JP4286637B2 - Microphone device and playback device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a microphone unit capable of stably operating even under a plurality of noises in the practical environment and of realizing a high S/N. <P>SOLUTION: A signal generating part generates a main signal and a noise reference signal. A judgement part judges whether or not a level ratio is greater than a predetermined value. An adaptive filter part generates a signal indicating a signal component of a target sound included in the noise reference signal generated by the signal generating part and learns a filter coefficient only when it is judged by the judgement part that the level ratio is greater than the predetermined value. A subtraction part subtracts the signal generated by the adaptive filter part. A noise suppressing part uses the main signal and the noise reference signal after subtraction by the subtraction part to suppress a signal component of noise contained in the main signal. <P>COPYRIGHT: (C)2004,JPO&amp;NCIPI

Description

  The present invention relates to a microphone device and a sound reproduction device, and more specifically, a microphone device and a sound reproduction device that detect noise coming from a predetermined direction while suppressing noise.

A configuration of a conventional microphone device will be described with reference to FIGS.
FIG. 24 is a diagram showing a configuration of the microphone device of the first conventional example. In FIG. 24, the microphone device includes a first microphone unit 1010, a second microphone unit 1020, a signal adding unit 1030, a first signal subtracting unit 1031, a signal amplifying unit 1050, and an adaptive filter unit 1060. , And a second signal subtracting unit 1062. Both microphone units 1010 and 1020 are arranged to face the front direction (left direction in FIG. 24). The signal adding unit 1030 adds the signal output from the first microphone unit 1010 and the signal output from the second microphone unit 1020. The first signal subtracting unit 1031 subtracts the signal output from the second microphone unit 1020 from the signal output from the first microphone unit 1010. The signal amplification unit 1050 multiplies the signal output from the signal addition unit 1030 by a factor of two. The adaptive filter unit 1060 receives the signal output from the first signal subtracting unit 1031 and outputs a signal filtered by the adaptive filter. The second signal subtraction unit 1062 subtracts the signal output from the adaptive filter unit 1060 from the signal output from the signal amplification unit 1050. The output from the second signal subtracting unit 1062 becomes the output of the microphone device. The adaptive filter unit 1060 learns filter coefficients based on the signal output from the second signal subtracting unit 1062 and the signal output from the first signal subtracting unit 1031.

  Next, the operation of the microphone device of Conventional Example 1 will be described. When detecting sound coming from the front direction, the microphone units 1010 and 1020 output substantially equal signals. Further, when detecting sound coming from a direction other than the front direction, the microphone units 1010 and 1020 output signals having different phases. Output signals from the microphone units 1010 and 1020 are added by a signal adding unit 1030. The level of the added signal is normalized by the signal amplification unit 1050, that is, the amplitude is halved. As described above, a main signal having a sound component coming from the front direction can be obtained. On the other hand, by the output from the first signal subtracting unit 1031, the main directivity axis is oriented in the direction of 90 degrees with respect to the front direction, and the front direction becomes a blind spot of directivity (that is, the front direction is the minimum directivity). Directivity characteristics such as a sensitivity direction can be obtained. That is, the signal output from the first signal subtraction unit 1031 is a noise reference signal that does not include a sound component coming from the front direction. The adaptive filter unit 1060 realizes adaptive directivity by using the main signal output from the signal amplification unit 1050 and the noise reference signal output from the first signal subtraction unit 1031. That is, a directivity blind spot is automatically formed with respect to a noise source in one direction coming from other than the front direction.

  FIG. 25 is a diagram showing a configuration of a microphone device of Conventional Example 2. In FIG. 25, the microphone device includes a first microphone unit 1010, a second microphone unit 1020, a first adaptive filter unit 1040, a first signal delay unit 1041, and a first signal subtracting unit 1042. , A second adaptive filter unit 1060, a second signal delay unit 1061, and a second signal subtraction unit 1062.

  The first adaptive filter unit 1040 receives the output signal from the second microphone unit 1020 as an input and outputs the filtering result by the adaptive filter. The first signal delay unit 1041 delays the signal output from the first microphone unit 1010. The first signal subtraction unit 1042 subtracts the signal output from the first adaptive filter unit 1040 from the signal output from the first signal delay unit 1041. The first adaptive filter unit 1040 learns filter coefficients based on the signal output from the first signal subtracting unit 1042 and the signal output from the second microphone unit 1020. The second signal delay unit 1061 gives a delay to the signal output from the first signal delay unit 1041. The second adaptive filter unit 1060 receives the signal output from the first signal subtracting unit 1042 as an input and outputs a filtering result by the adaptive filter. The second signal subtracting unit 1062 subtracts the signal output from the second adaptive filter unit from the signal output from the second signal delay unit 1061 to obtain the output of the microphone device. The second adaptive filter unit 1060 learns filter coefficients based on the signal output from the second signal subtracting unit 1062 and the signal output from the first signal subtracting unit.

  The operation of the microphone device of Conventional Example 2 will be described below. The first adaptive filter unit 1040, the first signal delay unit 1041, and the first signal subtraction unit 1042 in the conventional example 2 perform a cancel operation on the sound waves that have arrived at the microphone units 1010 and 1020. . That is, the signal output from the first signal subtraction unit 1042 is a noise reference signal for the second adaptive filter unit 1060. The signal output from the first signal subtracting unit 1042 is a target signal similar to the signal output from the first signal subtracting unit 1031 illustrated in FIG. However, the conventional example 1 has a fixed directivity, whereas the conventional example 2 is different in that the directivity can be changed by using an adaptive filter.

  FIG. 26 is a diagram showing a configuration of a microphone device of Conventional Example 3. 26 includes a first unidirectional microphone unit 1011, a second unidirectional microphone unit 1012, a first FFT unit 1070, a second FFT unit 1080, and two inputs. A type spectrum subtraction unit 1090 and a speech recognition unit 2000 are provided.

  In FIG. 26, the first unidirectional microphone unit 1011 is arranged such that the directional main axis faces the front direction. The second unidirectional microphone unit 1012 is arranged such that the directional main axis faces the back direction. The first FFT unit 1070 obtains a frequency spectrum using the signal output from the first unidirectional microphone unit 1011 as an input. The second FFT unit 1080 obtains a frequency spectrum using the signal output from the second unidirectional microphone unit 1012 as an input. The 2-input type spectrum subtraction unit 1090 receives the signals output from the FFT units 1070 and 1080 as inputs, and the signal spectrum derived by the second FFT unit 1080 from the signal spectrum derived by the first FFT unit 1070. Is subtracted in the power spectrum region to output the spectrum of the target signal. The voice recognition unit 2000 performs voice recognition using the spectrum of the target signal output from the two-input spectrum subtraction unit 1090 as an input.

The operation of the microphone device of Conventional Example 3 will be described below. In Conventional Example 3, the first unidirectional microphone unit 1011 has directivity characteristics for collecting the target sound in the front direction. The second unidirectional microphone unit 1012 has a directivity characteristic that mainly collects noise. As a result, the main signal m1 is obtained from the first unidirectional microphone unit 1011 and the noise reference signal m2 is obtained from the second unidirectional microphone unit 1012. In each of the FFT units 1070 and 1080, spectra of the main signal m1 and the noise reference signal m2 are obtained. In the 2-input spectrum subtraction unit 1090, the power spectrum of the signal component is estimated by subtracting the power spectrum of the noise reference signal from the power spectrum of the main signal. In the one-input type spectral subtraction method, the noise spectrum is estimated on the assumption that the noise is stationary in the time interval in which the target sound does not arrive. Therefore, in the one-input type spectral subtraction method, only stationary noise can be suppressed. On the other hand, according to the configuration of the conventional example 3 employing the two-input type spectral subtraction method, the spectrum of the noise reference signal can always be obtained by the second unidirectional microphone unit 1012. Noise can be suppressed. As described above, according to the microphone device of Conventional Example 3, it is possible to improve the speech recognition rate of the subsequent speech recognition unit 2000 by suppressing not only stationary noise but also non-stationary noise. Note that the apparatus shown in FIG. 26 uses voice recognition. Here, by performing IFFT in the final stage, the spectrum can be returned to a time signal, and a waveform signal can be obtained while performing frame overlap.
Japanese Patent No. 3084433 Bernard Widrow, “ADAPTION SIGNAL PROCESSIN”, Prentice Hall, 1985, p. 414, 419, 423 Nakadai, Tsunemura, Nakatsu, “Voice recognition in a car using a noise reduction technique with two inputs”, IEICE Technical Report, 1989, SP89-81, pp. 41-48

  In the configuration of the conventional example 1 described above, a large noise suppression effect can be obtained in an environment where noise comes from one direction. However, the device of Conventional Example 1 cannot cope with noise coming from a plurality of directions. Therefore, in an actual noise environment in which noise sources are simultaneously present in various directions, the configuration of the conventional example 1 obtains only a noise suppression effect equivalent to the performance of a conventionally used unidirectional microphone device. I can't.

  In the configuration of Conventional Example 2, the noise reference signal is obtained by using the first adaptive filter. Here, in order to stably operate the first adaptive filter in the actual environment, it is necessary to learn the first adaptive filter only when the voice from the speaker is sufficiently louder than the surrounding noise. Therefore, with the configuration of Conventional Example 2, the noise suppression effect cannot be obtained until the convergence of the filter is completed. In addition, it is difficult to converge the filter in a noisy environment. Further, like the conventional example 1, the configuration of the conventional example 2 cannot cope with a plurality of noise sources. Further, since the device of the conventional example 2 is invented for the purpose of suppressing wind noise having no correlation between unit signals, the direction of the target sound cannot be limited. That is, the loudest sound among the arriving sounds becomes the target sound, and the sound in a specific direction cannot be emphasized and collected.

  The configuration of Conventional Example 3 is a system that converts a main signal and a noise reference signal into a spectrum and suppresses noise using a spectrum subtraction method in a power spectrum. This method is a method capable of simultaneously suppressing noise even when there are noise sources in a plurality of directions. However, this method has a problem that even if the target sound is mixed into the noise reference signal, a large problem occurs in the sound quality of the processed sound, or the target sound itself is canceled. Further, in an actual sound field, it is conceivable that reflected waves wrap around and enter even when the directional blind spot of the unidirectional microphone unit is directed to the target sound direction. Furthermore, since a normal microphone unit is not an infinite attenuation of directivity but an attenuation of about 10 to 15 db, the direct wave of the target sound may not be completely removed and may be mixed into the noise reference signal. is there. In addition, in the case of the spectrum subtraction method, a processing delay due to frame processing occurs, so that there is a problem that it cannot be used for applications such as simultaneous calls and loudspeakers.

  Further, the above-described conventional example focuses on suppressing additive noise, which is noise different from the target sound. In the conventional example described above, multiplicative noise that arrives after the target sound is reflected on a reflecting surface such as a wall, desk, or floor cannot be removed. Therefore, the frequency characteristics of the target sound may be distorted due to the influence of reflection or the like in the sound field where the microphone device is actually used. For this reason, particularly in applications such as speech recognition, it has not been possible to solve the problem of misrecognition that causes inconsistency in matching during recognition.

  SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a microphone device that can stably operate even under a plurality of noises in an actual use environment and can realize a high S / N.

  Another object of the present invention is to provide a microphone device that suppresses both multiplicative noise caused by reflected waves of the target sound and additive noise caused by noise.

  Another object of the present invention is to generate a main signal and a noise reference signal used in a process for suppressing noise by a simple method.

In order to achieve the above object, the present invention employs the following configuration. That is, the first invention is a microphone device that detects a target sound coming from a target sound direction. The microphone device includes a signal generation unit, a determination unit, an adaptive filter unit, a subtraction unit, and a noise suppression unit. The signal generation unit generates a main signal indicating a result detected with sensitivity to the target sound direction , and a noise reference signal indicating a result detected with the sensitivity blind spot directed toward the target sound direction. The determination unit determines whether or not a level ratio indicating a ratio of the signal level of the main signal to the signal level of the noise reference signal generated by the signal generation unit is greater than a predetermined value. The adaptive filter unit generates a signal indicating the signal component of the target sound included in the noise reference signal generated by the signal generation unit by filtering the main signal generated by the signal generation unit with an adaptive filter, and determines Only when it is determined by the unit that the level ratio is larger than a predetermined value, the filter coefficient is learned. Subtraction unit, the noise reference signal, generated by the adaptive filter subtracts the signal indicating the signal components of the target sound included in the noise reference signal. The noise suppression unit suppresses a signal component of noise included in the main signal using the main signal and the noise reference signal after subtraction by the subtraction unit. The noise suppression unit includes a noise suppression filter coefficient calculation unit and a time-varying coefficient filter unit. The noise suppression filter coefficient calculation unit calculates a filter coefficient of a noise suppression filter for suppressing signal components other than the target sound signal from the main signal based on the main signal and the noise reference signal after subtraction by the subtraction unit. . The time-varying coefficient filter unit performs filtering on the main signal, reflecting the filter coefficient calculated by the noise suppression filter coefficient calculation unit.

