JP2948304B2 - Sound receiving method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound receiving method for receiving a sound signal having a good SN ratio in a sound field in which noise and a sound signal to be received are mixed. .

"Conventional technology" In recent years, there has been remarkable development of a voice communication system such as a teleconference and a speech recognition device. In these systems, desired sound (voice, music, etc .; hereinafter, referred to as a target signal)
Is received by a microphone installed at a position slightly away from the sound source (for example, the mouth of the speaker). However, at this time, if there is noise (general term of unnecessary sounds such as mechanical noise and disturbing sound) in the surroundings, the noise is also received at the same time as the target signal. It is a problem.

A conventional sound receiving device for preventing such ambient noise from being mixed will be described with reference to FIG. In FIG. 5, a curve 1 represents sensitivity directional characteristics of various microphones (hereinafter, simply referred to as directional characteristics). When the microphone is located at the center of the circle 2, the curve 1 extends from the center of the circle 2 to the curve 1. The distance indicates the magnitude of the sensitivity in each direction. The case where the direction of maximum sensitivity is the direction 3 of the target signal and the direction 4 of the noise is in a direction different from this direction is shown.

FIG. 5A shows the directivity characteristics of a sound pressure temporary gradient type unidirectional microphone (hereinafter abbreviated as a unidirectional microphone) mainly used in the related art. As shown in FIG. 5A, since the unidirectional microphone has high sensitivity in the ± 90 ° direction toward the target signal direction 3, when the noise comes from the direction 4 within ± 90 ° with respect to the target signal direction 3. Is received with almost no suppression of the noise.
As a result, a satisfactory S / N ratio improvement effect was not obtained with the conventional unidirectional microphone.

As a first method for obtaining a high SN ratio improving effect, a phase synthesis type microphone is known. As shown in FIG. 5B, the phase-combining microphone sharpens the directional characteristics in the target direction 3, thereby lowering the sensitivity in the noise direction 4 and improving the SN ratio.

FIG. 6 shows a configuration using a microphone array (a sound receiver composed of a plurality of microphone elements) as a conventional example of a phase synthesis type microphone. In FIG. 6, M
The number of microphone element 11 ₁ to 11 _M are linearly arranged, in this example are arranged vertically constitute a microphone array 12, the sound receiving the output of the microphone element 11 ₁ to 11 _M adder 13 Is added and output. Microphone element 11 ₁ to 11 _M of SEQ perpendicular direction, a target signal (sound waves) 14 arriving from the horizontal direction in this example, the noise (sound waves) 15 arrives obliquely with respect to the arranging direction of the microphone element 11 ₁ to 11 _M I do.
Thus, while the target signal 14 arriving from the horizontal direction is emphasized because it is summed in phase noise 15 coming from an oblique direction is added after being received sound with a time difference (phase difference) in each microphone element 11 ₁ to 11 _M Be suppressed.

As described above, the result of phase synthesis is to add a plurality of audio signals so that the target signal has the same phase, and a sound arriving from other directions has a phase difference. , The sharp directional characteristics shown in FIG. 5B can be realized. Such a sharp directional characteristic is often called super-directivity, and a microphone using such a sound receiving method is often called a super-directional microphone. However, hereinafter, in this specification, this system is called a phase synthesis type sound receiving system, and a sound receiving device based on this system is called a phase synthesis type microphone.

As other means for constructing the phase synthesis type microphone, a method using a parabolic reflector, a method using an acoustic tube, and the like are known (for example, edited by Nakajima: “Acoustic Engineering Course 2: Applied Electroacoustics”, Showa) 54, Corona).

Now, the phase synthesis type microphone has a good SN ratio improving effect. However, in order to sharpen its directional characteristics,
A sound receiver scale several times the spatial wavelength of the target frequency, that is, the length L of the microphone array 12 (FIG. 6) is required. For example, at 170 Hz, which is near the lowest frequency of sound, the spatial wavelength of the sound wave is 2 m. To achieve sharp directional characteristics in this frequency band, the microphone array 12 and the reflection There was a problem that a board was required. In other words, the directivity characteristic of the phase synthesis microphone has a strong frequency dependence, and the effect of improving the sn ratio is small at low frequencies. For example, in FIG. 6, when the sound receiver scale L is 50 cm,
The directional characteristics at 0 Hz, 680 Hz, and 1360 Hz are shown in FIGS. 7A, 7B, and 7C, respectively. As in FIG. 5, curve 1 represents the directivity characteristic, and arrow 3 represents the direction of the target signal. From this figure, 680Hz
In the following, it can be seen that a large SN ratio improvement effect cannot be expected. Since a linear microphone array is used, the directional characteristics are vertically symmetric.

