JPH0462597B2 - - Google Patents

Description

[Detailed description of the invention]

[Technical Field] The present invention relates to a voice message identification method for operating electronic equipment using voice messages. [Background Art] Since phonological information depends on the articulation method, it is mainly included in the envelope of the frequency spectrum that corresponds to changes in cross-sectional area, and is especially contained in the envelope of the frequency spectrum corresponding to changes in cross-sectional area. ) and its bandwidth (50-110Hz). The frequency spectrum of speech is almost determined by the vocal tract transfer characteristics and the shape of the sound source waveform, but the vocal tract transfer characteristics include resonance points determined by the vocal tract cross-sectional area and resonance points determined by the vocal tract length. Articulation, or phonology, is determined mostly by the cross-sectional area of the vocal tract, and the length of the vocal tract varies between men, women, children, and individuals. Furthermore, the sound source waveform (particularly due to vocal fold vibration of voiced sounds) is thought to depend on the pitch and strength of the voice. Therefore, for phoneme identification, by extracting the envelope of the frequency spectrum excluding the influence of the sound source waveform and vocal tract length, high discrimination ability (recognition rate) with little individual difference can be obtained. By the way, when performing speech recognition using the method of obtaining the spectral envelope, a high recognition rate and a large number of recognized words can be obtained, but known methods (for example, linear predictive analysis using an all-pole model) require a huge amount of calculation, so Requires expensive systems such as, and its uses are limited. By the way, the prediction residuals include impulse trains with flat frequency spectra (voiced sounds) and random noise (unvoiced sounds).
Contains the sound source waveform. Here, in the case of voiced sounds, the 16th
This is because when the sound source waveform of the speech waveform shown in Figure G is the sound source waveform of Figure A, it is regarded as an impulse sound source as shown in Figure b. By the way, when looking at the actual voice waveform and the sound source waveform shown in FIG. 16(f), it intuitively appears that the differential waveform (FIG. 16(c)) of the sound source waveform in FIG. 16(a) overlaps. Since the falling slope of the sound source waveform is gentle, but the falling slope is steep, the magnitude of the differential waveform is large in the downward direction and small in the upward direction. On the other hand, in order to remove the influence of voiced sound sources in order to estimate Holtmant and spectral envelope, homomorphic processing is known in which the product on the frequency spectrum is converted into a sum. Since a large amount of fast Fourier transform had to be repeated twice, it could not be called simple. A short interval analysis method is also known as a method for determining the spectral envelope and Holt mant. This is a linear predictive analysis performed within an interval that is slightly longer than the period of a voiced sound or as short as 3 msec (especially the glottal closure interval), and is not affected by the vocal fold frequency.
It is said that Holtmant can be calculated, but there is a problem in that it requires a large number of equal multiplications to calculate a correlation function for linear prediction and to calculate a covariance matrix for short interval determination. Therefore, the present inventors proposed a method in which the spectral envelope can be calculated with simple interval determination and a small number of multiplications using short-interval fast Fourier transform that centers the analysis window at the point of large amplitude of the voice. 1st
If we consider that the analysis is performed using something like the differential waveform of the sound source waveform shown in Figure 6c, it must be said that it is influenced by individual differences. (However, it is practical as a simple type when the spectral envelope is only rough.) [Objective of the Invention] The present invention has been made in view of the above-mentioned problems, and its purpose is to reduce the amount of calculation. To provide a voice message identification method which is less affected by individual differences among speakers and vocal cord vibration. [Disclosure of the Invention] First, let us consider the basic principle of the present invention. If we consider the vocal tract model as a linear system as in linear predictive analysis, the speech waveform is a convolution calculation of the sound source waveform and the vocal tract transfer characteristics (impulse response). The system that makes an impulse response to the vocal tract can be said to be the actual vocal tract model. For this reason, it is extremely difficult to create a sound source model for an impulse train by actually observing the sound source waveform because it involves double convolution. Therefore, I thought that it would be good if it was simple and practical, and if it was effective as a result of actual experiments, and as mentioned above, when looking at the time waveforms of the five vowels (i, e, a, o, u) of the voice, Since it appears to include the differential waveform of the sound source as shown in Figure 16c, the simplest way to think of it is to convert the waveform in Figure 16e, which is the differential waveform when the sound source waveform is Figure 16f, to the waveform in Figure 16g. The basic technical idea of the present invention is to perform fast Fourier transform by subtracting from the waveform of . Here, the analysis interval may be within the interval i and j shown in FIG. 16e, but in this case, it corresponds to treating the response to the impulse shown in FIG. 16d as if it were subjected to Fourier transform. Embodiment FIG. 1 is a circuit configuration diagram based on the processing flow of the present invention. In the figure, 1 is a high frequency emphasis section, and this high frequency emphasis section 1 is for emphasizing the high frequency range of input audio. . Reference numeral 2 denotes an A/D converter that performs A/D conversion of the high-frequency emphasized input audio, and the output from this A/D converter 2 is input to a section compensator 3 and a pitch detector 4. The pitch detection unit 4 detects the absolute value IPOW of the amplitude of the time waveform of the A/D converted audio within one frame.