  The “main signal indicating the result of detection with sensitivity to the target sound direction” is not only the signal itself output from the microphone unit, but also a predetermined process applied to the signal detected by the microphone unit. This also includes the signal obtained as a result. That is, the main signal may be a signal itself output from a microphone unit having a directional main axis directed to the target sound direction, or may be a microphone unit (which may be omnidirectional or in a predetermined direction). It may be a signal obtained by processing a signal output from a directivity main axis may be directed. Similarly, the “noise reference signal indicating the result of detecting the sound coming from other directions with higher sensitivity than the target sound” may be the signal itself output from the microphone unit or the microphone unit. It may be a signal obtained by processing the signal output from.

The second invention is a microphone device for detecting a target sound coming from a target sound direction. The microphone device includes a signal generation unit, a determination unit, an adaptive filter unit, a subtraction unit, a reflection information calculation unit, and a reflection correction unit. The signal generation unit generates a main signal indicating a result detected with sensitivity to the target sound direction , and a noise reference signal indicating a result detected with the sensitivity blind spot directed toward the target sound direction. The determination unit determines whether or not a level ratio indicating a ratio of the signal level of the main signal to the signal level of the noise reference signal generated by the signal generation unit is greater than a predetermined value. The adaptive filter unit generates a signal indicating the signal component of the target sound included in the noise reference signal generated by the signal generation unit by filtering the main signal generated by the signal generation unit with an adaptive filter, and determines Only when it is determined by the unit that the level ratio is larger than a predetermined value, the filter coefficient is learned. Subtraction unit, the noise reference signal, generated by the adaptive filter subtracts the signal indicating the signal components of the target sound included in the noise reference signal. The reflection information calculation unit calculates information related to the arrival time difference between the direct wave and the reflected wave of the target sound based on the filter coefficient of the adaptive filter unit. The reflection correction unit corrects the distortion of the frequency characteristics generated in the main signal due to the reflected wave of the target sound based on the information calculated by the reflection information calculation unit.

In the third invention, the signal generation unit includes a first microphone unit and a second microphone unit. The first microphone unit is arranged with the directional main axis directed in the target sound direction. The second microphone unit is arranged with the directional blind spot direction directed to the target sound direction. The output signal from the first microphone unit is the main signal, and the output signal from the second microphone unit is the noise reference signal.

  In the fourth invention, the microphone device further includes a signal delay unit. The signal delay unit is provided between the output end of the noise reference signal in the signal generation unit and the subtraction unit, and delays the noise reference signal so as to satisfy the convergence condition of the adaptive filter of the adaptive filter unit.

  In the fifth invention, the predetermined value can be changed.

In the sixth invention, the signal generation unit includes a first microphone unit, a second microphone unit, a delay unit, an amplification unit, a first subtraction unit, and a second subtraction unit. . The second microphone unit has the same characteristics as the first microphone unit. The delay unit delays and outputs the signal output from the first microphone unit by a predetermined delay amount. The amplification unit amplifies the signal output from the delay unit. The first subtracting unit generates a main signal by subtracting the signal amplified by the amplifying unit from the signal output from the second microphone unit. The second subtracting unit generates a noise reference signal by subtracting the signal output from the delay unit from the signal output from the second microphone unit. The delay amount of Jo Tokoro is dead angle directivity characteristic noise reference signal output from the second subtracting unit has is set to face the target sound direction. The amplification factor in the amplification unit is set so that the main signal has higher sensitivity in the target sound direction than the noise reference signal .

  In the seventh invention, the microphone device further includes a setting unit that changes a predetermined delay amount set in the delay unit.

  In the eighth invention, the signal generation unit includes a first microphone unit, a second microphone unit, and a synthesis unit. The second microphone unit has the same characteristics as the first microphone unit. The synthesizer generates a main signal based on the signals output from the first and second microphone units so as to be sensitive to the target sound direction, and minimizes the sensitivity in the target sound direction. A noise signal component is generated.

In the ninth invention, the signal generation unit includes a first microphone unit, a second microphone unit, a signal addition unit, and a signal subtraction unit.
The second microphone unit is arranged with the directional main axis directed in a direction different from that of the first microphone unit. The signal adding unit generates a main signal by adding the signal output from the first microphone unit and the signal output from the second microphone unit. The signal subtracting unit generates a noise reference signal by subtracting the other from one of the signal output from the first microphone unit and the signal output from the second microphone unit.

  In the tenth invention, the signal generation unit includes a first microphone unit, a second microphone unit, a stereo signal generation unit, an inverse synthesis unit, and a synthesis unit. The second microphone unit has the same characteristics as the first microphone unit. The stereo signal generation unit generates a stereo signal including a right channel signal and a left channel signal based on the first and second microphone units. The inverse synthesizer generates each signal output from each microphone unit based on the stereo signal. Based on each signal generated by the inverse synthesizer, the synthesizer outputs a main signal indicating a result detected with sensitivity to the target sound direction and a sound arriving from a direction other than the target sound direction. A noise reference signal indicating the detection result with higher sensitivity than the target sound is generated.

  In the eleventh aspect of the invention, the signal generation unit includes a first microphone unit, a second microphone unit, a stereo signal generation unit, a signal addition unit, and a signal subtraction unit. The second microphone unit has the same characteristics as the first microphone unit. The stereo signal generation unit generates a stereo signal including a right channel signal and a left channel signal based on the first and second microphone units. The signal adder generates a main signal by adding the right channel signal and the left channel signal of the stereo signal. The signal subtracting unit generates a noise reference signal by subtracting the other from one of the right channel signal and the left channel signal of the stereo signal.

  In the twelfth aspect, the microphone device further includes a reflection information calculation unit and a reflection correction unit. The reflection information calculation unit calculates information related to the arrival time difference between the direct wave and the reflected wave of the target sound based on the filter coefficient of the adaptive filter unit. The reflection correction unit corrects the distortion of the frequency characteristics generated in the main signal due to the reflected wave of the target sound based on the information calculated by the reflection information calculation unit. The noise suppression unit suppresses the signal component of the noise included in the main signal using the main signal after correction by the reflection correction unit and the noise reference signal after subtraction by the subtraction unit.

In the thirteenth invention, the noise suppression filter coefficient calculation unit includes a first frequency analysis unit, a second frequency analysis unit, a power spectrum ratio calculation unit, a multiplication unit, and a coefficient calculation unit. It is out. The first frequency analysis unit calculates a power spectrum of the main signal. The second frequency analysis unit calculates a power spectrum of the noise reference signal after subtraction by the subtraction unit. The power spectrum ratio calculation unit is calculated by the power spectrum calculated by the first frequency analysis unit and the second frequency analysis unit only when the determination unit determines that the level ratio is smaller than a predetermined value. The time average of the power spectrum ratio to the power spectrum is calculated. The multiplication unit multiplies the time average of the power spectrum ratio calculated by the power spectrum ratio calculation unit and the power spectrum calculated by the second frequency analysis unit. The coefficient calculation unit calculates a filter coefficient of the noise suppression filter based on the power spectrum calculated by the first frequency analysis unit and the multiplication result by the multiplication unit.

In addition, an audio reproduction device according to a fourteenth aspect includes an audio recording unit, a signal generation unit, a determination unit, an adaptive filter unit, a subtraction unit, a noise suppression unit, and a reproduction unit. The audio recording unit records audio signals of at least two types of channels. Based on the audio signal recorded in the recording unit, the signal generator detects the main signal indicating the detection result with sensitivity to the target sound direction and the sensitivity blind angle toward the target sound direction . A noise reference signal indicating the result is generated. The determination unit determines whether or not a level ratio indicating a ratio of the signal level of the main signal to the signal level of the noise reference signal generated by the signal generation unit is greater than a predetermined value. The adaptive filter unit generates a signal indicating the signal component of the target sound included in the noise reference signal generated by the signal generation unit by filtering the main signal generated by the signal generation unit with an adaptive filter, and determines Only when it is determined by the unit that the level ratio is larger than a predetermined value, the filter coefficient is learned. Subtraction unit, the noise reference signal, generated by the adaptive filter subtracts the signal indicating the signal components of the target sound included in the noise reference signal. The noise suppression unit suppresses a signal component of noise included in the main signal using the main signal and the noise reference signal after subtraction by the subtraction unit. The reproduction unit reproduces the main signal in which the noise signal component is suppressed by the noise suppression unit. The noise suppression unit includes a noise suppression filter coefficient calculation unit and a time-varying coefficient filter unit. The noise suppression filter coefficient calculation unit calculates a filter coefficient of a noise suppression filter for suppressing signal components other than the target sound signal from the main signal based on the main signal and the noise reference signal after subtraction by the subtraction unit. . The time-varying coefficient filter unit performs filtering on the main signal, reflecting the filter coefficient calculated by the noise suppression filter coefficient calculation unit.

In the fifteenth aspect of the invention, the audio reproduction device includes a video recording unit that records a video signal related to the audio signal recorded in the audio recording unit, and a video that reproduces the video signal recorded in the video recording unit. A playback unit and a direction receiving unit that receives an input of a direction in which the sound should be emphasized from a user are further provided. At this time, the signal generation unit generates the main signal and the noise reference signal with the direction received by the direction reception unit as the target sound direction.

  According to the first aspect, the signal component of the target sound included in the noise reference signal is removed from the noise reference signal, and then noise suppression processing is performed based on the main signal and the noise reference signal. Therefore, since noise suppression processing can be performed using an ideal noise reference signal, high S / N can be realized. According to the first invention, all sounds other than the target sound can be suppressed as noise. Therefore, not only noise in one direction but also noise in all directions can be handled.

  In addition, according to the second invention, since the influence of the reflected wave on the main signal can be corrected, a microphone device having stable sensitivity vs. frequency characteristics can be realized without being influenced by the sound field around the microphone device. can do. In addition, since there is no change in sound quality due to the reflector, the effect of improving the recognition rate is great particularly in speech recognition applications.

  Further, according to the third invention, the main signal and the noise reference signal can be easily generated. Furthermore, since the two microphone units can be arranged close to each other until they come into contact with each other, the microphone device can be miniaturized.

  Further, according to the fifth aspect, it is possible to control how many times the sound collection range of the microphone device can be collected left and right with the target sound direction as the center. Therefore, it is possible to set the sound collection angle width according to the purpose, or to make the sound collection angle width variable like a zoom microphone.

  Further, according to the sixth aspect, a directivity pattern in which the sensitivity characteristics of the main signal and the noise reference signal substantially match in directions other than the target sound direction can be obtained. Therefore, consistency in the subsequent noise suppression process is improved, and the voice quality after the process is improved.

  According to the seventh aspect, the sound collection direction can be controlled by changing the delay time.

  According to the ninth aspect, the main signal and the noise reference signal can be obtained using, for example, a signal output from a one-point stereo microphone.

  According to the tenth and eleventh aspects, the main signal and the noise reference signal can be obtained using the stereo signal.

  Further, according to the twelfth aspect, it is possible to simultaneously suppress both noise that is additive noise and reflected wave that is multiplicative noise. Therefore, it is possible to realize a microphone frequency characteristic that is always free from the influence of the sound field and has a high S / N ratio.

(Embodiment 1)
First, the microphone device according to Embodiment 1 of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram showing the configuration of the microphone device according to Embodiment 1. In FIG. In FIG. 1, the microphone device includes a first microphone unit 1, a second microphone unit 2, a determination unit 10, an adaptive filter unit 20, a signal subtraction unit 30, a noise suppression filter coefficient calculation unit 40, And a time-varying coefficient filter unit 50.

