JPH0527320B2 - - Google Patents

Description

[Detailed Description of the Invention] The present invention relates to a sound collection device for speech recognition, and in particular,
The present invention relates to a sound collection device for voice recognition that collects human voices in order to recognize them.

2. Description of the Related Art Recently, speech recognition devices that control controlled devices by recognizing speech produced by humans have been used in various fields. Such a speech recognition device is provided with a sound collection device for collecting sounds. As a sound collecting device, an acoustoelectric transducer such as a microphone having unidirectionality is generally used in many cases. However, in environments where the surrounding noise is extremely loud, it is not easy to extract only the human voice since the collected voice contains noise.

Therefore, the main object of the present invention is to provide a sound collection device for speech recognition with sharp directionality that can collect speech with a good signal-to-noise ratio even at a distance from the speaker in an environment with very loud surrounding noise. The goal is to provide the following.

To summarize, this invention arranges an even number of four or more acoustoelectric transducers at predetermined intervals on a straight line, and an axis that passes through the center of the row of acoustoelectric transducers and intersects with the straight line. The output signals of each pair of acoustoelectric transducers located at symmetrical positions as the axis of symmetry are added by at least two first adding means, each output signal is weighted by a predetermined ratio, and then the second adding means is added. The structure is such that it has a sharp directivity characteristic in the direction of the axis of symmetry by adding .

The above objects and other objects and features of the present invention will become more apparent from the detailed description given below with reference to the drawings.

FIG. 1 is a schematic block diagram of an embodiment of the present invention, and FIG. 2 is a block diagram of the microphone 1a shown in FIG.
It is a figure for explaining the arrangement state of 1 to 1b.

First, the configuration will be explained with reference to FIG. Microphones 1a to 1d serving as acoustoelectric transducers are arranged in a straight line at predetermined intervals. The microphones 1b and 1c at the center and the microphones 1a and 1d at both ends form pairs, respectively, and the microphone output signals of each pair are added in phase by first adders 2 and 3. The output signals of adders 2 and 3 are individually applied to amplifiers 4 and 5, respectively. The amplification factors of the amplifiers 4 and 5 are weighted to a predetermined ratio, and the microphone output signals of each pair are amplified by the amplifiers 4 and 5, respectively, and then added by the second adder 6.

As mentioned above, in the microphones 1a to 1d arranged at a predetermined distance d, the midpoint between the microphones 1b and 1c is called the microphone center point O, and the vertical bisector of the microphones 1b and 1c is defined as the front direction of the microphone. I will call it that. A sound source S of the sound collected by these microphones 1a to 1d is provided at a distance R from the microphone center point O. The sound source angle between the sound source S and the front direction of the microphone is represented by θ, and the sound source S
The legs of perpendicular lines drawn from the microphones 1b and 1c to the line connecting the microphone center point O and the microphone center point O are represented by C and D, respectively. Here, if the sound source S is a point sound source with intensity Q〓e〓〓 ^t , then the sound pressure P at a point distance R from the sound source S is the well-known formula P=jωQ〓/4πR・e〓 ⁽ 〓 ^t-kR) ...(1) It is expressed by the formula. Here, ω is each frequency, ρ
is the air density, k (=ω/C) is the wave constant, and C is the speed of sound. Now, if the sound source distance R is sufficiently large compared to the microphone interval d, the distance between the microphone 1b and the sound source S is shorter than the sound source distance R by OC.
Further, the distance between the microphone 1c and the sound source S is longer than the sound source distance R by OD. OC=OD=d/2・sinθ
Assuming that the phase shift due to the distance difference is larger than the attenuation of the sound wave due to distance, the distance term in the denominator in equation (1) above is expressed as R≫d/2・sinθ
Using the approximation formula, microphones 1b, 1c
The respective sound pressures P2 and P3 of are P ₂ ≒jθρQ〓／4πR・e ^j {〓 ^tk(Rd/2sin 〓 ⁾ } …(2) P ₂ ≒jθρQ〓／4πR・e ^j {〓 ^{tk(R+ d/2sin} 〓 ⁾ } …(3). Therefore, the sound pressure signal of the microphone pair consisting of microphones 1b and 1c is P ₂₃ ≒2jωρQ〓/4πR·e ^j( 〓 ^t−kR) ×cos(kd/2·sinθ) (4). Similarly, the sound pressure signal of the microphone pair consisting of microphones 1a and 1d is expressed as P ₁₄ ≒2jωρQ〓/4πR・e ^j( 〓 ^t-kR) × cos (3kd/2・sinθ) …(5) . If the amplification factors of the amplifiers 4 and 5 are G ₁ /2 and G ₂ /2, the output sound pressure signal of the sound collector for the sound source S in the θ direction is P _T ≒[−]jωρQ〓/4πR・e ^j( 〓 ^t-kR) × {G ₁ cos (kd/2・sinθ) +G ₂ cos(3kd/2・sinθ)} …(6).