The first large downward amplitude point IPN1 where the average value of the predetermined number of samples is the lowest in the first half of the frame is detected, and the second large downward amplitude point IPN3 that follows this is found, and these large downward amplitude points IPN1 are found. and
This is to obtain pitch IPit from IPN3. The interval compensator 3 is for performing pitch compensation, and detects the waveform shown in FIG.
Compensation is made by subtracting between
As shown in figure b, point A where the amplitude is −IPOW,
Assume that a straight line X connects point B where the amplitude IPOW is obtained at two points of IPN. Note that IPN0 in FIG. 4 is the closest large amplitude point. Also, as the offset value, the first
The point C, which is half the peak value at the large downward amplitude point IPN1, and the point D, which is half the peak value at the large upward amplitude point IPN2, are connected by a straight line, or the large upward amplitude point IPN0 and the large downward amplitude point Even if a straight line connects point E shown in Fig. 5b, which is -IPOW between amplitude point IPN1, and point F, which is IPOW halfway between upward large amplitude point IPN2 and downward large amplitude point IPN3, good. Note that FIGS. 4a and 5a show the frequency spectra of the voices in FIGS. 4b and 5b, respectively. Reference numeral 5 denotes an analysis interval determination unit, and this analysis interval determination unit 5 selects 64 points or 32 points from the downward large amplitude point IPN1.
within points, or 1/2, 1/ of (IPN2-IPN1)
Set it to a value of 4. Here, the number of samples for fast Fourier transform is 64, 128, and 256.
Generally speaking, the lengths of the short sections are 32 points and 64 points for window calculation reasons. Reference numeral 7 denotes a fast Fourier transform section. This fast Fourier transform section 7 is used to obtain the envelope of the frequency spectrum by fast Fourier transform, and during calculation, the spectrum window determined by the analysis window calculation section 6 is multiplied. It will be done. The analysis window calculation unit 6 optimizes the length and position of the spectral window to be applied to the fast Fourier transform so that the fast Fourier transform unit 7 can extract the spectral envelope more accurately and with fewer calculations (number of multiplications). It is for the purpose of Reference numeral 8 denotes a frequency band division section, and this frequency band division section 8 converts the frequency spectrum extracted by the fast Fourier transform section 7 into a logarithmic power spectrum, and then converts the short-term average power of each frequency component, such as UV.V, VH, into a logarithmic power spectrum. ,VL,VF,
It is used to obtain the 6 components of VB. here,
V indicates the short-time average power in the frequency band of 0 to 1 KHz during voice input, and corresponds to the energy of voice with voice. UV is 5-12KHz during audio input
It shows the short-term average power in the frequency band of
It corresponds to the energy of silence. There is also a voice
VL, VH, VB, VF are respectively during audio input, 0~
0.5KHz, 0.5~1.0KHz, 1.0~2.0KHz, and 2.0~
It shows the short-term average power in the 4.0KHz frequency band, and corresponds to the energy of narrow jaw sounds, wide jaw sounds, back tongue sounds, and front tongue sounds, respectively. 9 is a difference signal vector converter, and this difference signal vector converter 9 converts 5 phonemes (i, e, a,
o, u) are respectively eao/iu, a/eo, e/o, i/
UV/V, VH/VL,
This is to find the difference signal vectors of VF/VB, VB/VL, and VF/VH. Reference numeral 18 is a formant locus converter for obtaining a formant vector, whereas the frequency band dividing section 8 and the difference signal vector converting section 9 are for obtaining a difference signal vector by frequency band division. The peak frequency (formant frequency) of the spectrum envelope is determined and used as a formant vector, and the components of the formant vector are the differences with respect to the average value for each formant, and the frequency axis is expressed on a logarithmic or linear scale. The recognition rate can be improved by switching the above-mentioned average value, which serves as the reference frequency for each formant, to classes such as male, female, and child through pitch detection. 15th
Figures a and b show the formant distribution of five vowels and the positions of the peaks. 10 is a symbol vector converter, and this symbol vector converter 10 converts the above difference signal vector or formant vector and a conversion matrix into a symbol vector {i, e, a, o, u, h, l, f, b, w}.