As a second method for obtaining a high SN ratio improving effect, an adaptive microphone array (hereinafter, abbreviated as an adaptive array) is known. As shown in FIG. 5C, the adaptive array detects the direction 4 of noise and automatically lowers the sensitivity in that direction to improve the S / N ratio (for example, Monzingo and Miller). , “Introduction to Ad
aptive Arrays, "John Wiley & Sons, 1980).

FIG. 8 shows a configuration example of an adaptive array. In Figure 8, the microphone array 22 is constituted by M microphones elements 21 ₁ through 21 _M, the output of the microphone element 21 ₁ through 21 _M are output are added in the adder 24 through the filter 23 ₁ ~ 23 _M, respectively You. Each filter coefficient of the filter 23 ₁ ~ 23 _M are controlled by the filter control unit 25.

In this configuration, the filter control unit 25, noise power PN included in the output signal 26 of the adder 24 controls the filter 23 ₁ ~ 23 _M to minimize. However, in this case, if the noise power PN is simply minimized, the filter coefficients become all zero, and the noise power PN becomes zero, but there is no meaning that the target signal is not output. Therefore, usually, a certain constraint condition is given to the filter coefficient, and the noise power PN
Is minimized.

Examples of the constraint conditions include a non-distortion constraint condition (a condition that the target signal is not distorted at all), a secondary constraint condition (a condition that the target signal is not distorted beyond a certain allowable range), and the like. . Minimizing noise under such a constraint is equivalent to reducing sensitivity in the noise direction while maintaining sensitivity in the direction of the target signal, and the directional characteristic shown in FIG. Is realized. For specific filter coefficient calculation means, see, for example, Japanese Patent Application
No. 149500.

The SNR improvement effect of the adaptive array is smaller than that of the phase-combined microphone, but the noise can be reduced without realizing sharp directional characteristics, so the SNR can be improved with a small microphone array. There is. Also, the frequency dependency of the SN ratio improvement is smaller than that of the phase synthesis type microphone, and the SN ratio can be improved over a wide frequency band.

However, according to the conventional experimental results, it is known that good characteristics cannot be obtained even in an adaptive array in a low frequency range of several hundred Hz or less (hereinafter, this frequency band is referred to as an extremely low frequency range in this specification). Have been. For example, when the non-distortion constraint condition is used, the effect of improving the S / N ratio in such a low frequency range cannot be obtained, and when the secondary constraint condition is used, the distortion of the target signal in such a low frequency range is not obtained. Has been known to be a problem as a high-quality sound / music sound receiving device.

Further, the adaptive array has a problem that the hardware scale becomes large because the adaptive signal processing accompanying the control of the filter is required.

An object of the present invention is to provide a sound receiving system with a high S / N ratio that solves the problems of the conventional sound receiving system as described above.

[Means for Solving the Problems] The present invention divides a frequency band of an acoustic signal to be received into two and, for example, divides a received signal by two at a frequency fc determined based on a required sound receiver scale. One frequency band (these are referred to as a high frequency band and a low frequency band)
This high frequency range is received by the phase synthesis type sound receiving method,
It is characterized in that the low frequency range is received by an adaptive array system.

In the present invention, given the size of the sound receiver,
A frequency band in which a sharp directional characteristic (a directional characteristic having a sharpness required according to the purpose of use) can be formed by the phase synthesis type sound receiving method is regarded as a high frequency region, and a band lower than that is a low frequency region. As a result, in the high frequency range, a desired SN ratio improving effect can be obtained by the phase synthesis type sound receiving system. On the other hand, by receiving sound in the low frequency range by the adaptive array method, a good improvement in the SN ratio can be obtained except for the extremely low frequency range.

Further, if the applicable frequency bandwidth of the adaptive array system is limited to a low frequency band, the sampling frequency can be reduced, and as a result, there is an advantage that the signal processing amount in the adaptive array system is reduced. This will be described below.

The filter 23 ₁ ~ 23 _M shown in the configuration example of FIG. 8 digital
Consider the case of realizing with an FIR filter. Here, for example, considering an adaptive array having a two-element microphone array, it is necessary to realize at least a delay time of d / c (d: distance between microphone elements, c: sound speed) in each filter. It is known. The number of taps required to realize a certain delay time with the FIR filter is proportional to the sampling frequency. Furthermore, the processing amount of the adaptive array method is proportional to the number of taps of the filter,
Reducing the sampling frequency leads to a reduction in the amount of arithmetic processing.

