JP2012109799A - Noise suppression type microphone - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a noise suppression type microphone by which an excellent narrow directional can be obtained, a S/N ratio of voice uttered in a noisy location at a distance is improved, the number of used microphones is reduced to be suitable for miniaturization.SOLUTION: A directional characteristic is made smaller than a conventional directional microphone by multiplying a microphone output by a coefficient calculated from a power spectrum value of the output of a directional microphone having maximum sensitivity in a direction toward a target sound and an ambient power spectrum value, thereby improving a S/N ratio of the target sound. The microphone does without increasing the number of elements of the microphone in order to sharpen the directivity as performed in the conventional microphone.

Description

  The present invention relates to a noise suppression type microphone unit using a plurality of microphones. For example, ambient noise and reverberation such as a video conference hands-free telephone, a car navigation voice input, a voice recognition device input device, a robot hearing sensor, etc. The present invention relates to a technique that is effective when applied to input of an audio signal with reduced sound.

  With recent advances in speech recognition technology, the realization of a “human-machine interface” that can be easily used by anyone, anywhere is expected, and a microphone as a sound entrance is also required to be highly convenient. It was.

  In general, when collecting sound, if the microphone is located away from the target sound source, clear sound collection becomes difficult due to the influence of ambient noise and reverberation sound. This is due to the fact that when the microphone is moved away from the mouth, the direct sound ratio of the voice is rapidly reduced, and the ratio of noise and reverberant sound is increased. Headsets and phones that always have a microphone close to the mouth have a very high direct sound ratio, so those that didn't bother with ambient noise or reverberant sound could be worn on their body or in their hands. In the “human-machine interface” that cannot be held, noise and reverberation due to separation of sound sources become a problem.

  The sound collection problem in such a scene is not so simple that it can be solved simply by changing the omnidirectional microphone to a microphone having sound direction selectivity such as a unidirectional microphone. Traditional techniques such as the old method of obtaining a sharper directivity using a long acoustic tube to remove noise and reverberation, and the recent array microphone that performs digital processing by arranging multiple microphones are large in size. However, there are still problems such as increasing the system scale because of a large amount of computation and high-speed processing, and it is difficult to say that it is easy to use.

  As conventional techniques for obtaining directivity characteristics, there are beam forming techniques described in Non-Patent Documents 1 and 2 and Patent Document 1.

JP 2004-279390 A

-IEICE technical report EA98-78-100, pp.57-62, Nov.1998 Kazuhiko Kawahara, Kimitoshi Fukudome "Comparison of broadband beamforming characteristics between spherical baffle array and circular array" Journal of the Acoustical Society of Japan VOL.62 NO.10 2006 pp.726−737 Hiroshi Nakajima “Realization of minimum beamforming with average sidelobe energy using indefinite terms”

  The beam forming technique described in Non-Patent Document 1 provides excellent directivity when the number of microphone elements used is large, but if the number of elements is small, side lobes in the direction opposite to the main lobe direction can be suppressed. The desired directivity cannot be obtained.

  Non-Patent Document 2 and Patent Document 1 attempt to reduce the side lobe while keeping the main lobe narrow, but although the improvement of the low frequency band can be obtained, the desired directivity characteristic is obtained when the frequency band is 1500 Hz or higher. It becomes difficult.

  In order to remove directional noise and reverberation, a microphone with sharper directivity is required, which can be dealt with by increasing the number of microphone elements, but the number of elements increases. In some cases, the shape becomes so large that it is impossible to mount or mount the device.

  An object of the present invention is to achieve an excellent narrow directivity, to improve the S / N ratio of speech uttered in a distant and noisy place, and to reduce the number of microphones used. Therefore, it is possible to provide a noise suppression type microphone suitable for the above. In other words, a miniaturized noise suppression type microphone suitable for sound source separation of a plurality of speakers and sensing of a plurality of sound sources is provided.

  The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

  The following is a brief description of an outline of typical inventions disclosed in the present application.

  In other words, by multiplying the microphone output by the coefficient calculated from the power spectrum value of the output of the directional microphone that has the maximum sensitivity in the direction toward the target sound and the surrounding power spectrum value, it is more directional than the conventional directional microphone. Narrow the characteristics to improve the S / N ratio of the target sound. Conventionally, it is not necessary to cope with the problem by increasing the number of microphone elements in order to sharpen the directivity.

  The effects obtained by the representative ones of the inventions disclosed in the present application will be briefly described as follows.

  In other words, excellent narrow directivity can be obtained, the S / N ratio of speech uttered in a distant and noisy place can be improved, and it is suitable for downsizing that requires a small number of microphones. A noise suppression type microphone can be realized.