The values of the transformation matrix need only have row components corresponding to the magnitude of each component of the difference signal vector or formant vector corresponding to the symbol. Reference numeral 11 denotes a start/end detection section, and this start/end detection section 11 detects the UV/V difference signal at a certain set value Ru.
It has a voiced/unvoiced determination function that determines UV when it is more positive, V when it is more negative than a certain set value Rv, and S when it is in between, and detects the beginning of the voice by determining UV and V. If silence continues for a number of samples equal to or greater than a certain set value, it is detected as the end. 12 is a symbol conversion processing unit, and this symbol conversion processing unit 12 outputs the symbol when the maximum component of the symbol vector is greater than a certain set value in the interval of V,
If the value is less than the set value, m is output. Further, in the interval between UV and S, UV and S are output respectively. 13
is a formatting processing section, and this formatting processing section 13 converts repetitions of the same symbol into a list of one symbol and its duration, and further omits symbols with short durations. Reference numeral 14 denotes a word standard pattern storage section, and this word standard pattern storage section 14 is used to register a voice pattern in a registration mode and use it as a standard pattern at the time of recognition and verification. The preliminary selection unit 15 is used in the recognition mode to perform preliminary selection to limit the objects of comparison by performing primary identification based on the number of UVs or the like before comparison. 16 is a time axis normalization/verification unit, and this time axis normalization/verification unit 16 normalizes the duration so that the total duration of the above list becomes a constant value, such as 200 (or 1000). The distance between corresponding symbols on the time axis (correlation value) is determined using the time axis normalization function for It consists of a distance calculation function that uses a distance table to match a standard pattern.

[Table] In Table 1, the horizontal and vertical columns correspond to the standard pattern symbol and input pattern symbol, respectively.For example, if the standard pattern symbol is a, and the input pattern symbol is also a When , the output of the distance table is -2, indicating that the degree of approximation is low. Therefore, in the distance calculation function, the degree of approximation of the input pattern and the standard pattern as a whole can be easily calculated by simply adding the outputs from the distance table in sequence. Reference numeral 17 denotes a significant difference testing unit, which detects the closest pattern when it is closer than a certain setting value and is further away from the second closest pattern by more than a certain setting value. It is equipped with a significant difference test function that considers the input patterns to be the same and rejects them as recognition failure in other cases, and a result output function that outputs the recognition result. 19 is an optimization feedback section, and this optimization feedback section 1
9 is the speaker's /i, e,
The time series of the occurrences of a, o, and u/ is stored, and the dividing frequency is varied in the vicinity of a preset standard dividing frequency, and the direction and amount of variation are determined according to the sensitivity characteristics of the symbol vector. In this case, the slope of the spectrum is compensated by the offset of the difference signal vector, and in particular, when the input voice is an A note, the i component is prominent, and when the input voice is an A note, the a note is prominent. /e/, /u/
The gain balance of the difference signal input is adjusted so that identification of the difference signal becomes more reliable. In this case, first VH/VL
Then, the optimum adjustment of VF/VB is performed, and then the optimum adjustment of VB/VL is performed. In the example, the sampling period is 80 μsec (sampling frequency 12.5 KHz), and the frame length is
There were 512 samples. The period of the fundamental frequency is the lowest
If it is 90Hz, there will be 139 samples, and in order to calculate the frequency spectrum of 256 points, normal fast Fourier transform will require calculation of 512 points, and the number of multiplications will be ^29.