Next, it will be described that limiting the applicable frequency bandwidth of the adaptive array system has a great advantage in that the SN ratio improvement effect in an extremely low frequency range is improved.

First, the results of a simulation performed to investigate the cause of deterioration of the SN ratio improvement effect in an extremely low frequency range will be described. In the simulation, it was assumed that the target signal arrived from a linear direction connecting the two microphone elements using a two-element microphone array, and that noise came from a direction perpendicular to the straight line. The microphone element spacing d was 4.25 cm. The sampling frequency is 8k
And Hz (application frequency band 4 kHz), the filter 23 ₁ ~ 23 _M shown in the configuration example of FIG. 8 with the 16-tap FIR digital filter. The filter coefficients were calculated using a distortion-free constraint.

The experimental results are shown in FIG. In FIG. 9, curve 27 represents the frequency response of the adaptive array to the target signal, and curve 28 represents the frequency response of the adaptive array to noise. From this figure, in the frequency band of 500Hz or more,
It can be seen that there is a difference of about 20 dB or more between the response to the target signal and the response to noise, and the SN ratio is improved by 20 dB or more. However, it can be seen that the effect of improving the S / N ratio is small in the extremely low frequency range of 500 Hz or less.

 The following considerations can be made from FIG.

a. Since the directivity characteristic becomes omnidirectional in principle for the DC component, the value of the response to the target signal and the value of the response to the noise coincide at the point where the frequency is zero.

b. Deterioration of performance in extremely low frequency range
This is largely due to the influence of the peak starting from the frequency of zero. If the width of this peak is reduced, the performance of reducing the extremely low frequency range is improved.

c. The number of peaks in the response 28 to noise is eight, which is equal to the frequency resolution of the FIR filter in the one-input one-output system, that is, 1/2 of the number of filter taps.

From the above considerations, if the frequency resolution is improved by increasing the number of taps in the FIR filter, the number of peaks in the response 28 to noise increases, and the width of the peaks decreases accordingly, and the low-frequency characteristics can be improved. Is expected. Therefore, the simulation was performed with the number of taps of the filter being doubled (32). The results are shown in FIG. From this figure, it is clear that the frequency bandwidth in which the SN ratio improvement effect is degraded is almost halved, and is almost inversely proportional to the increase in the number of taps of the filter.

Therefore, it was found that the number of taps of the FIR filter should be increased in order to improve the low-frequency characteristics of the adaptive array, but the amount of computation increases in proportion to the number of taps, and the computation accuracy also increases. In many cases, this is difficult in practice.

However, since the frequency resolution of the filter has a relationship of (applied frequency band × 2) ÷ (the number of taps of the filter), the applied frequency band of the adaptive array can be increased without increasing the number of taps as in the present invention. If it is reduced, the frequency resolution is improved, and the performance in the extremely low frequency range is expected to be improved.

To confirm this, set the sampling frequency to 1.
Assuming 6kHz (applicable frequency band 800Hz) and 16 taps
The same simulation as above was performed. The result is
Figure 11 shows. From this figure, the applicable frequency band is 4 kHz (9th
In the case of the figure), the band where the S / N ratio improvement effect is degraded is reduced to 800 Hz from the band of about 500 Hz or less to about 100 Hz.
It can be seen that it has decreased to the following band.

From the above description, it has been clarified that the present invention has a great improvement effect on the performance deterioration in the extremely low frequency range of the adaptive array.

"Embodiment" FIG. 1 shows a first embodiment of the present invention. A phase matching sound receiver 41 and an adaptive array sound receiver 42 are provided, and the output of the phase matching sound receiver 41 is supplied to a high-pass filter 43,
The output of the adaptive array receiver 42 is supplied to a reduction pass filter 44, and the output of the high pass filter 43 and the output of the low pass filter 44 are added by an adder 45 and output as an output signal 46.

The high-pass filter 43 and the low-pass filter 44 have substantially the same cutoff frequency fc. As a result, the output signal
The high frequency range of 46 is synthesized from the output of the phase matching sound receiver 41 and the low frequency range is synthesized from the output of the adaptive array sound receiver 42. In this embodiment, each of the filters 43 and 44 can be incorporated in each of the sound receiving devices 41 and 42 as a part thereof. Thus, the high frequency component is received by the phase synthesis type sound receiving method, and the low frequency component is received by the adaptive array method.