If the above output sound pressure signal is normalized by the output temperature and pressure signal for the sound source in the front direction of the microphone (θ=0), then
The expression representing the sound collector directional characteristic is as follows: D≈1/(G ₁ +G ₂ ) |G ₁ cos (kd/2·sin θ)+G ₂ cos (3kd/2 sin θ) | (7). Here, if we transform the absolute value term in equation (7) above as y, then y=4G ₂ cos ³ (kd/2・sinθ) + (G ₁ −3G ₂ )cos(kd/2sinθ)... (8), and the amplification factors G ₁ and G ₂ can be determined so that equation (8) corresponds to the third-order Tievisiev polynomial. The third-order Tievisiev polynomial T ₃ is T ₃ = 4X ³ −3x …(9), so x = b cos (kd/2sinθ) (where b
>1), the above-mentioned equation (9) becomes T ₃ ′=4b ³ cos ³ (kd/2·sin θ) −3b cos (kd/2 sin θ) (10). Comparing the relationship between equations (8) and (10) above, G ₂ =b ³ G ₁ −3G ₂ =−3b (11). The amplification factors G ₁ and G ₂ are parameters b as numerical values that give the sensitivity ratio between the front direction and other directions.
It is only necessary to decide so that the above-mentioned equation (11) is satisfied.

FIG. 3 is a diagram showing the sound collection characteristics of the embodiment shown in FIG. 1. Next, an overview of the directional characteristics and sensitivity ratio of the sound collector shown in FIG. 1 will be explained with reference to FIG. 3. If equation (10) representing the directional characteristic is shown in a graph, it will be as shown in FIG. This third
In the figure, the horizontal axis is X=b cos(kd/2sinθ), and the vertical axis is the sensitivity ratio. In the front direction, θ=O, so x=b, and the maximum main lobe appears as a sensitivity ratio, with a peak value of 4b ³ -3b. X=±1/
At 2, a side lobe appears and the peak value becomes 1. Therefore, the ratio I of the peak values of the main lobe and the sublobe described above is a function of the parameter b, and is as follows: I=1/(4b ³ -3b) (12). As an example, if we choose b=2.117, I=
1/31.6, that is, the sound pressure due to the sound source in the other direction with respect to the front direction is -30 dB or less. At this time, the amplification factor G ₁ =22.113, G ₂ =9.488, and the ratio of the amplification factors G ₁ /G ₂ is 2.331. Furthermore, if b=6.34 is selected, the amplification factors _G1 =745.5 and _G2 =254.8, and the sensitivity ratio I becomes -60 dB. In this way, by determining the amplification factor using the function shown by the above equation (11),
Various sensitivity ratios can be selected.

On the other hand, the horizontal axis X=±b or kd/2sinθ=nπ
When (n=0, 1, 2...), the main lobe with the maximum sensitivity ratio appears. Therefore, in order to increase the sensitivity only in the front direction, it is necessary to select the microphone interval d so that kd/2<π (13). For example, if the upper limit frequency of the audio band is 3500 Hz, it is necessary to choose within the range of d<9.7 cm.