× (2 ⁴ + 2 ⁵ ) = 512 × (16 + 32) = 24576 times,
When 64 samples in an interval shorter than the fundamental period are extracted from a frame of 512 samples and analyzed, 128
Fast Fourier transform of a point is sufficient, so 2 ⁷ × (2 ³ + 2 ⁴ ) =
128×(8+16)=3072 multiplications are enough. Furthermore, since the multiplication of the analysis window in the preprocessing of the fast Fourier transform is the same as the number of samples of the frequency spectrum, it can be seen that the short interval analysis is effective as a simple method. FIG. 2 shows a flowchart of the characteristic part consisting of the pitch detection section 4 and the analysis interval determination section 5 of the embodiment of FIG.
Calculate IPOW in (1), and in (2) find the downward large amplitude point where the average value of each 30 samples is the lowest in the first half of the frame.
Detect IPN1, and then proceed to the next downward large amplitude point in (3)
Detect IPN3, and in (4), from these downward large amplitude points IPN1 and IPN3, pitch IPit=IPN3−
Find IPN1. After detecting the pitch (5), the previous upward large amplitude point IPN0 is detected from the downward large amplitude point IPN1, and from the point with amplitude 0 between the two large amplitude points IPN0 and IPN1, the large amplitude point IMiD is determined in (6). Large amplitude point
A short section including one cycle consisting of a half cycle of large upward amplitude and a half cycle of large downward amplitude is determined with IMiD as the center. Next, perform DC compensation in (7), perform analytical windowing in (8), perform fast Fourier transform in (9), select mode between difference signal vector or formant vector in (10), and (11) Find the frequency band division with (12) and the formant locus. FIG. 3 shows a specific circuit diagram of the present invention, in which audio is input from a microphone 18, amplified by a preamplifier 19, and adjusted for gain and offset by an adjustment amplifier 20. Next, the A/D conversion circuit 21 digitally converts the audio input, and the digitally converted audio frame is stored in the audio frame memory 23. Reference numeral 24 denotes an FFT processor, and this FFT processor 24 is a general type having a control section 24a, an arithmetic register 24b, a built-in RAM 24c, and a coefficient ROM 24d in which coefficients are stored.
Consists of FFT chip, audio frame memory 23
The audio frame read from is captured and fast Fourier transform is performed using a window. 25 is a spectral frame memory, and FFT processor 24
This is for storing the calculated spectral frame. 22 is an audio frame memory 23;
This is a timing circuit that provides operation timing for the FFT processor 24 and spectrum frame memory 25. 26 is a CPU that performs control calculations based on the operation program written in advance in the program ROM 27, and in the verification mode operates the verification calculation circuit 30 to encode data stored in the spectral frame memory 25 and register it in advance. In the mode, it is possible to perform a comparison calculation with the standard pattern stored in the standard pattern RAM 31, or in the registration mode, the input voice pattern can be stored as a standard pattern in the standard pattern RAM 31, and furthermore, in the learning mode, the above-mentioned optimization can be performed. We send feedback. In the figure, 32 is the terminal section, 33 is the microcomputer bus, and 28 is the working section.
RAM 29 is a control input/output section. Figure 6 shows the simulation result of fast Fourier transform (256 points) multiplied by a window of 128 points between point A and point B mentioned above. This is the frequency spectrum. Figure 7 shows the fast Fourier transform (256 points) multiplied by the window of 128 points between point C and point D mentioned above.