  In FIG. 1, the first microphone unit 1 is a unidirectional microphone unit. The main directivity axis of the first microphone unit 1 is directed in the front direction. The second microphone unit 2 is a bidirectional microphone unit. The main directivity axis of the second microphone unit 2 is oriented in a direction perpendicular to the front direction. The microphone device detects sound coming from a desired direction. In the following, the sound to be detected is called a target sound, and the desired direction is called a target sound direction. In the first embodiment, the front direction is the target sound direction.

  The determination unit 10 uses the signal m1 output from the first microphone unit 1 and the signal m2 output from the second microphone unit 2 as input signals, and whether or not the target sound has arrived according to the level ratio between the input signals Determine. The adaptive filter unit 20 outputs a signal obtained by filtering the signal m1 with the filter coefficient. The signal subtracting unit 30 subtracts the signal output from the adaptive filter unit 20 from the signal m2.

  The noise suppression filter coefficient calculation unit 40 receives the signal m1 as a main signal, and inputs the signal m3 output from the first signal subtraction unit 30 as a noise reference signal. The noise suppression filter coefficient calculation unit 40 calculates a filter coefficient indicating filter characteristics for noise suppression using the main signal and the noise reference signal. The calculated filter coefficient is output to the time-varying coefficient filter unit 50. The time varying coefficient filter unit 50 receives the signal m1. The input signal is filtered according to the filter coefficient calculated by the noise suppression filter coefficient calculation unit 40 and output.

  The operation of the microphone device configured as described above will be described. In the following description, the direction in which the target sound arrives is the front direction unless otherwise specified.

  In FIG. 1, the first microphone unit 1 is disposed close to the second microphone unit 2. By arranging the microphone units 1 and 2 close to each other, the second microphone unit 2 can collect sounds other than the target sound (that is, noise) at substantially the same position as the first microphone unit 1. it can. The microphone device according to Embodiment 1 realizes high S / N sound collection by suppressing noise mixed in the first microphone unit 1 by the time-varying coefficient filter unit 50. At that time, the signal m2 is used as a noise reference signal. Therefore, it is ideal that the microphone units 1 and 2 collect sound in the same place. That is, it is desirable that the microphone units 1 and 2 are placed in contact with each other on the condition that the directivity of the microphone units 1 and 2 is not affected by each other. Therefore, in the first embodiment, the microphone units 1 and 2 are arranged close to each other.

  In the first embodiment, a noise suppression processing method using a time-varying coefficient filter is employed in the subsequent stage (noise suppression filter coefficient calculating unit 40 and time-varying coefficient filter unit 50) of the microphone device. When the target sound is mixed into the second microphone unit 2, due to the nature of the method, adverse effects such as distortion and level reduction occur in the processed sound. Therefore, when using this method, how to remove the mixing of the target sound into the noise reference signal becomes a problem. In view of this, the first embodiment is configured such that the blind spot of the directivity of the second microphone unit 2 faces the front direction in order to reduce the mixing of the target sound into the noise reference signal as much as possible. The reason why the bidirectional microphone unit 2 is used as the second microphone unit 2 is that the bidirectional microphone unit has a manufacturing variation related to characteristics such as the direction of blind spot and sensitivity attenuation, etc. This is because there is a feature that it is less than that of the microphone unit.

  In addition, although the 2nd microphone unit 2 is comprised as mentioned above, mixing of the target sound into a noise reference signal can be suppressed, but mixing of the target sound into a noise reference signal cannot be eliminated completely. . This is because, in an actual usage environment, the reflected wave of the target sound is detected by the second microphone unit 2 due to acoustic effects such as a casing to which the microphone unit is attached and a reflector around the microphone device. Because. In addition to the influence of the reflected sound of the target sound, the direct wave of the target sound is detected slightly even if the directivity blind spot of the second microphone unit 2 is directed in the front direction (the target sound remains unerased). Because there is. For the reasons described above, the target sound component is mixed in the signal m2. Therefore, in the first embodiment, the canceller is configured by the determination unit 10, the adaptive filter unit 20, and the first signal subtraction unit 30. By this canceller, the target sound component mixed in the noise reference signal is removed. As a result, an ideal noise reference signal, that is, a noise reference signal in which the target sound is not mixed can be obtained.

  In FIG. 1, the signal output from the first microphone unit 1 has a higher proportion of the target sound component than the noise component. The microphone device according to Embodiment 1 adaptively equalizes the signal m1 to the signal of the target sound component mixed in the signal m2 in the waveform equalization processing in the canceller. That is, the main signal is equalized to the signal of the target sound component included in the noise reference signal. Thereby, the canceller can be operated with high accuracy.

  Further, in the first embodiment, the adaptive filter unit 20 of the canceller performs the adaptive filter learning operation only when the target sound is sufficiently large. Specifically, the determination unit 10 detects whether the target sound is larger than the noise. The adaptive filter unit 20 performs an adaptive filter learning operation according to the detection result of the determination unit 10. Thereby, the filter coefficient of the adaptive filter unit 20 can be converged stably. In addition, the determination part 10 needs to detect both the arrival direction and level of a sound. The detailed configuration of the determination unit 10 will be described later (see FIG. 2).

  Next, the details of the operation together with the detailed configuration of each component of the microphone device will be described. FIG. 2 is a diagram illustrating a configuration of the determination unit illustrated in FIG. 1. In FIG. 2, the determination unit 10 includes a first signal level calculation unit 11, a second signal level calculation unit 12, a signal division unit 13, and a target sound arrival determination unit 14.

  In FIG. 2, the first signal level calculation unit 11 receives the signal m1 as an input, calculates the short-time average of the signal level of the signal m1, and outputs the first signal level x1a. The second signal level calculation unit 12 receives the signal m2 as an input, calculates a short-time average of the signal level of the signal m2, and outputs a second signal level x2a. The signal divider 13 obtains a signal ratio (level ratio) between the first signal level x1a and the second signal level x2a. Specifically, the signal dividing unit 13 outputs the signal ratio Va by dividing Va = x1a / x2a. Based on the output from the signal divider 13, the target sound arrival determination unit 14 determines whether the target sound is sufficiently large, that is, whether the target sound is larger than the noise. Specifically, the target sound arrival determination unit 14 compares the magnitude relationship between the signal ratio Va and the predetermined threshold value th1, and outputs a determination result Vx indicating the magnitude relationship. More specifically, Vx is a value indicating that the signal ratio Va is larger than the predetermined threshold th1 (here, “1”), and the signal ratio Va is equal to or less than the predetermined threshold th1. It takes a binary value, which is a value indicating that it is present (here, “0”).

  In FIG. 2, first, consider a case where the sound coming from the θ0 direction (front direction) is dominant. Here, “the sound from the θ0 direction is dominant” means that the sound coming from the θ0 direction is much louder than the sound coming from the other direction, and the sound coming from the other direction can be ignored. Mean small. In this case, the θ0 direction coinciding with the front direction is the maximum sensitivity direction of the first microphone unit 1 and the minimum sensitivity direction of the second microphone unit. Accordingly, the value of the first signal level x1a is large (relative to the case described later), and the value of the second signal level x2a is small (relative to the case described later). Therefore, in this case, the signal ratio Va (= x1a / x2a) is a large value (relative to the case described later).

  Next, consider the case where the sound coming from the θ1 direction is dominant. Here, the directivity characteristic of the first microphone 1 is unidirectional with the main directivity axis oriented in the θ0 direction. The directivity characteristic of the second microphone 2 is bi-directional with the directivity main axis directed in the θ2 direction. Therefore, when the sound arriving from the θ1 direction is dominant, the value of the first signal level x1a is decreased compared to the case where the sound arriving from the θ0 direction is dominant, and the second signal level x2a The value increases. As a result, the signal ratio Va is smaller than when the sound arriving from the θ0 direction is dominant. Further, when the direction of the dominant sound moves from the θ1 direction to the θ2 direction, the value of the first signal level x1a further decreases and the value of the second signal level x2a further increases. As a result, the signal ratio Va is smaller than when the sound arriving from the θ0 direction is dominant.

  Next, consider the case where the sound coming from the θ3 direction is dominant. Here, with respect to both microphone units 1 and 2, the θ3 direction is a direction that becomes a directivity blind spot. Both the first signal level x1a and the second signal level x2a become small, and as a result, the signal ratio Va does not become a large value.

  FIG. 3 is a diagram illustrating an example of a sound detection state in a case where the dominant sound direction is the θ1 to θ3 direction. The waveforms of the first signal level x1a, the second signal level x2a, and the signal ratio Va are the signal waveforms shown in FIG. Here, by setting the threshold th1 to the level shown in FIG. 3, it is possible to detect that the sound in the θ0 direction is dominant as the determination result Vx. That is, when the threshold value th1 is set to the level shown in FIG. 3, the value of the determination result Vx is “1” only when the sound in the θ0 direction is dominant. In Embodiment 1, since the sound coming from the front direction (θ0 direction) is the target sound, it can be detected from the value of Vx that the target sound is dominant. If the target sound is not only the sound arriving from the θ0 direction but also the sound arriving from the θ1 direction, the threshold value may be set to th2 shown in FIG. When the threshold is set to th2, the value of Vx is “1” not only in the θ0 direction sound but also in the θ1 direction sound.

  Next, an operation of removing the target sound mixed in the noise reference signal (signal m2) in the adaptive filter unit 20 and the signal subtracting unit 30 will be described. The adaptive filter unit 20 equalizes the signal m1 to the signal of the target sound component included in the signal m2 by the adaptive filter. That is, the adaptive filter unit 20 generates a signal of the target sound component included in the signal m2 from the signal m1. As an adaptive filter method, for example, an LMS method (learning identification method) or the like can be used. The signal subtracting unit 30 subtracts the signal generated by the adaptive filter unit 20 from the signal m2. As a result, the signal m3 becomes a noise reference signal from which the target sound component is removed.

  Here, the adaptive filter unit 20 determines whether or not to learn the filter coefficient according to the determination result Vx by the determination unit 10. Specifically, when the determination unit 10 determines that the target sound is dominant, that is, when the determination result Vx indicates “1”, the adaptive filter unit 20 performs learning. On the other hand, when the determination unit 10 determines that the target sound is not dominant, that is, when the determination result Vx indicates “0”, the adaptive filter unit 20 does not perform learning.

  First, consider the case where the target sound is dominant. In this case, the adaptive filter unit 20 performs learning. Here, when the target sound is dominant, the noise can be ignored, so the second microphone unit 2 does not detect the noise, but the target sound component (the reflected wave of the target sound or the direct sound of the target sound). It can be considered that only components (such as unerased waves) are detected. That is, the signal m2 can be regarded as including only the target sound component without including the noise component. In this case, the adaptive filter unit 20 may output the signal m2 as a result of filtering the signal m1. That is, the filter coefficient may be learned so that the signal m3 becomes zero. As a result of this learning, the adaptive filter unit 20 can obtain a filter coefficient for generating a signal of the target sound component included in the signal m2 with high accuracy based on the signal m1.

  On the other hand, consider the case where the target sound is not dominant. In this case, the signal m2 includes a noise component having a magnitude that cannot be ignored in addition to the target sound component. Therefore, in this case, the adaptive filter unit 20 cannot obtain an appropriate filter coefficient even if the filter coefficient is learned so that the signal m3 becomes zero. That is, the filter coefficient for generating the signal of the target sound component included in the signal m2 cannot be obtained based on the signal m1. Furthermore, if learning is performed in such a case, the filter coefficients may diverge. For the above reasons, the adaptive filter unit 20 should not learn filter coefficients. Therefore, the adaptive filter unit 20 does not perform learning when the target sound is not dominant.

  As described above, adaptive filter learning is performed only when the target sound is louder than surrounding noise by using the determination result of the determination unit 10. Thereby, the adaptive filter unit 20 can converge the filter coefficients stably.

  As described above, the microphone device according to Embodiment 1 first separates the target sound and the noise to some extent as preprocessing using the directivity characteristics of the microphone units 1 and 2. In addition, by using the canceller, the target sound component mixed in the noise reference signal, which cannot be suppressed by the configuration using the microphone units 1 and 2, is removed. As described above, the microphone device according to Embodiment 1 can obtain an ideal noise reference signal.