FIG. 2 shows a second embodiment of the present invention. The output of the sound receiver 51 is supplied to a high-pass filter 52 and a low-pass filter 53, and the output of the high-pass filter 52 is supplied to a processing circuit 54 for implementing phase synthesis sound reception.
53 outputs are processing circuits to realize adaptive array sound reception
The output of the processing circuit 54 and the output of the processing circuit 55 are added by an adder 56 and output as an output signal 57.

In this embodiment, first, an acoustic signal is received by one receiver 51, for example, a microphone array. next,
The received signal is divided into two frequency bands by a high-pass filter 52 and a low-pass filter 53 having substantially the same cutoff frequency fc. After that, the processing circuit 54 performs signal processing for performing phase-synthesized sound reception for the high-frequency band, and performs signal processing for performing adaptive array sound reception for the low-frequency band in the processing circuit 55. Finally the processing circuit
The outputs of 54 and 55 are added by adder 56 to synthesize output 57.

Now, given the scale L of the sound receiver, the frequency fc that separates the low frequency region and the high frequency region is determined so as to satisfy the following conditional expression when c is the sound speed.

fc> c / L (1) The minimum frequency fc0 that satisfies this condition is, for example, 680 Hz when the size of the sound receiver is 50 cm, and the directivity shown in FIG. It has characteristics.

If the value of fc is smaller than c / L, the directional characteristics of the phase-synthesized sound reception will be close to non-directionality, and improvement of the SN ratio cannot be expected. On the other hand, if the value of fc is set to a sufficiently large value, the directional characteristics of the phase-synthesized sound reception become sharp, but the bandwidth of the low-frequency band for performing the adaptive array sound reception increases, and the signal processing amount increases. There is a disadvantage that the performance in the frequency range is also deteriorated. Therefore, in practical use, the value of fc is selected from a range that satisfies at least the condition of the expression (1) in consideration of the required sharpness of the directional characteristics, the allowable signal processing amount, and the like.

Next, a processing circuit 54 for realizing phase-synthesized sound reception and a processing circuit 55 for realizing adaptive array sound reception will be described.

As the sound receiver 51, a microphone array in which microphone elements are arranged on a straight line similarly to the microphone array 12 shown in FIG.
As in the case of the drawing, it is assumed that the direction is perpendicular to the straight line in which the microphone elements are arranged. A plurality of signals received by the plurality of microphone elements are output from the sound receiver 51, and supplied to the processing circuits 54 and 55 through a high-pass filter 52 and a low-pass filter 53, respectively.

FIG. 3 shows an example of the configuration of a processing circuit 54 for realizing phase synthesis sound reception. Signals corresponding to the respective sound receiving outputs of the M microphone elements supplied from the high-pass filter 52, 61 _{1 to} 61 _M are added by the adder 63 through the processing filters 62 _{1 to} 62 _M, respectively, and output as an output signal 64. Is output. This processing circuit 54
When the characteristics of the processing filters 62 _{1 to} 62 _M are all 1 (the same characteristic as when no filter is passed), the configuration is the same as the configuration shown in FIG. 6, and the super directional characteristic is realized. When appropriate characteristics are given to the processing filters 62 _{1 to} 62 _M , variations in characteristics of individual microphone elements can be corrected, and fluctuations in the directional characteristic beam width depending on the frequency can be reduced. Further, if an appropriate delay characteristic is given to the processing filters 62 _{1 to} 62 _M , it is possible to control the target direction (direction of the directional characteristic beam).

FIG. 4 shows a configuration example of a processing circuit 55 for realizing an adaptive array sound reception. The signals u _{1 to} u _M corresponding to the respective sound receiving outputs of the M microphone elements supplied from the low-pass filter 53 are subtracted from their adjacent signals by subtractors 72 _{1 to} 72 _M-1 , respectively. These subtractor outputs v _{1 to} v _M−1 are output from the adaptive filter 74.
The difference from the signal u ₁ is supplied to the subtractor 75 through ₁ to 74 _M−1 and output as the output signal 76.