These are the simulation results obtained by performing the above, and the figure a shows the frequency spectrum shown as the time waveform in the figure b. The case shown in FIG. 7 is not much different from the case shown in FIG. Figure 8 shows the fast Fourier transform (128 points) multiplied by the 64-point window between points C and D mentioned above.
These are the simulation results obtained by performing the above, and the figure a shows the frequency spectrum shown as the time waveform in the figure b. Figure 9 shows fast Fourier transform (64 points) by multiplying the window of 32 points between point C and point D mentioned above.
These are the simulation results obtained by performing the above, and the figure a shows the frequency spectrum shown as the time waveform in the figure b. As can be seen from these FIGS. 8 and 9, resonance points corresponding to the second and third formants are clearly visible. Figure 10 assumes that there is an impulse in the rising section of vocal cord vibration, and shows a section shorter than the large upward amplitude point IPN0 (no pitch compensation).
This is a simulation result obtained by multiplying a window of 128 points and performing fast Fourier transform (256 points). Figure a is the frequency spectrum shown as the time waveform in figure b. In addition, Fig. 11 assumes that there is an impulse in the rising section of the vocal cord vibration, and analyzes it in a shorter section (no pitch compensation) than the upward large amplitude point IPN0.It is multiplied by a window of 64 points and fast Fourier transform (12 point)
These are the simulation results obtained by performing the above, and the figure a shows the frequency spectrum shown as the time waveform in the figure b. Comparing these FIGS. 10 and 11 with the above-mentioned FIG. 8, it is clearly seen that the resonance point is clear in the case shown in FIG. 8. In the case shown in Figure 8, the cutout section is IPN2−IPN1=LWP=
This effect can be said to be large since the section in Figure 11 has 64 points, compared to 48 points, which is half of 96 points.
Figure 12 is an analysis of 64 frame length points at positions that appear to be the glottis closure section.In this case, the resonance characteristics are not clear even compared to Figure 11, so the effect of this method shown in Figure 8 is effective. I understand something very clearly. Note that a in the same figure is a frequency spectrum shown as a time waveform in the same figure b. Incidentally, FIG. 13 shows the symbolization process. In the figure, V indicates the short-time average power in the frequency band of 0 to 1 KHz during voice input, and corresponds to the energy of voiced sound. Furthermore, UV indicates the short-term average power in the frequency band of 5 to 12 KHz during audio input, and corresponds to the energy of unvoiced sound. Furthermore, VL, VH, VB, and VF are respectively 0~0.4KHz, 0.4~0.8KHz, and 1.8~3.2K during audio input.
It shows the short-term average power in the Hz frequency band, and corresponds to the energy of narrow jaw sounds, wide jaw sounds, back tongue sounds, and front tongue sounds, respectively. S ₀ to _{S 4} are differential amplification means, which output difference signals V/UV, Veao/Viu, and
It calculates Va/Veo, Ve/Vo, and Vi/Vu. C ₀ is a comparison means, and the above differential amplification means
When the difference signal component output from S ₀ is smaller than the reference value Rv, it is assigned the sign of a voiced sound V, when it is larger than the reference value Ru, it is assigned the sign of an unvoiced sound UV, and otherwise it is silent. Determined as S. However, Ru>O>Rv. MY ₀ is a symbolization processing section, and this symbolization processing section MY ₀ is silent,
S, V, UV for each case of voiced and unvoiced sounds
Enter one of the codes.