  If a noise reference signal is to be obtained only by the configuration of the canceller without performing preprocessing using the directivity characteristics of the microphone units 1 and 2, there are the following disadvantages. Since it is difficult to detect the target sound in an environment where noise is generated, there is a disadvantage that the accuracy of learning control is deteriorated. Further, since the target sound is not emphasized using the directivity of the microphone unit, there is a disadvantage that the correlation of the learning signal (target sound) is lowered and it is difficult to converge the filter coefficients.

  Next, the operation of suppressing the noise component from the main signal (signal m1) by the noise suppression filter coefficient calculation unit 40 and the time varying coefficient filter unit 50 will be described. Note that the noise suppression effect similar to that of the noise suppression filter coefficient calculation unit 40 and the time-varying coefficient filter unit 50 can be obtained also by the configuration that performs the two-input type spectral subtraction method. However, when the spectral subtraction method is performed, a frame process for finally returning the spectrum to the waveform signal is required, which causes a processing delay. Note that, as a method for reducing the signal delay in the frame processing, it is conceivable to shorten the frame length, increase the frame overlap, or the like. However, the former is not realistic in that the frequency resolution is lowered, and the latter is not realistic in that the processing amount is increased. Therefore, in the first embodiment, a configuration using a time-varying coefficient filter, which is a method with less processing delay, is employed.

  FIG. 4 is a diagram illustrating a configuration example of the noise suppression filter coefficient calculation unit 40. In FIG. 4, a noise suppression filter coefficient calculation unit 40 includes a first frequency analysis unit 41, a second frequency analysis unit 42, a spectrum ratio calculation unit 43, a signal averaging unit 44, a signal multiplication unit 45, A filter transfer characteristic estimation unit 46 and an impulse response design unit 47 are provided.

  In FIG. 4, the first frequency analysis unit 41 calculates the power spectrum X (ω) of the signal m1, which is the main signal. The second frequency analysis unit 42 calculates a power spectrum N1 (ω) of the signal m3 that is a noise reference signal. Here, each of the frequency analysis units 41 and 42 can be realized by using a known method capable of deriving the power of the frequency component, such as FFT, filter bank, wavelet transform, and DCT.

  The spectrum ratio calculation unit 43 receives the power spectrum X (ω) calculated by the first frequency analysis unit 41 and the power spectrum N1 (ω) calculated by the second frequency analysis unit 42 as input. H (ω) = X (ω) / N1 (ω) is derived. The signal average unit 44 receives the spectrum ratio H (ω) derived by the spectrum ratio calculation unit 43 and the determination result Vx by the determination unit 10 as inputs. Then, a time average Ha (ω) for each frequency component when the ambient noise is more dominant than the target sound (that is, when the value of Vx is “0”) is calculated. The signal multiplication unit 45 multiplies the power spectrum N1 (ω) calculated by the second frequency analysis unit 42 and the time average Ha (ω) calculated by the signal averaging unit 44 for each frequency component. Then, the multiplication result is output as Nx (ω). Note that the shape and level of the noise component spectrum other than the target sound component included in the main signal spectrum X (ω) due to the different directivity patterns, the characteristics of the microphone unit, and the like are the spectrum N1 of the noise reference signal. It is not necessarily equal to the shape and level of (ω). The spectrum ratio calculation unit 43, the signal averaging unit 44, and the signal multiplication unit 45 described above include the spectrum of the noise component other than the target sound component included in the spectrum X (ω) of the main signal and the spectrum N1 of the noise reference signal. This is a configuration for matching (ω). Accordingly, Nx (ω) obtained as a multiplication result of the signal multiplier 45 becomes a noise component included in the spectrum X (ω) of the main signal. Therefore, this Nx (ω) is referred to as an estimated noise spectrum Nx (ω).

  The filter transfer characteristic estimator 46 receives the power spectrum X (ω) calculated by the first frequency analyzer 41 and the estimated noise spectrum Nx (ω) calculated by the signal multiplier 45 as inputs, and receives a noise suppression filter. Transfer characteristic Hw (ω) is calculated. The transfer characteristic Hw (ω) of the noise suppression filter can be obtained by, for example, Hw (ω) = (X (ω) −Nx (ω)) / X (ω) based on the Wiener filter method.

  The impulse response design unit 47 uses the transfer characteristic Hw (ω) calculated by the filter transfer characteristic estimation unit 46 as a target characteristic, and outputs a filter coefficient hw (n) so as to gradually approach the target characteristic every sample. .

  The time-varying coefficient filter unit 50 filters the signal m1 according to the filter coefficient hw (n) output from the impulse response design unit 47, and generates an output signal y of the microphone device. Hereinafter, a specific configuration example of the time-varying coefficient filter unit 50 will be described with reference to FIGS. 5 and 6.

  FIG. 5 is a diagram illustrating a configuration example of the time varying coefficient filter unit 50. In FIG. 5, the time-varying coefficient filter unit 50 includes n signal delay units, n + 1 signal amplification units, and n signal addition units. In FIG. 5, the first signal delay unit 501, the second signal delay unit 502, the nth signal delay unit 503, the first signal amplification unit 504, the second signal amplification unit 505, the nth Only the signal amplification unit 506, the first signal addition unit 508, and the nth signal addition unit 509 are shown.

  In FIG. 5, each signal delay unit is connected to each other and delays an input signal by one sample. Each signal amplifying unit amplifies and outputs the input signal. The first signal amplification unit 504 amplifies the signal m <b> 1 input to the time varying coefficient filter unit 50. The second signal amplification unit 505 amplifies the signal output from the first signal delay unit 501. Thereafter, the signal amplifying unit subsequent to the second signal amplifying unit 505 performs the same operation as that of the second signal amplifying unit 505. In other words, the (i + 1) -th signal amplification unit amplifies the signal output from the i-th signal delay unit (i is an integer from 1 to n). The first signal adder 508 adds the signal output from the first signal amplifier 504 and the signal output from the second signal amplifier 505. The second signal adder (not shown) adds the signal output from the first signal adder 508 and the signal output from the third signal amplifier (not shown). Thereafter, the signal adding unit at the subsequent stage of the second signal adding unit (not shown) performs the same operation as the second signal adding unit. That is, the jth signal adding unit adds the signal output from the j−1th signal adding unit and the signal output from the i + 1th signal amplifying unit (j is an integer from 2 to n). ). The signal output from the nth signal adding unit 509 is the output signal y. The configuration shown in FIG. 5 is a configuration of a general FIR filter, and the coefficients of the first to (n + 1) th signal amplification units change according to the filter coefficient hw (n) from the impulse response design unit 47. .

  FIG. 6 is a diagram illustrating another configuration example of the time-varying coefficient filter unit 50. In FIG. 6, the time-varying coefficient filter unit 50 includes n band-pass filters, n signal amplification units, and a signal addition unit 517. In FIG. 6, the first bandpass filter 511, the second bandpass filter 512, the nth bandpass filter 513, the first signal amplification unit 514, the second signal amplification unit 515, the nth Only the signal amplifier 516 and the signal adder 517 are shown.

  In FIG. 6, each bandpass filter is provided in parallel in the subsequent stage of the input signal, and divides the band of the signal m <b> 1 input to the time-varying coefficient filter unit 50 into n and outputs it. Each signal amplifying unit amplifies the signal output from each bandpass filter. The signal adder 517 adds the signals output from the signal amplifiers and outputs the addition result as an output signal y. The amplification factor of each signal amplifying unit can be determined based on the transfer function Hw (ω) output from the filter transfer characteristic estimating unit 46. Also with the above configuration, the same effect as in FIG. 5 can be obtained.

  FIG. 7 is a diagram showing a specific example of each signal shown in FIG. Specifically, the signal m1 output from the first microphone unit 1, the signal m2 output from the second microphone unit 2, the signal m3 output from the first signal subtraction unit 30, and the time-varying coefficient filter The specific example of the output signal y output from the part 50 is shown. As shown in FIG. 7, the signal m <b> 3 is a signal that includes only components other than the target sound, i.e., only noise components, by removing the influence of reflected sound and the like from the signal m <b> 2. Further, by performing filtering using the main signal m1 and the noise reference signal m3 in the time-varying coefficient filter unit 50, only the target sound can be extracted as the output signal y. As is apparent from comparing the output m1 of the conventional directional microphone unit with the output signal y of the microphone device of the first embodiment, the microphone device of the first embodiment generates a target sound. The ambient noise can be greatly suppressed regardless of whether the target sound is generated or the target sound is not generated.

  Note that, depending on the positional relationship between the first microphone unit 1 and the second microphone unit 2 and the circuit of each component provided in the subsequent stage of each microphone unit 1 and 2, the causality for convergence of the adaptive filter may be satisfied. For the purpose, a signal delay unit may be provided between the signal subtracting unit 30 and the second microphone unit 2. The amount of delay in the signal delay unit is determined based on a criterion that the distance between the microphone units 1 and 2 is equal to or greater than the amount obtained by dividing the distance by the sound speed.

  In the first embodiment, a unidirectional microphone unit is used as the first microphone unit 1, but an omnidirectional microphone or a superdirectional microphone may be used.

  In the above description, the determination unit 10 outputs a numerical value expressed in binary as the determination result Vx. Here, the determination unit 10 may output the signal ratio Va expressed in multiple values. Further, in this case, the adaptive filter unit 20 changes the learning speed according to the determination result (signal ratio Va). Specifically, when the signal ratio Va is larger than the threshold value, the adaptive filter unit 20 increases the learning speed as the signal ratio Va increases. More specifically, the adaptive filter unit 20 brings the value of the step gain parameter closer to 0.5 as the signal ratio Va increases. On the other hand, when the signal ratio Va is equal to or less than the threshold value, the adaptive filter unit 20 does not perform learning. More specifically, the value of the step gain parameter is set to zero.

  As described above, the microphone device according to Embodiment 1 can obtain an ideal noise reference signal even in a noise environment and in a reflected sound field. Therefore, the noise suppression unit using the main signal and the noise reference signal can greatly improve the sound collection S / N compared to the conventional directional microphone. Furthermore, the microphone device according to Embodiment 1 can reduce the processing delay as compared with the case of using the spectral subtraction method by adopting a method using a time-varying coefficient filter as a noise suppression method. Therefore, the microphone device according to Embodiment 1 can also be applied to uses that require processing with little delay, such as a loudspeaker use and a call use.

(Embodiment 2)
Next, a microphone device according to Embodiment 2 will be described with reference to FIGS. The microphone device according to Embodiment 1 is intended to suppress noise that is mixed when detecting the target sound. The microphone device according to the second embodiment is intended to correct the frequency characteristic distortion of the target sound caused by detecting the reflected wave of the target sound.

  In FIG. 8, the microphone device includes a first microphone unit 1, a second microphone unit 2, a determination unit 10, an adaptive filter unit 20, a signal subtraction unit 30, a reflection information calculation unit 60, and reflection correction. Part 70. In FIG. 8, the same components as those in the first embodiment are denoted by the same reference numerals as those in FIG. 1, and detailed description thereof is omitted.

  In FIG. 8, the filter information of the adaptive filter unit 20 is input to the reflection information calculation unit 60. The reflection information calculation unit 60 uses the input filter coefficients to estimate the presence / absence, distance, and influence of the reflecting object. The reflection correction unit 70 receives the signal m1 and corrects the frequency characteristic distortion generated in the signal m1 due to the influence of the reflection of the target sound based on the estimation result of the reflection information calculation unit 60.

  Hereinafter, the operation of the microphone device according to Embodiment 2 will be described.

  In the microphone device shown in FIG. 8, the signal m1 is the main signal. Here, when the directivity of the first microphone unit 1 is unidirectional, the directivity of the first microphone unit 1 is not so sharp that the reflected wave of the target sound can be removed. Therefore, when a reflective object is present in the vicinity of the microphone device, the reflected wave is simultaneously collected in addition to the direct wave of the target sound. Will be disturbed by. The microphone device according to the second embodiment corrects the frequency characteristic distorted by the influence of the reflection of the target sound by using the fact that the information of the reflected wave appears in the filter coefficient of the adaptive filter unit 20. This enables automatic correction of the frequency characteristics of the detected sound.