FIG. 1 is an explanatory view illustrating a first embodiment of a noise suppression type microphone according to the present invention. FIG. 2 is an explanatory diagram illustrating the second embodiment of the noise suppression type microphone according to the present invention. FIG. 3 is a characteristic diagram illustrating the directivity characteristics of the directional microphone of FIG. FIG. 4 is a characteristic diagram that presents directional characteristics of three omnidirectional microphones. Figure 5 is a characteristic diagram showing the directional characteristics of the improved output S _k of the resulting radiation pattern as the effect of the first embodiment. 6 is a characteristic comparison diagram comparing the characteristics before and improved directional characteristics S _k of the directional improvements output. FIG. 7 is a characteristic diagram showing the directivity characteristics of the directivity improvement output SU _k obtained as an effect of the second embodiment. FIG. 8 is a comparative explanatory diagram comparing the directivity obtained by the conventional beam forming technique and the directivity of the microphone according to the present invention by using the same number of three microphones.

1. First, an outline of a typical embodiment of the invention disclosed in the present application will be described. Reference numerals in the drawings referred to in parentheses in the outline description of the representative embodiments merely exemplify what are included in the concept of the components to which the reference numerals are attached.

  [1] A noise suppression type microphone according to a typical embodiment of the present invention includes a plurality of directional microphones (M1, M2) each having a maximum sensitivity in different directions, and the plurality of directional microphones. Inputs the audio signal to be output, and generates the target audio signal by processing the input audio signal to suppress the audio and noise from other sound sources for the audio in the maximum sensitivity direction of the specified directional microphone And a data processing unit (1). The data processing unit adds the first power spectrum value of the received signal of the predetermined directional microphone, the power spectrum of the received signal of each of the other directional microphones, and the first power spectrum value. An omnidirectional power spectrum value, a coefficient related to noise control that is the inverse of the component of the omnidirectional power spectrum value, a spectrum value of a received signal of the predetermined directional microphone, and the first power spectrum value The first product data and the second product data of the first product data and the coefficient related to the noise control are calculated, and the second product data is output as the target speech signal. Used for.

  [2] The noise suppression type microphone according to [1], wherein the plurality of directional microphones include two directional microphones, and a main axis of the predetermined directional microphone is directed in a direction of a target sound.

  [3] A noise suppression type microphone according to another embodiment of the present invention receives a plurality of omnidirectional microphones (M3, M4, M5) and an audio signal output from the plurality of omnidirectional microphones. Then, the input audio signal is subjected to a process for suppressing audio and noise from other sound sources for audio in a predetermined direction on a straight line obtained by connecting two omnidirectional microphones, and the target audio signal A data processing unit (11) for generating The data processing unit includes a first power spectrum value of a received signal from the predetermined direction of a directional microphone obtained from two omnidirectional microphones, and a directional microphone obtained from another omnidirectional microphone. An omnidirectional power spectrum value obtained by summing up the respective power spectra of the received sound signal and the first power spectrum value, a coefficient relating to noise control that is the inverse of the component of the omnidirectional power spectrum value, and the two The first product data of the spectrum value of the received signal from the predetermined direction of the directional microphone obtained from the omnidirectional microphone and the first power spectrum value, the data of the first product, and the A second product data with a coefficient related to noise control is calculated, and the second product data is used to output the target speech signal.

  [4] The noise suppression type microphone according to item 3, wherein the plurality of omnidirectional microphones include three omnidirectional microphones, and corresponds to each side of a triangle having the three omnidirectional microphones as vertices. The direction of one straight line is the direction of the target sound in the predetermined direction, and the direction of the straight line corresponding to the other side is the direction of the non-target sound.

  [5] In the noise suppression type microphone according to any one of Items 1 to 4, the data processing unit includes a digital signal processor, a ROM holding a program executed by the digital signal processor, and a digital signal processor. RAM including a work area is included.

2. Details of Embodiments Embodiments will be further described in detail.

Embodiment 1
FIG. 1 illustrates an embodiment of a noise suppression type microphone according to the present invention. The noise suppression type microphone shown in the figure includes, for example, two primary gradient type unidirectional microphones M1 and M2 having directivity characteristics shown in FIG. 3, and the primary gradient type unidirectionality. The sound signals output from the microphones (also simply referred to as unidirectional microphones) M1 and M2 are input, and the maximum sensitivity direction of the specific unidirectional microphone M1 with respect to the input sound signals (the direction of the main axis of the directivity) ) Voice (target sound source) is processed to suppress the sound and noise from other sound sources to generate a target sound signal.

