs and then a mapping process to determine which of the stored sets of reference colour coordinates most closely corresponds.

Thus, the processor 135 calculates luminous reflectivity value Y as a weighted average of the three sample values at the three LED wavelength bands as:

Y - ΣJ AJRJ (5) where Y is the luminous reflectivity of the surface 113 being measured;

R_j , j = 1 to 3 are the three spectrum sample values; and A_j , j = 1 to 3 are three weighting coefficients which depend upon the set of three LEDs 15, 116 and 117 used in colorimeter 101.

The three CIE chromaticity values P; , i = 1 to 3, are calculated from the measured spectrum sample values as follows:

where Pj , i = 1 to 3, are the three CIE chromaticity values; r_j = R_j/∑_jR_j, j = 1 to 3, are the three normalized spectrum sample values; and Aj_j , i = 1 to 3, j = 1 to 3, is a colour transform matrix having matrix elements that are dependent on the set of the three LEDs 115, 116 and 117. The weighting coefficients A_j and Ay will have been stored in the processor unit 135, having been determined previously, conveniently empirically, as will be described in more detail later.

Once the three spectrum sample values have been processed, processor 135 determines which of the 24 colours is closest to the colour of the surface 113. As illustrated in Figure 7, it does so by calculating the best fit between the colorimetric values determined for the surface

113 with the stored colorimetric values of the 24 reference coloured surfaces. The best fit is found using a least mean square algorithm involving both reflectivity and chromaticity.

The 24 reference colours selected from the GMCC chart are characterized by luminous reflectivity values Y_m and chromaticity values P--_J where "m" takes the values from 1 to 24 and "i" the values from 1 to 3.

Where a measured luminous reflectivity value from the surface 113 is "Y", its mean square separation from the stored value Y_m is given by:

ΔY_ra ² = (Y - Y ² (7) where ΔY_ra ² can take values between zero and unity.

Similarly where the measured set of chromaticity values of the surface 113 are P_; for values of "i" from 1 to 3, the mean square separation from the colour with stored parameters P.,,_! is given by:

ΔH_ra ² = Σ_jCPj - P_mi)²/2 (8) where Δϊ j² can similarly take values from zero to unity.

The composition mean square separation parameter is here defined as:

S^ wΔY^ + ΔIL.² (9) where "w" is a weighting parameter, the preferred value of which is unity. S_m ² therefore can take values between zero and "1 + w"

The value of S_m ² is calculated for each value of "m" and the smallest value obtained identifies the stored colour "m" that corresponds to the best fit to the colour of the surface 113 covering the aperture 110. The mapping or selection process, as shown in Figure 7, begins, in step 7.01, with the processor 135 setting m = 1 and S^ = 1 + w. S^ is the updated value of minimum separation found so far during the selection process and C is the number associated with the best fit colour. Both occupy stores that can be updated. (S(m) is the same as S_m ² as is used for convenience in the diagram.)

In step 7.02, the processor 135 calculates S(m) for the colour m and in step 7.03 determines whether it is greater than the stored value of S-^. If it is, step 7.04 updates the stored values of S^ and C and step 7.05 increments the value of m. If the calculated value for S(m) is not greater than S^ m is incremented.

In step 7.06, the processor 135 determines whether or not the calculation has been done for all 24 reference colours. If it has not, loop 7.07 returns to step 7.02 and the calculations are repeated for the next colour. When the measured coordinates have been compared with those of all 24 colours, step 7.08 determines the colour identified as C to be the best fit and sends the identifier C to the speech synthesizer unit 128, which uses it to select the corresponding name of the reference colour, synthesizes the name as an analog signal, and supplies the analog signal via speaker drive amplifier 129 to loudspeaker 122 for audible announcement to the user. The empirical determination of the weighting coefficients A_j and A_y, which is done during the design process, may use this same algorithm and a similar colorimeter, i.e., with LEDs and a photodiode having the same characteristics and with the same geometry as those used in the production colorimeter 101, to measure each colour of the reference chart in turn. (While it is convenient to use the same used in the production models, it is not essential). The measurements from the three channels, i. e. , the colorimetric coordinate set for each reference coloured surface derived from the corresponding three sample values, for each reference colour are entered into a Microsoft Excel spreadsheet that contains the transform formulae with the coefficients A_j and Ay as variables. Using the standard best fit procedure which is available in Excel, these variables are optimized such that the colorimetric coordinates calculated from the measurements are as close as possible to the colorimetric coordinates for the reference colours, as supplied by the manufacturer of the reference chart. During this empirical determination process, the 5 grey scale squares are used to caHbrate the gain of each channel.