  As described above, the adaptive filter unit 20 generates a target sound component mixed in the signal m2, that is, a signal of a target sound unerased component due to imperfect directivity and a reflected wave component of the target sound. That is, the transfer characteristic (impulse response) from the signal m1 containing a large amount of the direct wave component of the target sound to the signal m2 containing the large amount of the reflected wave component of the target sound is expressed in the filter coefficient of the adaptive filter unit 20. It will be. Therefore, by detecting the peak of the coefficient from the filter coefficient, it represents the time difference dt (sec) between the time when the direct wave of the target sound arrives at the position of the microphone unit and the time when the reflected wave arrives, and the reflected wave. The peak level Lr and the intensity of reflection are known. Further, from the time difference dt, the distance difference dt × c (where c is the speed of sound) between the path from which the reflected wave of the target sound arrives and the path from which the direct wave arrives is known.

  Here, for the sound of the frequency whose wavelength is equal to the distance difference (wavelength λ satisfies the relationship of λ = dt × c), the direct wave and the reflected wave are added in the same phase, and therefore detected by the microphone unit. Increased sound pressure level. Conversely, for a sound having a frequency whose wavelength is equal to 1/2 of the distance difference (wavelength λ satisfies the relationship of λ / 2 = dt × c), the direct wave and the reflected wave are in opposite phases. The sound pressure level detected by the microphone unit decreases, and a dip occurs in the frequency characteristics of the main signal. If complete reflection occurs on the reflecting surface, the signal output from the first microphone unit 1 is emphasized by a harmonic part having a fundamental frequency of fa (= c / λ = 1 / dt). A comb-like filter-like frequency characteristic appears.

  FIG. 9 is a diagram for explaining the difference in the internal state of the microphone device when there is a reflector and when there is no reflector. In FIG. 9, the positional relationship between the microphone unit, the target sound source (speaker), and the reflector, and the value of the adaptive filter coefficient hadf (n) in the adaptive filter unit 20 when there is a reflector and when there is no reflector. And frequency characteristics of the signal m1 are shown.

  In FIG. 9, in the state where there are no reflectors in the vicinity of the speaker and the microphone unit as shown in (a1), as shown in (a2), the influence of the reflected wave on the filter coefficient of the adaptive filter unit 20 is as follows. It does not appear. Furthermore, as shown in (a3), the shape of the frequency characteristic of the main signal is relatively flat. On the other hand, in the state where there is a reflector near the speaker and microphone as shown in (b1), the filter coefficient of the adaptive filter unit 20 has a large value in the portion of the time difference dt as shown in (b2). Become. Further, as shown in (b3), the frequency characteristics of the main signal are also distorted according to the positional relationship between the microphone, the target sound source, and the reflector.

As described above, the time difference dt and the influence level Lr can be calculated from the coefficient peak of the adaptive filter. Furthermore, the correction amount of the frequency characteristic distorted by the influence of the reflected wave can be estimated using these. Actually, it cannot be considered that complete reflection occurs on the reflecting surface particularly in a high sound range. In the case where it is not possible to assume that complete reflection occurs on the reflection surface, it is conceivable to perform deconvolution filter design assuming the reflection characteristics of the reflection surface. Further, focusing attention only on the low frequency characteristics, a frequency where one wavelength is equal to the distance difference (fa = 1 / dt), a frequency where the half wavelength is equal to the distance difference (fb = 1/2 dt), etc. For the frequency, for example, the correction gain is calculated by the following equation.
Center frequency fa: correction gain = −β1 · 20 log (1 + α1 · Lr) (dB)
Center frequency fb: correction gain = + β2 · 20 log (1−α2 · Lr) (dB)
In this case, the correction characteristic Hr (ω) of the reflection correction unit 70 can be realized by an equalizer that can adjust the center frequency, the bandwidth, and the gain based on the information from the reflection information calculation unit 60.

  Note that, when the use environment can be limited, for example, when a microphone device is used for voice recognition for car navigation, the detection accuracy of the filter coefficient of the adaptive filter unit 20 can be increased. Specifically, only the initial reflection component is targeted, and the search range of the maximum value of the filter coefficient is limited based on the reflected wave delay amount calculated from the reflection surface position.

  In addition, depending on the directivity type of the microphone unit, the maximum value of the filter coefficient may depend on whether the peak due to the reflected wave is positive or negative depending on the direction of the reflected wave depending on the polarity of the directivity lobe. In such a configuration, it is necessary to search for the maximum value with respect to the absolute value of the coefficient.

  As described above, according to the second embodiment, it is possible to correct the frequency characteristics that are distorted by the influence of the reflected wave of the target sound. Therefore, it is possible to realize a microphone device that can obtain a stable and flat sound pressure sensitivity versus frequency characteristic in any use environment (sound field). Therefore, according to the second embodiment, it is possible to improve the sound quality in a telephone call or voice expansion. In particular, in frequency recognition applications, frequency characteristic distortion caused by reflected waves was one of the causes of misrecognition. However, the configuration of the second embodiment enables stable and high voice regardless of the presence or absence of nearby reflectors. The recognition rate can be realized.

(Embodiment 3)
Next, a microphone device according to Embodiment 3 will be described with reference to FIGS. The microphone device according to the third embodiment has a configuration obtained by combining the configuration of the first embodiment and the configuration of the second embodiment.

  FIG. 10 is a block diagram showing a configuration of the microphone device according to Embodiment 3. In FIG. 10, the microphone device includes a first microphone unit 1, a second microphone unit 2, a determination unit 10, an adaptive filter unit 20, a signal subtraction unit 30, a noise suppression filter coefficient calculation unit 40, A time-varying coefficient filter unit 50, a reflection information calculation unit 60, and a reflection correction unit 70 are provided. In FIG. 10, the same components as those in the first or second embodiment are denoted by the same reference numerals as those in FIG. 1 or 8, and detailed description thereof is omitted.

  The difference between the configuration shown in FIG. 10 and the configuration shown in FIG. 8 is that the noise suppression filter coefficient calculation unit 40 and the time-varying coefficient filter unit 50 shown in FIG. 1 are provided after the configuration shown in FIG. . Accordingly, the microphone device shown in FIG. 10 can correct the distortion of the frequency characteristic due to the reflected wave and perform noise suppression.

  FIG. 11 is a block diagram showing another configuration of the microphone device according to Embodiment 3. In FIG. In FIG. 11, the microphone device includes a first microphone unit 1, a second microphone unit 2, a determination unit 10, an adaptive filter unit 20, a signal subtraction unit 30, a time-varying coefficient filter unit 50, a reflection An information calculation unit 60, a reflection correction unit 70, and a noise suppression and reflection inverse characteristic filter coefficient estimation unit 80 are provided. The configuration shown in FIG. 11 is a configuration in which the processing amount is reduced by superimposing the characteristics of the reflection correction unit 70 on the characteristics of the time-varying coefficient filter unit 50.

  The operation of the configuration shown in FIG. 11 is different from the operation of the configuration shown in FIG. 10 in the operation of the noise suppression and reflection inverse characteristic filter coefficient estimation unit 80. The noise suppression and reflection inverse characteristic filter coefficient estimation unit 80 receives the signal m1 (main signal), the signal m3 (noise reference signal), and the signal output from the reflection information calculation unit 60 as inputs. Based on these signals, the noise suppression filter characteristic Hw (ω) = (X (ω) −Nx (ω)) / X (ω) and the reflection inverse characteristic Hr (ω) are calculated. Further, the filter coefficient having the target characteristic {Hw (ω) · Hr (ω)} is output to the time-varying coefficient filter unit 50. As a result, it is possible to simultaneously perform the correction process of the distortion of the frequency characteristic due to the reflected wave and the noise suppression process.

  As described above, according to the third embodiment, as in the first embodiment, an ideal noise reference signal from which the target sound is removed can be obtained. Similarly to the second embodiment, the two-input type noise suppression process using the main signal and the noise reference signal and the correction process of the frequency characteristic distortion due to the influence of the reflected wave can be performed simultaneously. As a result, even if the surrounding environment is a noisy environment or a reflected sound field, it is possible to obtain a flat frequency characteristic with a high S / N, and the effect of improving the voice quality of calls and loudspeakers, The effect of improving the recognition rate of voice recognition can be obtained.

(Embodiment 4)
Next, a microphone device according to Embodiment 4 will be described with reference to FIGS. In the fourth embodiment, the direction regarded as the target sound is changed from the sounds in all directions arriving at the microphone device.

  FIG. 12 is a block diagram showing a configuration of the microphone device according to Embodiment 4. In FIG. In FIG. 12, the microphone device further includes a detection threshold setting unit 90 in addition to the configuration shown in FIG. In FIG. 12, the same components as those in the third embodiment are denoted by the same reference numerals as those in FIG. 11, and detailed description thereof is omitted.

  The detection threshold setting unit 90 sets a threshold value used in the determination unit 10. That is, the configuration of the fourth embodiment is different from the configuration of the third embodiment in that the threshold set in the determination unit 10 can be controlled.

  In FIG. 12, the threshold set in the determination unit 10 can be changed. By changing this threshold value, it is possible to change to what angle the direction in which the sound regarded as the target sound arrives is included on the left and right sides from the front direction. That is, by changing this threshold value, the range of angles that can be collected as the target sound can be controlled.

  For example, consider the case where the detection threshold value setting unit 90 sets the threshold value as th1 (see FIG. 3). In this case, the sound coming from the θ1 direction (see FIGS. 2 and 3) is not regarded as the target sound. That is, the sound coming from the θ1 direction is regarded as noise, and the signal m3 that is the noise reference signal includes a sound component coming from the θ1 direction. As a result, in the final output, the sound coming from the θ1 direction is suppressed.

  On the other hand, when the threshold is set to th2 (see FIG. 3), the sound coming from the θ1 direction is regarded as the target sound. In this case, the signal m3, which is a noise reference signal, does not include a sound component coming from the θ1 direction. As a result, in the final output, the sound arriving from the θ1 direction is output as the target sound.

  As described above, by controlling the threshold value of the determination unit 10, it is possible to control the angle range in which the microphone device can collect sound. However, the angle range is limited to a certain angle range with respect to the directional blind spot direction of the second microphone unit 2, that is, the front direction.

  FIG. 13 is a diagram illustrating a directivity pattern of the microphone device. FIG. 13A shows a directivity pattern of the signal m1. FIG. 13B is a diagram showing a directivity pattern of the output signal y of the microphone device when the threshold is set to th2. FIG. 13C is a diagram showing a directivity pattern of the output signal y of the microphone device when the threshold is set to th1. In FIG. 13B, the angle range in which the microphone device can collect sound is wider than that in FIG. For example, when the threshold is set to th2, the sound coming from the angle θ1 is determined as the target sound. In addition, the sensitivity is greatly attenuated at portions outside the range. On the other hand, in FIG. 13C, the angle range in which the microphone device can collect sound is narrow, and a very sharp directional characteristic is realized. In this case, the sound coming from the angle θ1 is not determined as the target sound.

  As described above, according to the fourth embodiment, the sharpness of the directivity of the microphone device can be changed by changing the threshold value of the determination unit 10. In general, the directivity of a microphone is more difficult to form a sharp main beam than to form a sharp blind spot. However, according to the fourth embodiment, a microphone device having a sharp directivity that has not existed in the past is realized. can do.

  Here, in actual use, the sharpness of directivity is contrary to the ease of use of the microphone device. When using a microphone device with a sharp directivity, the user must use it with a strong awareness of the front direction. Therefore, in order to achieve both ease of use and noise suppression performance, the microphone device has a certain sensitivity characteristic from the front to a certain angle range, and a directivity characteristic that increases the sensitivity attenuation in other directions. It is desirable. In addition, it is desirable that the angle range in which sound can be collected can be freely set according to the use of the microphone device and the sound collecting situation. According to the fourth embodiment, the directivity of the microphone device changes as shown in FIG. As can be seen from FIG. 13, the microphone device according to Embodiment 4 can achieve both ease of use as a microphone device and high noise removal capability.

(Embodiment 5)
Next, a microphone device according to Embodiment 5 will be described with reference to FIG. In the microphone devices according to the first to fourth embodiments, the unidirectional microphone unit and the bidirectional microphone unit are arranged close to each other, and the signals output from the microphone units are the main signal and the noise reference signal. It was the composition to do. The merit of this configuration is that it can be miniaturized and can be realized at low cost because processing such as directivity synthesis is unnecessary.