Note that in the above embodiment, there are two microphone pairs.
In the case where there are M microphone pairs (M≧2), the amplification factors of the M amplifiers may be determined so as to correspond to the 2M-1 order Tievisiev polynomial. Also, an attenuator may be used instead of an amplifier. Further, in the above embodiment, the individual microphones 1a to 1d have omnidirectional characteristics, but when configured using microphones having unidirectional characteristics, the directivity coefficient As is well known as the above-mentioned directivity law, the unidirectivity of the microphone itself can be added to the above-mentioned directivity to obtain even more directivity characteristics.

FIG. 4 is a schematic block diagram of another embodiment of the invention. In this embodiment, five to ten microphones 1a to 1j are arranged at intervals d, and every even number of microphones constitutes a transducer array. That is, microphone 1a and eight microphones 1b to 1
Microphone 1j separated by i constitutes a pair of converter rows, and thereafter microphone 1
A converter array is constructed by b and 1i, 1c and 1h, 1d and 1g, and 1e and 1f, respectively.
The transducer row constituted by the microphones 1b to 1i constitutes a high-frequency directional sound collecting section, and the second microphone pair from the inside is 1d and 1g.
The outermost pair of microphones 1a and 1j constitute a low frequency sound collection section. Then, each microphone pair is added by a first adder 7 to 11. The output signals of adders 9 to 11 are individually amplified by amplifiers 12 to 14 and added to the output signal of adder 8 by a second adder 16. Adder 1
The addition output of 6 is applied to an adder 20 via a bandpass filter 19 that passes only the high frequency range. On the other hand, the outputs of the microphone pair consisting of microphones 1a and 1j are added by an adder 7. Furthermore, the summed output of the microphones 1d and 1g is added to the output of the adder 7 by an adder 17 via an amplifier 15.
The output signal of the adder 17 is applied to the adder 20 via a bandpass filter 18 that passes only the low frequency range. The adder 20 adds the low range output signal and the high range output signal and outputs the result. Note that the amplification factors of each of the amplifiers 12 to 14 are determined to correspond to a seventh-order Tievisiev polynomial,
The amplification factor of the amplifier 15 is determined to correspond to a third-order Tievisiev polynomial.

By the way, when ten microphones 1a to 1j are arranged in a straight line with an interval of d as shown in Fig. 4, as the interval between each microphone increases, the upper limit frequency of the frequency band where high directivity can be obtained is There is a tendency for the attenuation gradient near OHz to become large. The distance between the microphones 1b to 1i in the high-frequency sound collection section is d, and the distance between the microphones 1a and 1d and 1g and 1j that make up the low-frequency sound collection section is d.
Since it is 3D, the frequency band in which directivity can be obtained can be widened by combining sound collection sections with different microphone spacing.

FIG. 5 is a diagram showing an example of the directional frequency characteristic of the low-frequency sound collection section of the embodiment shown in FIG. 4, and FIG. 6 is a diagram showing an example of the directional frequency characteristic of the high-frequency sound collection section in the embodiment shown in FIG. It is.

The directional frequency characteristics shown in FIGS. 5 and 6 are based on examples when the microphone d=8.5 cm and I=30 dB.

As is clear from FIGS. 5 and 6, the output signals from the low-frequency sound collection section and the high-frequency sound collection section pass through bandpass filters 18 and 19, respectively, and are then added together in an adder 20. It has high directivity over the entire audio band required for the recognition device. In addition,
Microphone spacing d for audio band is 3cm≦d
It is preferable to choose a range of ≦10cm.

In addition, in the above-mentioned embodiment, the case where 10 microphones 1a to 1j were provided was explained, but the number of microphones is 10+6(N-1)(N=
1, 2,...), it is possible to configure a sound collecting section with a different microphone interval in addition to a sound collecting section with a microphone interval d. That is, the distance between the second microphone pair from the inside is 3d. If at least one pair of microphones with the same spacing as the second microphone pair is provided outside this, the length will be
Only 3dm (m is an integer of 3 or more) is required. on the other hand,
Since there are 10+(N-1) microphones with an interval of d, their total length is d{10+6(N-1)-1}.
For N=1, 2..., the total length of these two is 3dm=d{10+6(N-1)-1} =3d・{3+2(N-1)}...(14) There are several meters. Therefore, when the number of microphones is 10 + 6 (N-1), it is possible to configure two sound collection sections with the minimum number of microphones and two types of microphone spacing, d and 3d. can. Also, the distance between the third microphone from the inside is 5.
Similarly for microphone pair d, 5dm=d{10+6(N-1)-} ={3+2(N-1)}d・3 (15), and N=2+5(n-1)( n=1, 2,...)
The equality holds true for N. Similar inferences can be made regarding other microphone pairs located outside the microphone interval 7d. Therefore, when the number of microphones is 10+6(N-1), the number of microphones is the minimum and the microphone interval d
The present invention is characterized in that it is possible to configure a sound collection section having at least two different types of microphones, such as the distance between the microphones and the microphone interval 3d.