The resulting numerical values of A,-, j = 1 to 3, and Ay, i = 1 to 3, j = 1 to 3 are stored in the processor 135 and used when the colorimeter is actually in use. It is envisaged that the same weighting coefficients will be used for all production versions of the colorimeter, on the assumption that the variations between units wiU be so smaU that the quaUty of the colour determination will not be degraded unduly. Nevertheless, greater accuracy could be obtained by calibrating each colorimeter with the colour chart, during production, to select the best weighting coefficients for that particular colorimeter.

Once loaded with its weighting coefficients and reference colour information, but before use, the colorimeter 101 will be factory calibrated to correct for offsets and gain differences in the three colour channels. This calibration is a two step process. In the first caHbration step, the colorimeter 101 is operated with the aperture 110 left uncovered, so there are no reflections from an outside surface 113. The three spectrum sample values so acquired correspond to the offset values for each of the LED wavelength bands (orange, green, blue) and are primarily generated by stray reflections from the interior surfaces of the housing 106, both surfaces of the window 111 and, to a lesser extent, by the electronic circuitry. The offset values are stored in the microprocessor 130 for use in subsequent measurements to generate offset error compensated spectrum sample values.

In the second caHbration step, the aperture 110 is covered with a reference white surface with a defined high reflectivity and operated again. The three sample values now acquired will be higher than those acquired in the first calibration step. The differences correspond to the channel gain values for each of the three LED wavelength bands and the defined high reflectivity. The corresponding slope values also are stored in the microprocessor 130 for use in subsequent measurements to correct for differences in the channel gains according to the LED wavelength band being sampled and thereby generate gain calibrated spectrum sample values. This ensures that the noise and quantization levels of the electrical drive signals are about equal for each of the LED wavelength bands. If desired, the first caHbration step could be performed with the aperture 110 covered by a black surface having a defined low reflectivity rather than left uncovered. The second step would again be performed with the aperture 110 covered with a white surface having a high defined reflectivity. The 'black' measurement would yield spectrum sample values corresponding to the offset errors and the defined low reflectivity. The 'white' measurements would yield spectrum sample values corresponding to the offset errors and the defined high reflectivity. Using both 'black' and 'white' measurements, the offset errors and the gain values corresponding to the difference between the defined low and high reflectivities can be calculated and stored as described above. (The same factory calibration may be performed for colorimeters having three or more LEDs).

It should also be appreciated that the transformation algorithm for deterr ning colorimetric coordinates for surface 113, and the "best fit" algorithm for comparing measured colorimetric coordinates with those of a set of reference coloured surfaces, conveniently taken from a commercially available colour checker chart, could be used independently of each other and also could be employed with colorimeters which did not have the sensor unit described herein. Conversely, the sensor unit described herein could be used in colorimeters employing different algorithms to process the reflected light and determine the colour of the surface being scanned.

Various modifications may be made to the above-described embodiment without departing from the scope of the invention, some of which wiU now be described.

Thus, more than three LEDs may be used in the colorimeter 101 for higher precision in the colour determination for more stringent applications. Generally, a similar geometry as that of colorimeter 101 may be used provided that the LEDs are disposed so that diffuse reflections are directed to the PD 114 and specular reflections are directed away from the PD

114.

The factory caHbration with more than three LEDs will have to be adapted. Thus, assuming n LEDs, n>3, the factory caHbration in terms of offset value and gain value calibration would have to be performed for each of the n spectrum sample values rather than three as previously described.

Moreover, the temperature correction of the n spectrum sample values due to the drop of the light output of the n LEDs with increasing temperature may be similarly performed as described above.

Furthermore, the luminous reflectivity value Y would be calculated as a weighted average of the n spectrum sample values at the n LED wavelengths.

Y = Σ_jA_jR_j (10) wherein Y is the luminous reflectivity of the surface covering the aperture 5; R_j , j = 1 to n are the n spectrum sample values; and A,-, j = 1 to n are n weighting coefficients which depend on the set of n LEDs used in colorimeter 101. The three CIE chromaticity values Pj , i = 1 to 3, would be calculated from the n spectrum sample values as foHows. . P^ ∑_jAy (11) wherein Pj , i = 1 to 3, are the three CIE chromaticity values; r,- = R_j/Σ_jR_j, j = 1 to n, are the n normalized spectrum sample values; and Ay , i = 1 to 3, j = 1 to n, is a colour transform matrix having matrix elements that are dependent on the n spectral samples from the LEDs used in colorimeter 101.