  On the other hand, video movies and other devices with sound collection functions often have multiple microphone units with omnidirectionality or directivity with the same characteristics due to mounting problems and performance problems. In some cases, directivity is formed by synthesizing signals output from the. In the process of performing directivity synthesis on signals from a plurality of microphone units, a certain distance (usually 1 cm to 5 cm) is required as a distance between the microphone units due to problems such as circuit noise. Therefore, the method of performing the process is disadvantageous in terms of downsizing compared to the above-described first to fourth embodiments. However, the method of performing the processing has advantages in terms of mounting, such as a high degree of freedom in design of directivity and the ability to use variable characteristics using digital processing.

  Therefore, in the fifth embodiment, using a plurality of microphone units having the same directivity characteristics (two in the fifth embodiment) and the directivity synthesis unit 100, a main signal corresponding to the signal m1, A configuration for obtaining a noise reference signal corresponding to the signal m2 is adopted.

  FIG. 14 is a diagram showing a part of the configuration of the microphone device according to Embodiment 5. In FIG. In FIG. 14, the microphone device includes a third microphone unit 3, a fourth microphone unit 4, and a directivity synthesis unit 100. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  In FIG. 14, each of the microphone units 3 and 4 is arranged on an axis (a chain line shown in FIG. 14) facing the front direction. The distance between each microphone unit 3 and 4 is d. Each microphone unit 3 and 4 is arranged such that its directivity main axis faces the front direction.

  The directivity synthesis unit 100 includes a first signal delay unit 101, a first signal subtraction unit 103, a second signal delay unit 102, and a second signal subtraction unit 104. The first signal delay unit 101 delays the signal output from the fourth microphone unit 4. The second signal delay unit 102 delays the signal output from the third microphone unit 3. The first signal subtraction unit 103 subtracts the signal output from the first signal delay unit 101 from the signal output from the third microphone unit 3. As a result, a signal m1 is obtained. The second signal subtracting unit 104 subtracts the signal output from the second signal delay unit 102 from the signal output from the fourth microphone unit 4. Thereby, the signal m2 is obtained.

  Further, by setting the delay amount τ1 of the first signal delay unit 101 to 0 ≦ τ1 ≦ d / c (where c is the speed of sound), the superdirectivity of the secondary sound pressure gradient type in which the directivity main axis is the front direction. The characteristic can be obtained as the signal m1. Further, by setting the signal delay amount τ2 of the second signal delay unit 102 to τ2 = d / c, a signal in which a directional blind spot is formed in the front direction (a microphone in which a directional blind spot is formed in the front direction). (Resulting signal from the unit) m2.

  With the above configuration, by realizing superdirectivity in advance in the characteristics of the signal m1, it is possible to realize sharp directivity and noise suppression performance that greatly exceed conventional superdirective microphones in combination with subsequent noise suppression processing. it can.

(Embodiment 6)
Next, a microphone device according to Embodiment 6 will be described with reference to FIG. As in the fifth embodiment, the sixth embodiment employs a configuration in which a main signal and a noise reference signal are obtained using a plurality of microphone units having the same directivity characteristics.

  FIG. 15 is a diagram showing a part of the configuration of the microphone device according to Embodiment 6. In FIG. The microphone device includes a third microphone unit 3, a fourth microphone unit 4, and a directivity synthesis unit 100. Each of the microphone units 3 and 4 is arranged on an axis (a chain line shown in FIG. 15) perpendicular to a straight line (a dotted line shown in FIG. 15) facing the front direction. Each microphone unit 3 and 4 is arranged such that its directivity main axis faces the front direction. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  In FIG. 15, the directivity synthesis unit 100 includes a first signal addition unit 105 and a second signal subtraction unit 104. The first signal adding unit 105 adds signals output from the microphone units 3 and 4. As a result, the signal m1, which is the main signal, can be obtained. The second signal subtracting unit 104 subtracts the signal output from the third microphone unit 3 from the signal output from the fourth microphone unit 4. As a result, a signal m2 that is a noise reference signal can be obtained.

  In FIG. 15, when the distance between the microphone units 3 and 4 is narrow to some extent, the directivity characteristic of the signal m1 is not much different from that of the single microphone unit (Embodiments 1 to 4) except for the high frequency characteristics. Therefore, the configuration shown in FIG. 15 cannot obtain sharp directivity as compared with the configuration shown in FIG. 14, but on the other hand, an effect of reducing vibration noise and circuit noise can be obtained. Also, since the sound coming from the front direction is detected in the same phase by the microphone units 3 and 4, a signal m2 in which a directional blind spot is formed in the front direction can be obtained.

(Embodiment 7)
Next, a microphone device according to Embodiment 7 will be described with reference to FIG. As in the fifth embodiment, the seventh embodiment employs a configuration in which a main signal and a noise reference signal are obtained using a plurality of microphone units having the same directivity characteristics.

  FIG. 16A shows a part of the configuration of the microphone device according to the seventh embodiment. The microphone device includes a third microphone unit 3, a fourth microphone unit 4, and a directivity synthesis unit 100. The arrangement of the microphone units 3 and 4 is the same as the arrangement shown in FIG. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  In FIG. 16A, the directivity synthesis unit 100 includes a signal delay unit 111, a second signal subtraction unit 104, a signal amplification unit 150, and a first signal subtraction unit 103. The signal delay unit 111 receives the signal output from the third microphone unit 3 and delays the signal. The second signal subtracting unit 104 subtracts the signal output from the signal delay unit 111 from the signal output from the fourth microphone unit 4. As a result, a signal m2 that is a noise reference signal can be obtained. The signal amplification unit 150 multiplies the signal output from the signal delay unit 111 by a constant. The first signal subtracting unit 103 subtracts the signal output from the signal amplifying unit 150 from the signal output from the fourth microphone unit 4. Thereby, the signal m1 which is the main signal can be obtained.

  In FIG. 16A, the difference between the process of obtaining the signal m1 and the process of obtaining the signal m2 is that the signal amplification unit 150 exists in the process of obtaining the signal m1. Directional blind spot directions in the signal m1 and the signal m2 are determined by the delay amount τ1 of the signal delay unit 111. For example, when τ1 = 0, the directivity blind spot is the front direction, and when τ1 = d / c, the directivity blind spot is the direction perpendicular to the front direction. Here, the delay amount τ1 is set so that a blind spot is formed in the direction of the target sound. As a result, the signal m1 and the signal m2 contain more sound components coming from directions other than the target sound direction than the target sound components.

  Here, the directivity pattern formed by the directivity synthesis unit 100 preferably has a large sensitivity difference between the signal m1 and the signal m2 with respect to the direction of the target sound. On the other hand, it is preferable that there is no difference in sensitivity characteristics between the signal m1 and the signal m2 in directions other than the direction of the target sound. In order to suppress the noise component mixed in the main signal based on the noise reference signal in a situation where noise is simultaneously received from a plurality of directions, the output of the spectrum ratio calculation unit 43 shown in FIG. This is because it needs to be constant regardless of the direction of noise arrival. That is, if the output of the spectrum ratio calculation unit 43 changes depending on the noise arrival direction, only the estimated noise spectrum Nx (ω) in a specific direction can be accurately obtained. Therefore, it is preferable that the directivity patterns of the signal m1 and the signal m2 are different in shape only in the blind spot portion of directivity and the shape is the same in other portions.

  Here, if the balance of sensitivity between the third microphone unit 3 and the fourth microphone unit 4 is lost when the signals from the microphone units 3 and 4 are subtracted, the zero point that requires the most accuracy, that is, the directivity. The sensitivity of the blind spot increases. Utilizing this property, by providing the signal amplification unit 150 on the signal m1 side and setting the signal amplification factor to about 0.85, a directivity pattern as shown in FIG. 16B can be obtained. FIG. 16B is a diagram showing directivity patterns in the signal m1 and the signal m2 in FIG. As shown in FIG. 16B, in the seventh embodiment, it is possible to obtain a directivity pattern whose shape is different only in the blind spot portion of directivity and whose shape is substantially the same in other portions.

  As described above, according to the seventh embodiment, it is possible to obtain the signal m1 and the signal m2 having different sensitivity characteristics only with respect to the target sound direction. Therefore, a good suppression effect can be obtained in the subsequent noise suppression processing.

(Embodiment 8)
Next, a microphone device according to Embodiment 8 will be described with reference to FIG. As in the fifth embodiment, the eighth embodiment employs a configuration in which a main signal and a noise reference signal are obtained using a plurality of microphone units having the same directivity characteristics.

  FIG. 17A shows a part of the configuration of the microphone device according to Embodiment 8. In FIG. 17A, the directivity synthesis unit 100 further includes an angle setting unit 160 and a second signal delay unit 112 in addition to the configuration shown in FIG. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  The configuration shown in FIG. 17A is different from the configuration shown in FIG. 16A in that an angle setting unit 160 is further provided and a second signal delay unit 112 is provided after the fourth microphone unit 4. . The basic operation in FIG. 17A is the same as that in FIG. 17A is different from the operation in FIG. 16A in that the target sound direction can be changed by the angle setting unit 160.

  The angle setting unit 160 can change the signal delay amount τ1 of the first signal delay unit 111 within a range of 0 ≦ τ1 ≦ 2d / c (where d is the interval between the microphone units and c is the speed of sound). And Here, when the second signal delay unit 112 is not provided, even if the signal delay amount τ1 of the first signal delay unit 111 is changed in the above range, the range from 0 ° to + 90 ° with respect to the front direction. However, the target sound direction can only be changed. Therefore, by providing the second signal delay unit 112 and setting the signal delay amount τ2 to τ2 = d / c, the target sound direction is changed within a range of ± 90 ° with respect to the front direction.

  As described above, in the eighth embodiment, the sound collection direction (target sound direction) of the microphone device can be made variable. For example, the directivity pattern shown in FIG. 17B can be realized, or the directivity pattern shown in FIG. 17C can be realized by changing the signal delay amount of the signal delay unit. Is possible. The variable delay characteristic is easily realized by configuring the signal delay unit with an all-pass filter H (ω) = (A + z−1) / (1 + A · z−1) and setting the coefficient A to 0 ≦ A <1. can do. When changing the signal delay amount, the coefficient A is changed by the angle setting unit 160. If a large delay amount or linearity of delay frequency characteristics is required, a secondary all-pass filter and / or an all-pass filter may be connected in cascade.

(Embodiment 9)
Next, a microphone device according to Embodiment 9 will be described with reference to FIG. As in the fifth embodiment, the ninth embodiment employs a configuration in which a main signal and a noise reference signal are obtained using a plurality of microphone units having the same directivity characteristics.

  FIG. 18A illustrates a part of the configuration of the microphone device according to the ninth embodiment. The microphone device includes a third microphone unit 3, a fourth microphone unit 4, a directivity synthesis unit 100, and an angle setting unit 160. The arrangement of the microphone units 3 and 4 is the same as the arrangement shown in FIG. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  18A, the directivity synthesis unit 100 includes a third signal delay unit 121, a first signal delay unit 101, a fourth signal delay unit 122, a second signal delay unit 102, A first signal subtracting unit 103 and a second signal subtracting unit 104 are provided. The third signal delay unit 121 delays the signal output from the third microphone unit 3. The first signal delay unit 101 delays the signal output from the fourth microphone unit 4. The first signal subtracting unit 103 subtracts the signal output from the first signal delay unit 101 from the signal output from the third signal delay unit 121. Thereby, the signal m1 which is the main signal can be obtained. The fourth signal delay unit 122 delays the signal output from the fourth microphone unit 4. The second signal delay unit 102 delays the signal output from the third microphone unit 3. The second signal subtracting unit 104 subtracts the signal output from the second signal delay unit 102 from the signal output from the fourth signal delay unit 122. As a result, a signal m2 that is a noise reference signal can be obtained. The angle setting unit 160 independently controls the signal delay amount of the first signal delay unit 101 and the signal delay amount of the second signal delay unit 102.

  In FIG. 18A, the configuration on the signal m1 side is configured symmetrically with respect to the configuration on the signal m2 side. As a result, the directivity pattern of the signal m1 and the directivity pattern of the signal m2 are controlled independently. Therefore, the directivity patterns of the signal m1 and the signal m2 may be designed to emphasize the sensitivity in the target sound direction. it can. Specifically, as shown in FIG. 18B, the directivity pattern of the signal m1 has a directivity that is as sensitive as possible in the target sound direction and that provides a noise suppression effect. Further, as shown in FIG. 18C, the directivity pattern of the signal m2 is formed so that the blind spot direction of directivity coincides with the target sound direction.