FIG. 7 is a schematic block diagram of another embodiment of the invention. In the embodiment shown in FIG. 7, microphones 1b to 1e, adders 22 and 23,
The amplifier 25, the adder 27, and the bandpass filter 29 constitute a high-frequency sound collection section. Further, microphones 1a and 1f are arranged 3d apart from the outermost microphones 1b and 1e in this high-frequency sound collecting section, respectively, and these microphones 1a and 1f, the adder 21, the amplifier 24, the adder 26, and the band The pass filter 28 constitutes a low frequency sound collection section. By configuring the sound collecting device in this way, high directivity can be achieved in the same manner as the embodiment shown in FIG. 4 described above.

As described above, according to the present invention, an even number of four or more acoustoelectric transducers are arranged on a straight line with predetermined intervals, and each pair is symmetrically located with the center line as the axis of symmetry. By adding and weighting the output signals of the respective acoustoelectric transducers and further adding them, it is possible to have a sharp directivity characteristic in the direction of the axis of symmetry. Therefore, even in a very noisy environment, it is possible to collect only sounds coming from a specific direction within the necessary audio band and with a good SN ratio.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram of one embodiment of the present invention. FIG. 2 is a diagram showing the geometric positional relationship between the microphone arrangement and the sound source position shown in FIG. 1. FIG. 3 is a characteristic diagram of the embodiment shown in FIG. FIG. 4 is a schematic block diagram of another embodiment of the invention. FIG. 5 is a diagram showing an example of the directional frequency characteristics of the low-frequency sound collection section of the embodiment shown in FIG. 4. FIG. 6 is a diagram similarly showing an example of the directional frequency characteristics of the high-frequency sound collection section. FIG. 7 is a schematic block diagram of another embodiment of the invention. In the figure, 1a to 1j are microphones,
2, 3, 6 to 11, 16, 17, 20 to 23, 26, 27, 30 are adders; 4, 5, 12
15, 24, 25 are amplifiers, 18, 19,
28 and 29 indicate bandpass filters.

Claims (1)

[Scope of Claims] 1. A sound collection device for voice recognition for collecting the sounds produced by a speaker, comprising an even number of four or more acoustic and electrical devices arranged at predetermined intervals on a straight line. a transducer, each pair of acoustoelectric transducers located symmetrically with respect to an axis of symmetry passing through the center of the array of acoustoelectric transducers and intersecting the straight line, each pair of acoustoelectric transducers being connected to a respective input; two first adding means for adding the output signals of the acoustoelectric transducers; connected to the output of each of the first adding means, weighting the output signals of the corresponding first adding means respectively at a predetermined ratio; and a second addition means for adding up the output signals of each of the weighting circuit means, and having sharp directivity in the direction of the axis of symmetry. Sound collection device. 2 The number of the acoustoelectric transducers is 10 or more and 4+6N (N
= 1, 2,...), and every even number of acoustoelectric transducers among the respective acoustoelectric transducers constitute a transducer array, and the transducer array provides directivity at different frequencies. The sound collection device for speech recognition according to claim 1, wherein the sound collection device includes two types of sound collection sections for different bands. 3. said plurality of acoustoelectric transducers include a transducer arrangement with the addition of a transducer pair having a spacing three times that of any one of said symmetrically located transducer pairs, thereby providing directivity. Two types of sound collection sections with different frequency bands are configured to obtain
A sound collection device for speech recognition according to claim 1.