The determination of the numerical values of A_j, j = 1 to n, and A_y, i = 1 to 3, j = 1 to n would follow the same procedure as the one described for n = 3 hereinbefore. The Microsoft ^' Excel spreadsheet would require a small modification to factor in the n spectrum sample values to produce the n numerical values of _j and the 3n numerical values of Ay. The determination of the colour of the surface 113 from its luminous reflectivity and chromaticity values using a colorimeter 101 having n LEDs with n>3, may use the identical best fit method as described hereinbefore.

The colorimeter could use other non-visual forms of communicating the determined colour to a user, such as a Braille based output, either in addition to, or instead of, the audible announcement.

The wavelengths and spectral widths of the LEDs shift slightly with temperature in a range from 0.02nm/degree C to 0.1 nm degree C. This dependency may be compensated for by making the numerical values of A_j and Ay in equations 10 and 11, respectively, temperature dependent. To minimise noise, the PD 114 may be operated at zero bias.

The preamplifier 137 in analog signal processing circuitry 134 may feature a low-noise transimpedance pre-amplifier, conveniently mounted on the underside of PCB 108 opposite to the location of the PD 114. The interconnections between the PD 114 and the pre-amplifier 137 would then be very short, minimizing sensitivity to electromagnetic interference (EMI). A pre-amplifier 137 having a very low noise level enables satisfactory operation at low

LED light levels. This can be achieved with a FET front end and a high value transimpedance, typicaHy 10 Megohm. A suitable device is the LMC6484 operational amplifier manufactured by National Semiconductor Corp. This enables the use of inexpensive, low-power LEDs operated well below their ratings to ensure long life, stability and reliabiHty.

The synchronous demodulator 138 may multiply its input signal synchronously by +1 and -1, thereby using the entire signal rather than part of the signal.

It will be seen from Figure 3 that a modulation depth of 95% is used for greatly reducing the voltage swing on the LEDs 115, 116 and 117, thereby minimizing electromagnetic interference (EMI) to the sensitive preamplifier. It should be noted however that, instead of the LEDs 115, 116 and 117 being modulated from a baseline upwards, they may be modulated from a mid-level to opposite sides of it.

Although subcarrier modulation is not essential to obtain the spectrum sample values, the use of subcarrier modulation substantially eliminates the effects of interfering natural or artificial background light and DC drift of the preamplifier 137. This permits the colorimeter 101 to operate at very low signal levels, allowing the LEDs 115, 116, and 117 to be operated at low drive currents in the order of 1 mA, and removes the need for any Hght guide or focussing optics to conserve light energy by increasing the light transfer efficiency.

Although the above-described sequential driving of the three Hght sources is preferred, it would be possible to energize them simultaneously providing their respective sample values can be discriminated. Thus, a Frequency-Division Multiplex (FDM), or a Code-Division Multiplex (CDM) format may be used. In the FDM format, the three (or more) LEDs would be driven simultaneously but each at a different subcarrier modulation frequency. An FDM Electrical Signal Processor would demultiplex the three (or more) spectral samples based on the individual subcarrier modulation frequencies.

In the CDM format too, the three (or more) LEDs would be driven simultaneously, but each would be continuously amplitude modulated with a respective one of a set of mutuaUy orthogonal codes. A CDM Electrical Signal Processor would demultiplex the three (or more) spectral samples based on the individual orthogonal codes.

The FDM and CDM formats advantageously would provide an AC signal component such that no DC detection response would be required; moreover the AC component should be spectraUy separated from those AC signals caused by artificial lighting, usually powered by mains AC suppHes of 50 Hz or 60 Hz and generating spectral components of their respective harmonics.

The FDM and CDM formats require the frequency or code to demodulate the required component. For speed, a plurality of demodulators could be used in paraUel rather than having to use a single demodulator in a serial mode.

It wiU be appreciated that light sources other than LEDs could be used, such as electroluminescent devices.

Advantageously, colorimeters embodying the present invention are inexpensive and rugged because of the total absence of optical filters, light pipes/guides, lenses, mirrors, reflector cones, or other such optical elements.

The specular reflection from real surfaces that are not perfectly smooth is such that the angles of reflection have a narrow distribution of angles surrounding the nominal angle of reflection that equals the angle of incidence. For this reason, the above-described colorimeter, wherein 45 degrees separate the diffuse reflections from the specular reflections, advantageously provides an ample margin to ensure that no specular components are coUected.