  The data processing unit 1 includes a digital signal processor (DSP) 3, a RAM 4 used in a work area such as the DSP 3, and an analog / digital conversion circuit (converting analog signals output from the unidirectional microphones M 1 and M 2 into digital signals). ADC) 5, digital-analog conversion circuit (DAC) 2 that converts audio data digitally processed by DSP 3 into analog signals and outputs, and ROM 6 that holds programs executed by DSP 3, etc. Consists of multiple chips. The DSP 3 executes a program stored in the ROM 6 to perform data processing for generating the target audio signal. Hereinafter, the processing will be described.

  In the example of FIG. 1, the directional main axis of the unidirectional microphone M1 is directed to the target sound source, and the directional main axis of the unidirectional microphone M2 is directed opposite to the target sound source. The direction is made different. The characteristic of FIG. 3 draws a cardioid curve in polar coordinates.

Assume that the sample value of the received signal of the unidirectional microphone M1 is {x1 _n } and the sample value of the received signal of the unidirectional microphone M2 is {x2 _n }. n = 0, ..., N-1. Then, the power spectrum value of the received sound signal of each microphone is represented by | X1 _k , where {x1 _n } is represented by {X1 _k } and {x2 _n } is represented by {X2 _k }. | ² and | X2 _k | ² . Here, k = 0,..., N−1. Since the two unidirectional microphones M1 and M2 arranged as shown in FIG. 1 uniformly receive noise coming from all directions, the power spectrum values | X1 _k | ² and | X2 _k | PS _{k in} equation (1), which is the sum of ² , corresponds to power spectrum values in all directions.

Furthermore, the omnidirectional power spectrum value PS _{k ˜} obtained by adding the adjustment parameter α of the ambient noise to PS _k is expressed by Expression (2). α is a predetermined constant and a real value of 0 <α <10.

A value represented by Expression (3), which is composed of the reciprocal of a non-zero component of PS _k˜ of Expression (2), is _{defined as} a coefficient H _k for noise control. In equation (3), k = 0,..., N−1, and β is a real value sufficiently smaller than 1.

Finally, a desired transfer function capable of obtaining an output signal of a microphone having a target directivity from the received sound signal X1 _k is set as G _k , and the microphone output S _k at that time is _obtained by Expression (4). k = 0,..., N-1.

Here, as shown in Expression (5), G _k is a power spectrum value of the received signal in the target direction and a coefficient H _k related to noise control obtained from the power spectrum value of the received signal in the surrounding non-target direction, It is a product of.

The output signal S _k thus obtained has a characteristic as shown in FIG. 5 in which the directivity characteristic of the sound reception signal X1 _k of the unidirectional microphone M1 is improved.

It is shown in comparison with directional characteristics of the directional improvements output signal S _k, before improving the characteristics of the unidirectional microphone M1 in FIG. Directional characteristics of the output signal S _k as shown has been greatly improved.

<< Embodiment 2 >>
Next, a noise suppression microphone according to a second embodiment shown in FIG. 2 will be described. The noise suppression type microphone shown in the figure includes, for example, a plurality of, for example, three omnidirectional microphones M3, M4, and M5, and an audio signal output from the plurality of omnidirectional microphones M3, M4, and M5. A process for suppressing sound and noise from other sound sources with respect to sound in a predetermined direction (target sound source direction) on a straight line obtained by connecting two omnidirectional microphones to the input sound signal And a data processing unit 11 for generating a target audio signal.

  The data processing unit 11 is an analog-to-digital conversion circuit that converts analog signals output from the RAM 14 and the omnidirectional microphones M3, M4, and M5 used in work areas such as the digital signal processor (DSP) 13 and the DSP 13 into digital signals. (ADC) 15, a digital-analog converter circuit (DAC) 12 that converts the audio data digital signal processed by the DSP 13 into an analog signal and outputs the analog signal, and a ROM 16 that holds a program executed by the DSP 13. It consists of a chip or multichip. The DSP 13 executes a program stored in the ROM 16 to perform data processing for generating the target audio signal. Hereinafter, the processing will be described.

  Here, for example, three omnidirectional microphones M3, M4, and M5 are used. In FIG. 2, among the three microphones M3, M4, and M5, a line segment A connecting M3 and M4, a line segment B connecting M4 and M5, and a line segment C connecting M5 and M3 are in different directions. In addition, one of the line segments is arranged so that the target sound source is always located on the extension line. In the example of FIG. 2, each line segment opens at an arbitrary angle, the target sound source exists on the extension line of the line segment A connecting the microphone M3 and the microphone M4, and on the extension lines of the other line segments B and C. It is assumed that there is noise or a sound source other than the intended purpose. The arrival direction of noise and the target direction are different.