  As described above, in the ninth embodiment, the noise suppression process at the subsequent stage is supplementarily used, and noise is positively suppressed by the directivity synthesis at the previous stage. Therefore, in the ninth embodiment, the directivity pattern of the signal m1 is formed with priority. Here, since the directivity synthesis is a linear process, there is a feature that speech waveform distortion or the like hardly occurs. On the other hand, the noise suppression process is a non-linear process in which the filter coefficient changes with time, so that there are cases where speech waveform distortion occurs due to errors of various estimation units such as a noise spectrum. From this point of view, whether to use the directivity pattern shown in FIGS. 17B and 17C or the directivity pattern shown in FIGS. 18B and 18C depends on the usage environment (purpose It is preferable to select appropriately according to the volume of sound, ambient noise level, reflection, reverberation, etc.), application (calling, voice recognition, recording, etc.), required noise suppression amount, and the like.

(Embodiment 10)
Next, a microphone device according to Embodiment 10 will be described with reference to FIG. The object of the tenth embodiment is to obtain a main signal and a noise reference signal necessary for noise suppression processing of the present invention in a device in which the directivity main axes of two microphone units are oriented in different directions.

  FIG. 19 is a diagram illustrating a part of the configuration of the microphone device according to the tenth embodiment. The microphone device includes a third microphone unit 3, a fourth microphone unit 4, and a directivity re-synthesis unit 200. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  In FIG. 19, the positions where the microphone units 3 and 4 are arranged are the same as the positions shown in FIG. However, in FIG. 19, the third microphone unit 3 has a directional main axis directed in a direction rotated by a predetermined angle with respect to the front direction. The fourth microphone unit 4 has a directional main axis directed in a direction rotated by a predetermined angle with respect to the front direction (the direction of rotation opposite to that of the third microphone unit 3). Here, a signal output from the third microphone unit 3 is referred to as a right channel signal, and a signal output from the fourth microphone unit 4 is referred to as a left channel signal.

  In FIG. 19, the directivity re-synthesis unit 200 includes a signal addition unit 205 and a signal subtraction unit 204. The signal adding unit 205 adds the right channel signal and the left channel signal. As a result, the signal m1, which is the main signal, can be obtained. The signal subtracting unit 204 subtracts the right channel signal from the left channel signal. As a result, a signal m2 that is a noise reference signal can be obtained.

  Note that the configuration of FIG. 19 assumes that the present invention is applied to a device in which a one-point stereo microphone is used, such as a video movie. For example, in this device, stereo sound collection is usually performed, and when only the front direction is emphasized as the target sound, re-synthesis of directivity as described below may be performed.

  In a normal one-point stereo microphone, the left and right microphones are arranged so that the sound coming from the center (front direction shown in FIG. 19) has the same phase in the left and right microphone units, taking into account the sound image localization during playback. The unit has the same amplitude and phase characteristics. Further, as described above, the directivity angles of the microphone units 3 and 4 are set to be equal in the left and right microphone units. Therefore, by adding the right channel signal and the left channel signal in the signal adding unit 205, a signal m1 having directivity in the front direction can be obtained. The signal subtracting unit 204 subtracts the right channel signal from the left channel signal, thereby obtaining a signal m2 having a directional blind spot in the front direction. As described above, the signal m1 and the signal m2 generated by the directivity re-synthesis unit 200 are the same as the signal m1 and the signal m2 of Embodiment 1, respectively. Accordingly, it is possible to perform noise suppression processing and reflection characteristic distortion correction processing using the signal m1 and the signal m2.

  As described above, according to the tenth embodiment, the sound in the target sound direction can be enhanced using the signal output from the one-point stereo microphone. Therefore, a device having a one-point stereo microphone can function as a zoom microphone, for example. In Embodiment 10, since directivity recombination is performed based on a stereo signal, it can be applied to multi-channel sound collection in which a stereo signal and a signal in the front direction can be obtained simultaneously. Even if the stereo microphone is an analog circuit, the same effect as described above can be obtained.

(Embodiment 11)
Next, a microphone device according to Embodiment 11 will be described with reference to FIG. The object of the eleventh embodiment is to obtain a main signal and a noise reference signal necessary for noise suppression processing of the present invention in a device that generates a stereo signal.

  FIG. 20 is a diagram showing a configuration of the microphone device according to Embodiment 11. In FIG. 20, the microphone device includes a fifth microphone unit 5, a sixth microphone unit 6, a directivity synthesis unit 500, and a directivity resynthesis unit 200. Each microphone unit 5 and 6 is an omnidirectional microphone unit having the same characteristics. The arrangement positions of the microphone units 5 and 6 are the same as the arrangement shown in FIG. Directivity synthesis section 500 receives signals output from microphone units 5 and 6 as inputs, and outputs right channel signal Rch and left channel signal Lch. The directivity re-synthesis unit 200 receives the right channel signal Rch and the left channel signal Lch as input, and is a signal m1 that is a main signal having sensitivity in the target sound direction and a noise reference signal having a directivity blind angle in the target sound direction. The signal m2 is output. The target sound direction can be set to a direction other than the front direction.

  In FIG. 20, the directivity re-synthesis unit 200 includes an inverse directivity synthesis unit 250 and a directivity synthesis unit 100. The reverse directivity synthesis unit 250 receives the signals (right channel signal Rch and left channel signal Lch) output from the directivity synthesis unit 500. The reverse directivity synthesis unit 250 generates an omnidirectional signal from the right channel signal Rch and the left channel signal Lch. The directivity synthesis unit 100 is the same as that shown in the fifth embodiment. However, here, the angle setting unit 160 is not provided. In FIG. 20, the directivity synthesis unit 100 is configured as shown in FIG. 18A, but the directivity synthesis unit 100 is configured as shown in FIGS. 15, 16A, and 17A. There may be.

  In the eleventh embodiment, the stereo signals (right channel signal Rch and left channel signal Lch) obtained by directivity synthesis section 500 are re-converted into signals output from microphone units 5 and 6 by reverse directivity synthesis section 250. Converted. That is, the stereo signal is reconverted into two omnidirectional signals. Further, the omnidirectional signal obtained by the reconversion is converted into a main signal and a noise reference signal for detecting a target sound coming from a predetermined direction by the directivity synthesis unit 100.

ここで、左辺のｘ１およびｘ２は第５および第６のマイクロホンユニット５および６からそれぞれ出力される信号であり、右辺のＲｃｈおよびＬｃｈが指向性合成部５００から出力されるステレオ信号である。なお、指向性合成部５００については、一般に用いられている指向性合成のための構成であるので、詳細説明は省略する。なお、式（１）において、１／（１−Ｈτ４（ω））の部分は、６ｄｂ／ｏｃｔの周波数特性補正項になる。実際のマイクロホン装置では補正が行われるが、指向特性とは別に考えられるのでここでは無視している。指向性合成部５００によって得られたステレオ信号（信号Ｒｃｈおよび信号Ｌｃｈ）をマイクロホンユニットから出力された信号（信号ｘ１および信号ｘ２）に戻すには、式（１）の左辺第２項の行列の逆行列を両辺の左側から掛ければよく、いわゆる逆フィルタによって実現することができる。このことを数式で表現すると（２）、（３）のようになる。

従って、信号Ｒｃｈおよび信号Ｌｃｈに式（３）の処理を行うことによって、逆指向性合成を実現することができる。図２０に示す逆指向性合成部２５０は、式（３）を図示したものである。指向性合成部１００は、このようにして得られた信号ｘ１および信号ｘ２から、目的音方向に感度を持つ主信号ｍ１と、目的音方向に指向性死角を持つ雑音参照信号ｍ２とを生成する。 Here, the directivity synthesis unit 500 for outputting a stereo signal includes a first signal delay unit 501, a first signal subtraction unit 521, a second signal delay unit 502, and a second signal subtraction unit. 522. The first signal delay unit 501 delays and outputs the signal output from the sixth microphone unit 6. The first signal subtracting unit 521 subtracts the signal output from the first signal delay unit 501 from the signal output from the fifth microphone unit 5, and outputs the signal Rch obtained as a result of the subtraction. The second signal delay unit 502 delays and outputs the signal output from the fifth microphone unit 5. The second signal subtracting unit 522 subtracts the signal output from the second signal delay unit 502 from the signal output from the sixth microphone unit 6, and outputs the signal Lch obtained as a result of the subtraction. The operation of the directivity synthesis unit 500 described above is expressed as follows.

Here, x1 and x2 on the left side are signals output from the fifth and sixth microphone units 5 and 6, respectively, and Rch and Lch on the right side are stereo signals output from the directivity synthesis unit 500. Note that the directivity synthesis unit 500 is a commonly used configuration for directivity synthesis, and thus detailed description thereof is omitted. In Equation (1), the 1 / (1-Hτ4 (ω)) portion is a frequency characteristic correction term of 6 db / oct. Although correction is performed in an actual microphone device, it is ignored here because it can be considered separately from directivity characteristics. In order to return the stereo signals (signal Rch and signal Lch) obtained by the directivity synthesis unit 500 to the signals (signal x1 and signal x2) output from the microphone unit, the matrix of the second term on the left side of the equation (1) What is necessary is just to multiply an inverse matrix from the left side of both sides, and it can implement | achieve by what is called an inverse filter. This can be expressed by equations (2) and (3).

Therefore, reverse directivity synthesis can be realized by performing the processing of Expression (3) on the signal Rch and the signal Lch. The reverse directivity synthesis unit 250 illustrated in FIG. 20 illustrates Equation (3). The directivity synthesis unit 100 generates a main signal m1 having sensitivity in the target sound direction and a noise reference signal m2 having directivity blind angle in the target sound direction from the signals x1 and x2 obtained in this way. .

  As described above, in the eleventh embodiment, a signal output from the one-point stereo microphone is used. Even in this case, an effect similar to that of the tenth embodiment can be obtained. That is, the target sound coming from the front direction can be emphasized and frequency distortion due to reflection can be corrected. In the eleventh embodiment, it is possible to deal with a target sound coming from an arbitrary direction.

  The eleventh embodiment is particularly effective when a signal output from the microphone unit cannot be obtained and only a stereo signal is available. In other words, according to the eleventh embodiment, a configuration for obtaining a main signal of an objective sound and an ideal noise reference signal can be realized even in a device that generates a stereo signal.

  FIG. 21 is a diagram illustrating an application example of the eleventh embodiment. FIG. 21 is a diagram showing a system including an audio recording device 801 and an audio reproduction device 802. The audio recording device 801 includes fifth and sixth microphone units 5 and 6 and a directivity synthesis unit 500. The recording unit 803 is a recording medium that can be attached to and detached from the audio recording device 801 and the audio reproduction device 802. The audio reproduction device 802 includes a directivity resynthesis unit 200. Although not shown, the audio playback device 802 has the configuration of the microphone device according to any one of the first to fourth embodiments.

  In FIG. 21, a signal Rch and a signal Lch are recorded in the recording unit 803 of the audio recording device 801. As a result, audio information is recorded in the recording unit 803. When the recording unit 803 in which the audio information is recorded is attached to the audio reproduction device 802, the audio reproduction device 802 reads information recorded in the recording unit 803. Specifically, the signal Rch and the signal Lch are read by the directivity resynthesis unit 200. The directivity re-synthesis unit 200 generates a main signal and a noise reference signal from the read signal Rch and signal Lch. By using the main signal and the noise reference signal, it is possible to perform noise suppression processing on the target sound.

  As described above, even when the audio recording device 801 and the audio reproduction device 802 are separate bodies, the configuration of the eleventh embodiment can be realized. That is, it is possible to perform noise suppression processing during reproduction on a signal once recorded in the recording unit 803 such as a video movie.

  FIG. 22 is a diagram illustrating an application example of the audio reproduction device illustrated in FIG. 21. In FIG. 22, the audio playback device 802 includes an image display unit 900 and an angle setting unit 160 in addition to the configuration described in FIG. 21. That is, the audio reproducing device 802 shown in FIG. 22 has an image display function, and is realized by, for example, a digital video camera.

  In FIG. 22, the recording unit 803 records image information to be displayed on the image display unit in addition to the audio information described in FIG. The audio information and image information are information related to each other such as image (video) and audio information recorded simultaneously by a digital video camera. The audio information and the image information are simultaneously reproduced by the audio reproduction device 802. Here, during the reproduction of the audio information and the image information, the user instructs the angle using the angle setting unit 160. At this time, the user determines the angle while viewing the image displayed on the image display unit. For example, if the subject is displayed at the center of the screen of the image display unit, the user instructs an angle indicating a direction corresponding to the center of the screen (that is, the front direction). Thereby, the user can extract and listen to the sound coming from the front direction as the target sound.