The withholding of the announcement of the colour while the pushbutton 121 is still depressed facilitates use in a noisy environment or by a user who is hard of hearing, since it allows the user to hold the colorimeter close to one ear before releasing the pushbutton and listening to the announcement of the name of the colour. It should be noted that this advantageous feature could be employed with other hand-held sensors, including colorimeters, which do not have the specific sensor unit construction described herein or even employ the same algorithms as the colorimeter 101 described herein. Thus, this aspect of the invention is not limited to colorimeters at all, but could be applied to other devices for determining features other than colour, such as hand-held optical character recognition devices for scanning text and providing an aural output.

Although embodiments of the invention are of benefit for determining the colour of garments by the blind or colour-blind, it should be appreciated that embodiments of the invention may be used for other purposes, for example domestic consumer purposes (paint, furnishings, etc.) or as a colour learning aid for young children.

Claims

1. A colorimeter for determining the colour of a surface which, when muminated, will produce at least diffusely reflected light, comprising: - a housing (106) including a wall having an aperture (110) therein, interior surfaces of the housing being adapted to absorb light impinging thereon, the aperture to be covered externally of the housing by the surface when the colorimeter is in use;

- a pIuraHty of light sources (115, 116, 117), and a photodetector (PD) means (114) disposed in the housing and generally facing the aperture, the light sources being each responsive to an electrical drive signal to emit light in a predetermined spectral segment of the visible spectrum,

- the light sources and the PD being mutually spaced apart and oriented so that substantially aU of Hght from each light source that is specularly reflected by the surface is directed away from the PD and the PD will receive at least a portion of the diffusely reflected Hght from each light source and produce a corresponding electrical output signal having a plurality of values each representing the diffuse reflection characteristics of the surface for the spectral segment of the corresponding light source; the colorimeter further comprising: a drive unit for supplying said electrical drive signals to the light sources so as to cause them to emit said light and means for processing the corresponding electrical output signal from the photodetector means, the drive unit and processing means being so configured that said value for any one of the light sources is distinct from the values for the other light sources, the processing means being configured to derive from the plurality of values of the electrical signal an indication as to the colour of the surface.

2. A colorimeter according to claim 1, wherein the light sources are energized sequentially.

3. A colorimeter according to claim 1, wherein: - the aperture (106) lies substantially in an aperture plane; - emission surfaces of the Hght sources and a detection surface of the PD are generally in a plane parallel to the aperture plane at a pre-determined distance from the aperture plane;

- the PD is disposed opposite the aperture (106); - the Hght sources are substantiaUy equidistant from the PD; and

- a maximum lateral dimension of the aperture (106) is smaUer than the distance between each of the light sources and the PD.

4. A colorimeter according to claim 2, wherein the light sources are so disposed relative to the aperture that light from each of the light sources will be incident upon a said surface covering the aperture (106) at an angle of approximately 45 degrees from the normal to the surface.

5. A colorimeter according to any one of claims 1 to 4, wherein the plurality of light sources (115, 116, 117) having emission wavelengths corresponding to the colours orange, green, and blue, respectively.

6. A colorimeter according to claim 5, wherein the emission wavelengths have spectral maxima at approximately 610 nanometers, 555 nanometers, and 450 nanometers, respectively.

7. A colorimeter according to claim 5 or 6, wherein the light sources comprise LEDs, the LEDs and the PD being planar semiconductor devices, the LEDs (115, 116, 117) having wide angle emission characteristics and the PD (114) having a wide angle field of view.

8. A colorimeter according to claim 7, wherein the LEDs (115, 116, 117) and the PD (114) are surface mount devices surface mounted on a printed circuit board (PCB) (108).

9. A colorimeter according to any one of claims 1 to 8, further comprising an opaque shield (118) surrounding the PD (114) to prevent direct irradiation of the PD by the light sources.

10. A colorimeter according to any one of claims 1 to 9, further comprising means for subcarrier modulating the electrical drive signals and means for demodulating the electrical output signal.

5 11. A colorimeter according to claim 10, wherein the subcarrier modulation has a square waveform.

12. A colorimeter according to claims 10 or 11, wherein the demodulation means comprises a phase synchronous demodulator synchronized to the phase of the subcarrier 0 modulator.

13. A colorimeter according to claims 10, 11 or 12, wherein the modulation depth is less than unity.

15 14. A colorimeter according to claim 13, wherein the modulation depth is approximately 95%.