MC ₀ is each difference signal output Vea/Viu, Va/Vea,
A four-dimensional vector whose components are Ve/Vo, Vi/Vu is multiplied by a predetermined matrix Tm, and each vowel i, e, a, o, u and other voiced sounds h, i included in the audio input are , f, b, w, and the output of the matrix calculation unit MC ₀ is input to the maximum value determination unit MX ₀ to calculate each component i, e, a,
It is determined which of o, u, h, l, f, b, and w is the largest component, and the code of the largest component is input to the encoding processing unit _MY0 . However, when the difference between the largest component and the second largest component is small, the code m is output. When the code output from the comparison means _C0 is V, the symbolization processing unit _MY0 calculates the following:
i, e, a, output from the maximum value judgment unit MX ₀ ,
Outputs one of o, u, h, l, f, b, w, and m, and when the code output from comparison means _C0 is U or S, outputs that code as is. It is something to do. Matrix calculation part
As the transformation matrix Tm of MC ₀ , equations (1) to (3) can be used. [Tm]=-17 17, 17, 17, -17, 18, -18 0, 0, 13, 0, 0, 17, 0, 0, 0, 0, 0, 0, 0, 0, 17, 0 , -17, 0, 0, 0, 18, -18, 0,17 0 0 0 0 0 0 0 0 -13 ...(1) [Tm] = -16, 16, 16, 16, -16, 18, -18, 0, 0, 13, -8, -8, 16, -8, -8, 0, 0, 0, 0, 0,0, 16, 0, -16, 0, 0, 0, 18, −18, 0,16 0 −8 0 −16 0 0 0 0 −13 …(2) [Tm] = −14, 14, 14, 14, −14, 18, −18, 0, 0, 13, − 14, -14, 14, -14, -14, 0, 0, 0, 0, 0,0, 14, 0, -14, 0, 0, 0, 18, -18, 0,14 0 -14 0 −14 0 0 0 0 −13 …(3) First, the transformation matrix Tm in equation (1) is set to 0 except for the minimum necessary elements for identification to speed up the calculation, and the transformation matrix Tm in equation (2) (3) gives redundancy to the part where the absolute value of the element is 8, and when the detection of the difference signal is weak, it is possible to symbolize a wide range of five vowels. The elements of the 5 vowels for the difference signal related to F ₁ are all given the same weight (absolute value 14), and for the two difference signals related to the second formant F ₂ , one of the 5 vowels is given the same weight (absolute value 14). The first formant F 1 is given the necessary weight to identify each individual, and the first formant F ₁ is converted into the second formant.
It can be said that it was given more importance than _F2 . This transformation matrix
Tm can be arbitrarily set depending on the word to be identified. AP in FIG. 3 shows the characteristics of the adjustment amplifier 20 described above. In addition to the above-mentioned matching method, it is also possible to create a binary signal from the difference signal, encode it with this combination, and sequentially match. Examples of this method include: In other words, UV/V difference signal, Veao/Viu difference signal, Va/Veo difference signal, Ve/Vo difference signal, Vi/Vu obtained from the vector of short-term average power.
Extract the difference signal and assign the symbol Veao if the Veao/Viu difference signal is above a certain positive value, assign the symbol Viu if it is below a certain negative value, and assign the symbol S in other cases. When the difference signal is above a certain positive value, the symbol Va is assigned, when it is below a certain negative value, the symbol Veo is assigned, and in other cases, the symbol S is assigned,
When the Ve/Vo difference signal is above a certain positive value, the symbol Ve is assigned, when it is below a certain negative value, the symbol Vo is assigned, and in other cases, the symbol S is assigned, and the Vi/Vu difference signal is positive. When it is above a certain value, the symbol Vi is assigned, when it is below a certain negative value, the symbol Vu is assigned, and in other cases, the symbol S is assigned. Then, while storing these symbols in the temporary storage means and referring to the symbolization table shown in Table 2, the symbols a, e, o,
Convert to one of the symbols i, u, h, l, f, b, w, and m.

【table】

[Table] However, in Table 2, * indicates that it can be either 0 or 1, and 0/1 indicates 0 or 1.
The case is shown below. Such a symbolization table is constructed using, for example, a ROM, and by accessing the ROM by using the temporarily stored contents as an address input, codes for each symbol such as a, e, o, etc. can be obtained as data output. Alternatively, it is preferable to compare the temporarily stored contents and the contents of the symbolization table using exclusive OR, and output a symbol when they match. FIG. 14 is a flowchart showing a method for realizing the matching section using a sequential discrimination processing program of a microcomputer.The first step is to determine whether the Veo/Viu difference signal is at a high level H or at a medium level M. They are divided into three groups depending on whether they are low level or low level L.