In this configuration example, first, M-1 one subtractors 72 ₁ ~72 _M-1
Synthesizing M signals u ₁ ~u M-1 pieces of signal to remove the phase component contained in the _M v ₁ to v _M-1 used. According to the sound receiving condition, the target signal components included in the _M signals u _{1 to} u _M have the same phase, and the noise components have a phase difference. Therefore, signals v ₁ to
v _M-1 is a signal that does not include the target signal but includes only noise. These signals v _{1 to} v _M-1 are adaptive filters 74 _{1 to} 7 respectively.
4 Input to _M-1 . Outputs of the adaptive filters 74 ₁ to 74 _M−1 are input to a subtractor 75 and subtracted from the signal u ₁ . The output 76 of the subtractor 75 is fed back to the adaptive filters 74 ₁ to 74 _M−1 , and the filter coefficients are corrected so that the root mean square value of the output 76 is minimized. As a correction algorithm, an adaptive algorithm such as a learning identification method or an LMS method is used.

In this adaptive filter operation, minimizing the output 76 corresponds to minimizing the noise component contained in the output 76. Also, because it contains only noise in the signal v ₁ ~v _M-1, the target signal component contained in the signal u ₁ is not affected by subtracting the signal v ₁ ~v _M-1. Therefore, the configuration example of FIG. 4 is an example of a configuration of an adaptive array that minimizes noise under the constraint that the target signal is not distorted at all.

The configuration example of the processing circuit 55 for realizing the adaptive array sound reception has been described above with reference to FIG. 4, but it is also possible to use another adaptive array sound reception method. For example, a method that minimizes noise while allowing a certain range of distortion in the target signal (Kanada: "AMNOR (Adaptive Noise Suppression Microphone Array)")
The IEICE Technical Report, EA85-36, 1985) states that although the amount of computation increases compared to the method of FIG. 4 based on the constraint that the distortion of the target signal is not allowed, the method has a large SN ratio improvement effect. This is an effective method.

[Effects of the Invention] As described above, according to the present invention, a received signal is divided into two frequency bands, a high frequency region is received by a phase synthesis type sound receiving system, and a low frequency region is received by an adaptive array system. It is characterized by sound. According to the present invention, the performance degradation in the low frequency range, which is a disadvantage of the conventional phase synthesis type sound receiving system, and the performance degradation in the extremely low frequency range, which is a disadvantage of the adaptive array system, are improved. Is done. As a result, for example, from several tens of Hz to
High SNR reception over a wide band such as z can be realized.

In addition, since the sampling frequency can be reduced by applying the adaptive array method only to the low-frequency range, a significant reduction in the amount of computation can be achieved compared to the conventional adaptive array that covers the entire frequency band. Have. Further, the size of the sound receiver can be reduced as compared with the conventional phase synthesis type sound receiving method.

Therefore, the present invention can provide a sound receiving system that is very widely used, including application to a high-quality loudspeaker system and a voice recognition system.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention, FIG. 2 is a block diagram showing a second embodiment of the present invention, and FIG.
FIG. 4 is a block diagram showing a configuration example of a processing circuit 54 for realizing a phase-synthesized sound reception in FIG. 2, and FIG. 4 shows a configuration example of a processing circuit 53 for realizing an adaptive array sound reception in FIG. FIG. 5 is a block diagram, FIG. 5 is a diagram of directional characteristics for explaining a conventional sound receiving system, FIG. 6 is a diagram showing a configuration example of a phase synthesis type microphone, and FIG. 7 is phase synthesis using a linear array microphone array. FIG. 8 is a diagram showing a configuration example of the directional characteristics of a rectangular microphone, and FIG.
The figure is a block diagram showing an example of the configuration of an adaptive microphone array. Fig. 9 is an example of the frequency response characteristics of the adaptive microphone array for the target signal and noise (16 taps, 4 kHz bandwidth).
Figure 10 shows an example of the frequency response characteristics of the adaptive microphone array for the target signal and noise (32 taps, band
FIG. 11 is a diagram showing an example (16 taps, band 0.8 kHz) of a frequency response characteristic of the adaptive microphone array to a target signal and noise.

────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Naofumi Inmaki Nippon Telegraph and Telephone Corporation, 1-6-1, Uchisaiwaicho, Chiyoda-ku, Tokyo (56) Reference JP-A-58-151796 (JP, A) JP-A Sho 59-149494 (JP, A) JP-A-2-205200 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) H04R 3/00 320 H04R 1/40 320

Claims (1)

(57) [Claims] 1. An audio signal to be received is divided into two frequency bands. Of the two frequency bands, a target sound wave is received in phase with respect to a high frequency component, and an unnecessary sound wave is received. Is a phase-synthesizing sound receiving system that receives a sound with a phase difference. Out of the above two frequency bands, output of unnecessary sound is maintained while maintaining sensitivity to a target sound. A sound receiving method characterized in that sound is received by an adaptive array method that minimizes noise.