15. A colorimeter according to any one claims 1 to 14, wherein:

- the processor means (135) calculates luminous reflectivity (Y) and chromaticity 20 values (P) of the reflectivity of the surface covering the aperture (106) from the spectrum sample values;

- the luminous reflectivity value Y is calculated as a weighted average of the spectrum sample values as follows:

25 wherein Y is the luminous reflectivity value of the surface covering the aperture

(106);

R_j , j = 1 to n are the spectrum sample values; and

A_j , j = 1 to n are weighting coefficients which depend on the set of n

LEDs used in the colorimeter; 30 - the three CIE chromaticity values P; , i = 1 to 3, are calculated as a matrix transform of normalized spectrum sample values as follows:

wherem P; , i = 1 to 3 are the three CIE chromaticity values of the surface covering the aperture (106);

T_j = R_j/Σ_jR_j, j = 1 to n are the normalized spectrum sample values; and

5 Ay, i = 1 to 3, j = 1 to n is a colour transform matrix having matrix elements that depend on the set of n LEDs used in the colorimeter.

16. A colorimeter according to claim 15, wherein n = 3.

10 17. A colorimeter according to claim 16, wherein the processor (135) stores luminous reflectivity and chromaticity values for a plurality of reference surfaces, determines a best fit between the luminous reflectivity and chromaticity values measured for the surface covering the aperture (106) and the stored luminous reflectivity and chromaticity values, selects the reference surface corresponding to the best colorimetric fit to the surface covering the

15 aperture, and identifies that reference surface to be the colour of the surface covering the aperture.

18. A colorimeter according to claim 17, wherein the processor determines the best fit using a least mean square difference criterion as follows:

20 S^ wΔY^+ΔH- wherein S_m ² is a mean square difference criterion; w is a weighting factor with 0.1 < w < 10;

Δ Y_m ² = (Y - Y- ² is a reflectivity mean square difference criterion; and ΔH--,² = Σ (Pj - P_n v is a chromaticity mean square difference criterion; 25 and

- the least value of S_m ² for m = 1 to k, k being the number of the known reference colours in the store, is deemed to be the best fit.

19. A colorimeter according to claim 18, wherein w is unity. 30

20. A colorimeter according to any one of claims 1 to 19, further comprising a temperature sensor (119) for generating ambient temperature information for use by the processor (135) to correct for temperature dependency of the light output from the Hght sources when deriving said indication from said plurality of values.

21. A colorimeter according to claim 20 as dependent upon claim 8, wherein the temperature sensor (119) is a surface mount device surface mounted on the PCB (108).

22. A colorimeter according to any one of claims 1 to 21, further comprising output means responsive to the processor for communicating the determined colour of the surface covering the aperture (106) to a user non- visually.

23. A colorimeter according to claim 22, wherein the output means communicates the colour aurally.

24. A colorimeter according to any one of the preceding claims, further comprising user- operable activation means, wherein:

- upon activating of the activation means the colorimeter (101) determines the colour of the surface covering the aperture (106); and - upon de-activating the activation means the colorimeter (101) communicates the determined colour to the user.

25. A colorimeter sensor unit for use in a colorimeter for determining the colour of a surface which, when illuminated, will produce at least diffusely reflected Hght, the sensor unit comprising:

- a housing (106) including a wall having an aperture (110) therein, interior surfaces of the housing being adapted to absorb light impinging thereon, the aperture to be covered externally of the housing by the surface when the colorimeter is in use;

- a plurality of light sources (115, 116, 117), and a photodetector (PD) means (114) disposed in the housing and generally facing the aperture, the light sources being each responsive to an electrical drive signal to emit light in a predetermined spectral segment of the visible spectrum,

- the Hght sources and the PD being mutuaUy spaced apart and oriented so that substantiaUy all of light from each Hght source that is specularly reflected by the surface is

5 directed away from the PD and the PD will receive at least a portion of the diffusely reflected Hght from each light source and produce a corresponding electrical output signal having a pluraHty of values each representing the diffuse reflection characteristics of the surface for the spectral segment of the corresponding light source.

10 26. A colorimeter sensor unit according to claim 25, wherein the Hght sources are energized sequentially.

27. A colorimeter sensor unit according to claim 25, wherein:

- the aperture (106) lies substantially in an aperture plane;

15 - emission surfaces of the light sources and a detection surface of the PD are generaUy in a plane parallel to the aperture plane at a pre-determined distance from the aperture plane;

- the PD is disposed opposite the aperture (106);

- the light sources are substantially equidistant from the PD; and

- a maximum lateral dimension the aperture (106) is smaller than the distance between 20 each of the light sources and the PD.