As before, the sample value of the received signal of the microphone M3 is {x3 _n }, the sample value of the received signal of the microphone M4 is {x4 _n }, and the sample value of the received signal of the microphone M5 is {x5 _n }. In addition, a directional microphone having a characteristic in which the main axis is directed toward the target sound source is formed by a combination of the microphones M3 and M4. Here, the sound reception signal x4 _n of the microphone M4 is obtained through a delay circuit for providing a delay having a time value equal to or less than the sound wave propagation time of the distance between the two microphones M3 and M4. A sample value {u3 _n } of a new received sound signal obtained by taking a difference between the received delayed signal x4 ′ _n and the received sound signal x3 _n of the microphone M3 is obtained as Expression (6).

  The directional microphone obtained in this way has characteristics as represented by 4a in FIG.

The received sound signal of the second directional microphone is a combination of microphones M4 and M5 and the main directivity axis rotates counterclockwise with {u3 _n }. The same delay as described above is added to microphone M5. ones as x5 _'n, obtained with the new sample value taken the difference between x4 _n {u4 _n}, thereby having a characteristic as represented by 4b in Fig.

The received sound signal of the third directional microphone is a combination of microphones M5 and M3 whose main axis of directivity rotates clockwise by u4 _n and {u3 _n }, and is the same as described above for microphone M3. the plus delay as x3 _'n, obtained from the new sample values taken the difference between x5 _n {u5 _n}, consists in having the characteristic as represented by 4c in FIG. The sound reception signals u4 _n and u5 _n at this time are _expressed by equations (7) and (8). n = 0, ..., N-1.

Then the sample values of the received sound signals _{{u3 n}, {u4 n} }, {u5 n}, the same processing supra from.

The representation of the time-frequency domain of _{{u3 n} {U3 k}} , the representation of time-frequency domain of _{{u4 n} {U4 k}} , the representation of time-frequency domain of _{{u5 n} {U5 k}} ( where k = 0,..., N−1), the received sound signals {U3 _k } and {U4 _k } of the directional microphones made from the three omnidirectional microphones M3, M4 and M5 arranged as shown in FIG. , The direction of the principal axis of {U5 _k } is determined so as to uniformly receive noise coming from all directions, and the power spectrum value is calculated from the received signals of the respective directional microphones, and the sum thereof is calculated. The power PU _{k in} all directions is obtained. The respective power spectrum values are set as | U3 _k | ² , | U4 _k | ² , | U5 _k | ² , and the power spectrum value PU _k in all directions is _obtained as in Expression (9).

Furthermore, a power spectrum in all directions obtained by adding an adjustment parameter α _u of ambient noise to PU _k can be expressed as Equation (10) as PU _k˜ . α _u is a predetermined constant and a real value of 0 <α _u <10.

A value represented by Expression (11), which is composed of the reciprocal of a non-zero component of PU _k , is set as a coefficient HU _k for noise control. In equation (11), k = 0,..., N−1, and βu is a real value sufficiently smaller than 1.

Finally, a desired transfer function capable of obtaining an output signal of a microphone having a target directivity from the received sound signal U3 _{k is} defined as GU _k , and the microphone output SU _k at that time is _obtained by Expression (12). k = 0,..., N-1.

Here, GU _k is the product of the power spectrum value of the received signal in the target direction and the coefficient HU _k related to noise control obtained from the power spectrum value of the received signal in the surrounding non-target direction. Indicated.

The output signal SU _k obtained in this way has a characteristic as shown in FIG. 7 in which the directivity characteristic of the sound reception signal U3 _k of the directional microphone is improved. FIG. 8 shows a comparison between the directivity obtained by the conventional beam forming technique and the directivity of the microphone according to the present invention using the same number of three microphone elements.

  In the above description, the main part of the mathematical expression is the representation of the time frequency domain, but it is natural that a calculation having the same meaning can also be performed in the time domain. For example, the product of two signals obtained by FFT (high-speed Fourier transform) may be obtained as a convolution operation of each signal. These may be appropriately processed according to the scale of the system to be constructed and the calculation time. The result obtained in the time frequency domain can be converted into a time domain waveform by inverse Fourier transform and used.

  In the above description, the number of microphones is 2 to 3. However, the number may be increased within the range allowed for the size of the apparatus when the present invention is implemented. It has already been demonstrated that a sharper beam directivity can be obtained by increasing the number of microphones.