In the second stage, when the first stage is H, if the Va/Veo difference signal is H, the symbol /a/ is output, if it is M, the symbol /Vo/ is output, and if it is L, the third Step 1, check the Ve/Vo difference signal, and
If so, output the symbol /e/, if M, output /h/, and if L, output the symbol /o/. On the other hand, if the first stage is M, in the second stage, if the Ve/Vo difference signal is H, the symbol /f/ is output, if it is M, the symbol /m/ is output, and if it is L, the symbol /b/ Output. Furthermore, if the first stage is L, then in the second stage
If the Vi/Vu difference signal is AH, output the symbol /i/,
If M, output the symbol /l/; if L, output the symbol /u/
It outputs . Although the above-described embodiment uses fast Fourier transform as the orthogonal transform, Walsh transform may also be used. [Effects of the Invention] The present invention maintains the positive and negative phases of the time waveform of voice input so that they are constant with respect to the direction of vocal fold vibration, and corresponds to the falling edge of vocal fold vibration when the differential waveform of vocal fold vibration is negative. Find the first large downward amplitude point of the temporal waveform of the voice that corresponds to the fall of the vocal fold vibration where the differential waveform of the next vocal fold vibration is negative. and a first large downward amplitude point and an upward large amplitude point of the temporal waveform of the voice corresponding to the rise of the vocal fold vibration in which the differential waveform of the vocal fold vibration is positive before the second downward large amplitude point. Since the envelope of the frequency spectrum of the voice input is extracted by performing a short interval analysis in the middle of This has the effect of allowing high recognition.

[Brief explanation of the drawing]

FIG. 1 is a schematic circuit configuration diagram of an embodiment of the present invention, FIG. 2 is a flowchart for explaining the operation of the main parts of the same, FIG. 3 is a specific circuit diagram of the same, and FIGS.
2 is an explanatory diagram of the same operation as above, FIG. 13 is a circuit block diagram explaining the symbolization process of the embodiment of the present invention, FIG. 14 is a flowchart for explaining the operation of another collation example of the present invention, and FIG. The figure is a waveform diagram for explaining the Holtmann trajectory, and FIG. 16 is a waveform diagram for explaining the background technology of the present invention.
5 is a pitch detection unit, 5 is an analysis interval determination unit, 6 is an analysis window calculation unit, 7 is a fast Freeier transform unit, 8 is a frequency band division unit, 9 is a difference signal vector conversion unit, and 8 is a Holtmann locus conversion unit.

Claims (1)

[Claims] 1. Means for extracting the frequency spectrum of the audio input, frequency-dividing the logarithmic power spectrum, extracting the short-term average power for each frequency band, and extracting the five vowels i, e, a, o, u from these short-time average powers.
Extract the difference signal vector so that it is divided into the ratio of e, a, o and i, u, the ratio of a and e, o, the ratio of e and o, and the ratio of i and u, or extract the formant vector from the formant locus. matrix calculation means for calculating symbol vectors of five vowels and other voiced sounds by multiplying the difference signal vector or formant vector by a transformation matrix; A means for outputting the largest component as an onomatopoeic symbol of an analysis frame and a standard pattern in which input patterns consisting of symbols and durations are stored in advance based on the onomatopoeic symbol are compared on a time axis or by symbol. In a voice message identification method that identifies the standard pattern closest to the input pattern as the input message, the positive and negative phases of the time waveform of the voice input are maintained so that they are constant with respect to the direction of vocal cord vibration. , find the first downward large amplitude point of the voice time waveform corresponding to the falling edge of vocal fold vibration where the differential waveform of vocal fold vibration is negative, and find the next falling edge of vocal fold vibration where the differential waveform of vocal fold vibration is negative. means for determining a second downward large amplitude point of the temporal waveform of the voice corresponding to the second downward large amplitude point, and the differential waveform of the vocal fold vibration before the second downward large amplitude point is positive, and the voice corresponds to the rise of the vocal fold vibration. A voice message identification method characterized by extracting an envelope of a frequency spectrum of a voice input by performing a short interval analysis between an upward large amplitude point and a first downward large amplitude point of a time waveform.