28. A colorimeter sensor unit according to claim 26, wherein the light sources are so disposed relative to the aperture that Hght from each of the light sources wUl be incident upon a said surface covering the aperture (106) at an angle of approximately 45 degrees from the

25 normal to the surface.

29. A colorimeter sensor unit according to any one of claims 25 to 28, wherein the plurality of Hght sources (115, 116, 117) have emission wavelengths corresponding to the colours orange, green, and blue, respectively.

30

30. A colorimeter sensor unit according to claim 29, wherein the emission wavelengths have spectral maxima at approximately 610 nanometers, 555 nanometers, and 450 nanometers, respectively.

5 31. A colorimeter sensor unit according to claim 29 or 30, wherein' the light sources comprise LEDs, the LEDs and the PD being planar semiconductor devices, the LEDs (115, 116, 117) having wide angle emission characteristics and the PD (114) having a wide angle field of view.

10 32. A colorimeter sensor unit according to claim 31, wherein the LEDs (115, 116, 117) and the PD (114) are surface mount devices surface mounted on a printed circuit board (PCB) (108).

33. A colorimeter sensor unit according to any one of claims 25 to 32, further comprising 15 an opaque shield (118) surrounding the PD (114) to prevent direct irradiation ofthePD by the light sources.

34. A process for determining luminous reflectivity and chromaticity values from sample values obtained by irradiating a surface with light, receiving at least a portion of such light

20 reflected from the surface and producing a corresponding electrical signal having a plurality of sample values each representing the reflection characteristics of the surface for a corresponding spectral segment of the visible spectrum, the process comprising the steps of:

- calculating the luminous reflectivity value Y as a weighted average of the sample values as follows:

where Y is the luminous reflectivity value;

R_j , j = 1 to n are the sample values, n being the number of sample values; and

A_j, j = 1 to n are weighting coefficients which depend on the sampling 30 wavelengths, the number of sampling wavelengths also being n; and - calculating the three CIE chromaticity values P_; , i = 1 to 3, as a matrix transform of normalized sample values as follows:

where P_; , i = 1 to 3 are the three chromaticity values; i_j = R/∑_j R_j, j = 1 to n are the normalized sample values; and

Aj_j , i = 1 to 3, j = 1 to n, is a colour transform matrix having matrix elements that depend on the n spectral samples.

35. A process according to claim 34, for use in a colorimeter having stored luminous reflectivity and chromaticity values of a group of reference coloured surfaces, further comprising the steps of:

- determining a best fit between:

- the measured luminous reflectivity and chromaticity values; and

- the luminous reflectivity and chromaticity values stored for each of the known reference coloured surfaces.

36. A process according to claim 35, wherein the best fit is determined using a least mean square difference criterion as follows:

- the least value of S_m ² for m = 1 to k, being deemed as the best fit. where S_m ² is a mean square difference criterion; w is a weighting factor 0.1 < w < 10;

Δ Y_m ² = (Y - Y- ² is a reflectivity mean square difference criterion;

ΔH,-,² = ∑j (P; - J? fl is a chromaticity mean square difference criterion; and k is the number of the known reference coloured surfaces.

37. A process according to claim 34, 35 or 36, wherein n = 3.

38. A process for use in a colorimeter having stored colorimetric coordinates of a group of reference coloured surfaces, the process using luminous reflectivity and chromaticity values measured for a colour to be determined, comprising the steps of:

- determining a best fit between:

5 - the measured luminous reflectivity and chromaticity values; and

- corresponding luminous reflectivity and chromaticity values stored for each of the known reference surfaces;

- the best fit being determined using a least mean square difference criterion as follows:

10 - the least value of S_m ² for m - 1 to k, being deemed as the best fit. where S_ra ² is a mean square difference criterion; w is a weighting factor 0.1 < w < .10;

ΔY_m ² = (Y - Y- ² is a reflectivity mean square difference criterion;

ΔH_m ² = ∑j (Pj -

is a chromaticity mean square difference 15 criterion; and k is the number of the known reference coloured surfaces.

39. A process according to claim 38, wherein w is equal to unity.

20 40. A hand-held optical device having light source means for Ulurninating a surface whose are to be determined, photodetector means for receiving light from the surface and producing a corresponding electrical signal, means for processing the electrical signal and providing an aural indication of the feature so determined, and user-operable switch means for delaying the outputting of such aural indication.

25

41. A hand-held device according to claim 41, wherein the feature is colour.