Figure 5 shows the directional characteristics of the improved output S _k of the resulting radiation pattern as the effect of the first embodiment. Figure 6 is a comparison with the directivity characteristic S _k of the directional improvements output, before improving the properties, in comparison to the directional characteristics of the unidirectional microphone M1 before improvement, 4.8 at an incident angle of 60 ° of the wave The main lobe is improved by 11.7 dB at 120 ° and 90 dB at 120 °.

FIG. 7 shows the directivity characteristics of the directivity improvement output SU _k obtained as an effect of the second embodiment. FIG. 8 shows a comparison between the directivity obtained by the conventional beam forming technique and the directivity of the microphone according to the present invention by using the same number of three microphones. The main lobe improvement of dB is reduced by 22 dB or more at 180 °.

  In the conventional beam forming technique, the means for improving the directivity has been achieved by increasing the number of microphones, but in the present invention, it can be realized with a small number of microphones, which is very useful for downsizing of the apparatus.

  Although the invention made by the present inventor has been specifically described based on the embodiments, it is needless to say that the present invention is not limited thereto and can be variously modified without departing from the gist thereof.

  For example, the directional microphone may have a directivity of either a primary pressure gradient type bi-directional or unidirectional, or may be a higher-order pressure gradient microphone. In the first embodiment, the characteristic of the directional microphone may be a directional characteristic obtained with an acoustic tube.

  In the configuration using the omnidirectional microphones described in the second embodiment, the number of omnidirectional microphones is not limited to three, and a larger number can be adopted. In consideration of miniaturization of the noise suppression type microphone, for example, a number of up to nine may be employed. For example, a noise suppression type microphone has 4 to 9 omnidirectional microphones as a plurality of omnidirectional microphones, and connects two to four of the 4 to 9 selected microphones. The direction of the straight line is the direction of the target sound in the predetermined direction, and the direction of the straight line connecting the two to four microphones selected from the remaining microphones is the direction of the non-target sound.

  The data processing unit can include a central processing unit.

M1, M2 Unidirectional microphone M3, M4, M5 Omnidirectional microphone 1,11 Data processing unit 2,12 Digital analog conversion circuit (DAC)
3,13 Digital signal processor (DSP)
4,14 RAM
5,15 Analog to digital converter (ADC)
6,16 ROM

Claims (5)

A plurality of directional microphones with maximum sensitivities directed in different directions and audio signals output from the plurality of directional microphones are input, and the maximum sensitivity direction of a predetermined directional microphone is input to the input audio signals. A data processing unit that generates a target voice signal by performing a process of suppressing voice and noise from other sound sources to the voice of
The data processing unit adds the first power spectrum value of the received signal of the predetermined directional microphone, the power spectrum of the received signal of each of the other directional microphones, and the first power spectrum value. An omnidirectional power spectrum value, a coefficient related to noise control that is the inverse of the component of the omnidirectional power spectrum value, a spectrum value of a received signal of the predetermined directional microphone, and the first power spectrum value The first product data and the second product data of the first product data and the coefficient related to the noise control are calculated, and the second product data is output as the target speech signal. Noise suppression microphone used for   The noise suppression type microphone according to claim 1, wherein the plurality of directional microphones include two directional microphones, and the predetermined directional microphone has a main axis directed in a direction of a target sound. A plurality of omnidirectional microphones and an audio signal output by the plurality of omnidirectional microphones are input, and a predetermined line on a straight line obtained by connecting two omnidirectional microphones to the input audio signal A data processing unit that generates a target voice signal by performing processing to suppress voice and noise from other sound sources with respect to the voice in the direction;
The data processing unit includes a first power spectrum value of a received signal from the predetermined direction of a directional microphone obtained from two omnidirectional microphones, and a directional microphone obtained from another omnidirectional microphone. An omnidirectional power spectrum value obtained by summing up the respective power spectra of the received sound signal and the first power spectrum value, a coefficient relating to noise control that is the inverse of the component of the omnidirectional power spectrum value, and the two The first product data of the spectrum value of the received signal from the predetermined direction of the directional microphone obtained from the omnidirectional microphone and the first power spectrum value, the data of the first product, and the A noise suppression type that calculates data of a second product with a coefficient related to noise control, and uses the data of the second product for output of the target speech signal Ikurohon.   As the plurality of omnidirectional microphones, there are three omnidirectional microphones, and the direction of one straight line corresponding to each side of a triangle having the three omnidirectional microphones as apexes is the target sound in the predetermined direction. The noise suppression type microphone according to claim 3, wherein a direction of a straight line corresponding to the other side is a direction of a non-target sound.   5. The data processing unit according to claim 1, wherein the data processing unit includes a digital signal processor, a ROM holding a program executed by the digital signal processor, and a RAM serving as a work area of the digital signal processor. A noise suppression type microphone according to claim 1.

Images