  In other embodiments, the following configuration is also conceivable. FIG. 23 is a diagram illustrating a part of the configuration of a microphone device according to another embodiment. In FIG. 23, the fifth microphone unit 5, the sixth microphone unit 6, and the directivity synthesis unit 500 are the same as those shown in FIG. The directivity re-synthesis unit 200 has the same configuration as that shown in FIG. Also with the configuration shown in FIG. 23, the same effect as described above can be obtained. In addition, the structure after obtaining the signal m1 and the signal m2 uses the structure in any one of Embodiment 1-4.

  As described above, according to the present invention, with respect to the directional microphone output directed toward the target sound direction, the stationary and non-stationary noises are suppressed in directions other than the target sound direction. A microphone with can be obtained. At the same time, the influence on the frequency characteristic of the reflected wave received by the microphone device can be removed. In this way, both additive noise and multiplicative noise reflected waves can be suppressed simultaneously, and high S / N and always flat microphone frequency characteristics are not affected by the sound field. Can be realized. In addition, the noise suppression processing unit realizes a configuration that reduces a processing delay, thereby enabling application to a loud voice or a call that does not allow a large delay. In addition, sound in various directions can be extracted by combining combinations of directivity synthesis, reverse directivity synthesis, and directivity resynthesis, which are preprocessing, and similar effects can be obtained on the playback device side.

  As described above, the microphone device and the playback device of the present invention can be used for the purpose of stably operating even under a plurality of noises in an actual use environment and realizing a high S / N.

1 is a block diagram showing a configuration of a microphone device according to Embodiment 1. FIG. The figure which shows the structure of the determination part shown in FIG. The figure which shows the example of the state of an audio | voice detection in case the direction of the dominant sound is (theta) 1-theta (3) direction. The figure which shows the structural example of the noise suppression filter coefficient calculation part 40. The figure which shows the structural example of the time-varying coefficient filter part 50. The figure which shows the other structural example of the time-varying coefficient filter part 50. The figure which shows the specific example of each signal shown in FIG. FIG. 3 is a block diagram showing a configuration of a microphone device according to Embodiment 2. The figure explaining the difference in the internal state of a microphone apparatus with the case where there is a reflector, and the case where there is no reflector FIG. 4 is a block diagram showing a configuration of a microphone device according to Embodiment 3. FIG. 9 is a block diagram showing another configuration of the microphone device according to Embodiment 3. Block diagram showing a configuration of a microphone device according to Embodiment 4 The figure which shows the directivity pattern of the microphone device The figure which shows a part of structure of the microphone apparatus which concerns on Embodiment 5. FIG. The figure which shows a part of structure of the microphone apparatus which concerns on Embodiment 6. FIG. The figure which shows a part of structure of the microphone apparatus which concerns on Embodiment 7. FIG. The figure which shows a part of structure of the microphone apparatus which concerns on Embodiment 8. FIG. The figure which shows a part of structure of the microphone apparatus which concerns on Embodiment 9. FIG. FIG. 10 shows a part of a configuration of a microphone device according to Embodiment 10; The figure which shows the structure of the microphone apparatus based on Embodiment 11. FIG. FIG. 15 shows an application example of the eleventh embodiment. The figure which shows the application example of the audio | voice reproduction apparatus shown in FIG. The figure which shows a part of structure of the microphone apparatus in other embodiment. The figure which shows the structure of the microphone apparatus of the prior art example 1. The figure which shows the structure of the microphone apparatus of the prior art example 2. The figure which shows the structure of the microphone apparatus of the prior art example 3.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 1st microphone unit 2 2nd microphone unit 3 3rd microphone unit 4 4th microphone unit 10 Determination part 20 Adaptive filter part 30 Signal subtraction part 40 Noise suppression filter coefficient calculation part 50 Time-varying coefficient filter part 60 Reflection Information calculation unit 70 Reflection correction unit 90 Detection threshold setting unit

Claims (15)

A microphone device for detecting a target sound coming from a target sound direction,
A signal generation unit that generates a main signal indicating a result detected with sensitivity to the target sound direction , and a noise reference signal indicating a result detected by directing the sensitivity blind angle with respect to the target sound direction;
A determination unit that determines whether a level ratio indicating a ratio of the signal level of the main signal to the signal level of the noise reference signal generated by the signal generation unit is greater than a predetermined value;
The main signal generated by the signal generation unit is filtered by an adaptive filter to generate a signal indicating the signal component of the target sound included in the noise reference signal generated by the signal generation unit, and by the determination unit An adaptive filter unit that learns filter coefficients only when the level ratio is determined to be greater than a predetermined value;
Before Kizatsu sound reference signal, the generated by the adaptive filter section, and a subtraction unit for subtracting a signal indicating the signal components of the target sound included in the noise reference signal,
Using the main signal and a noise reference signal after subtraction by the subtraction unit, a noise suppression unit that suppresses a signal component of noise included in the main signal ,
The noise suppression unit calculates a filter coefficient of a noise suppression filter for suppressing signal components other than the target sound signal from the main signal based on the main signal and the noise reference signal after subtraction by the subtraction unit. A noise suppression filter coefficient calculation unit for
A microphone device , comprising: a time-varying coefficient filter unit that performs filtering on the main signal, reflecting the filter coefficient calculated by the noise suppression filter coefficient calculation unit . A microphone device for detecting a target sound coming from a target sound direction,
A signal generation unit that generates a main signal indicating a result detected with sensitivity to the target sound direction , and a noise reference signal indicating a result detected by directing the sensitivity blind angle with respect to the target sound direction;
A determination unit that determines whether a level ratio indicating a ratio of the signal level of the main signal to the signal level of the noise reference signal generated by the signal generation unit is greater than a predetermined value;
The main signal generated by the signal generation unit is filtered by an adaptive filter to generate a signal indicating the signal component of the target sound included in the noise reference signal generated by the signal generation unit, and by the determination unit An adaptive filter unit that learns filter coefficients only when the level ratio is determined to be greater than a predetermined value;
Before Kizatsu sound reference signal, the generated by the adaptive filter section, and a subtraction unit for subtracting a signal indicating the signal components of the target sound included in the noise reference signal,
Based on the filter coefficient of the adaptive filter unit, a reflection information calculation unit that calculates information on the arrival time difference between the direct wave and the reflected wave of the target sound;
A microphone device comprising: a reflection correction unit that corrects a distortion of a frequency characteristic generated in a main signal due to a reflected wave of a target sound based on information calculated by the reflection information calculation unit. The signal generator is
A first microphone unit arranged such that a directional main axis is directed to a target sound direction;
Directivity dead angle direction viewed contains a second microphone unit is arranged directed to the target sound direction,
The microphone device according to claim 1 or 2, wherein an output signal from the first microphone unit is the main signal, and an output signal from the second microphone unit is the noise reference signal . Provided between the output terminal and the subtraction of the noise reference signal in the signal generating unit further includes a signal delay unit for delaying the noise reference signal so as to satisfy the convergence condition of the adaptive filter of the adaptive filter section The microphone device according to claim 1 or 2. The microphone device according to claim 1, wherein the predetermined value is changeable. The signal generator is
A first microphone unit;
A second microphone unit having the same characteristics as the first microphone unit;
A delay unit that delays a signal output from the first microphone unit by a predetermined delay amount;
An amplifier for amplifying the signal output from the delay unit;
A first subtraction unit that generates a main signal by subtracting the signal amplified by the amplification unit from the signal output from the second microphone unit;
A second subtracting unit that generates a noise reference signal by subtracting the signal output from the delay unit from the signal output from the second microphone unit ;
Before SL predetermined delay amount, the blind spot direction of the directional characteristic noise reference signal output from the second subtracting unit has is set to face the target sound direction,
The microphone device according to claim 1 or 2, wherein an amplification factor in the amplification unit is set such that the main signal has higher sensitivity in the target sound direction than the noise reference signal .   The microphone device according to claim 6, further comprising a setting unit that changes a predetermined delay amount set in the delay unit. The signal generator is
A first microphone unit;
A second microphone unit having the same characteristics as the first microphone unit;
Based on the signals output from the first and second microphone units, a main signal is generated so as to be sensitive to the target sound direction, and a noise signal is used so that the sensitivity in the target sound direction is minimized. The microphone device according to claim 1, further comprising: a synthesis unit that generates The signal generator is
A first microphone unit;
A second microphone unit disposed with a directional main axis directed in a direction different from the first microphone unit;
A signal adder that generates a main signal by adding a signal output from the first microphone unit and a signal output from the second microphone unit;
And a signal subtracting unit that generates a noise reference signal by subtracting the other from one of the signal output from the first microphone unit and the signal output from the second microphone unit. The microphone device according to claim 1 or 2. The signal generator is
A first microphone unit;
A second microphone unit having the same characteristics as the first microphone unit;
A stereo signal generator for generating a stereo signal composed of a right channel signal and a left channel signal based on the first and second microphone units;
Based on the stereo signal, an inverse synthesis unit that generates each signal output from each microphone unit;
Based on each signal generated by the inverse synthesizing unit, a main signal indicating a result detected with sensitivity to the target sound direction and a sound arriving from a direction other than the target sound direction from the target sound. The microphone device according to claim 1, further comprising: a synthesizing unit that generates a noise reference signal indicating a detection result with high sensitivity. The signal generator is
A first microphone unit;
A second microphone unit having the same characteristics as the first microphone unit;
A stereo signal generator for generating a stereo signal composed of a right channel signal and a left channel signal based on the first and second microphone units;
A signal adding unit that generates a main signal by adding the right channel signal and the left channel signal of the stereo signal;
The microphone device according to claim 1, further comprising: a signal subtracting unit that generates a noise reference signal by subtracting the other from one of the right channel signal and the left channel signal of the stereo signal. Based on the filter coefficient of the adaptive filter unit, a reflection information calculation unit that calculates information on the arrival time difference between the direct wave and the reflected wave of the target sound;
Based on the information calculated by the reflection information calculation unit, further comprising a reflection correction unit that corrects distortion of the frequency characteristics generated in the main signal by the reflected wave of the target sound,
The noise suppression unit suppresses a signal component of noise included in the main signal by using the main signal after correction by the reflection correction unit and the noise reference signal after subtraction by the subtraction unit. The microphone device described. The noise suppression filter coefficient calculation unit is
A first frequency analysis unit for calculating a power spectrum of the main signal;
A second frequency analysis unit that calculates a power spectrum of the noise reference signal after subtraction by the subtraction unit;
Wherein the determination unit only when the level ratio is determined the Most smaller than a predetermined value, said the power spectrum calculated by the first frequency analysis unit, the power spectrum calculated by the second frequency analysis unit A power spectrum ratio calculation unit for calculating a time average of the power spectrum ratio with
A multiplier that multiplies the time average of the power spectrum ratio calculated by the power spectrum ratio calculator by the power spectrum calculated by the second frequency analyzer;
The microphone according to claim 1 , further comprising: a coefficient calculation unit that calculates a filter coefficient of the noise suppression filter based on a power spectrum calculated by the first frequency analysis unit and a multiplication result by the multiplication unit. apparatus. An audio recording unit for recording audio signals of at least two types of channels;
Based on the audio signal recorded in the recording unit, the main signal indicating the detection result with sensitivity to the target sound direction, and the detection result with the sensitivity dead angle directed to the target sound direction are shown. A signal generator for generating a noise reference signal;
A determination unit that determines whether a level ratio indicating a ratio of the signal level of the main signal to the signal level of the noise reference signal generated by the signal generation unit is greater than a predetermined value;
The main signal generated by the signal generation unit is filtered by an adaptive filter to generate a signal indicating the signal component of the target sound included in the noise reference signal generated by the signal generation unit, and by the determination unit An adaptive filter unit that learns filter coefficients only when the level ratio is determined to be greater than a predetermined value;
Before Kizatsu sound reference signal, the generated by the adaptive filter section, and a subtraction unit for subtracting a signal indicating the signal components of the target sound included in the noise reference signal,
Using the main signal and the noise reference signal after subtraction by the subtraction unit, a noise suppression unit that suppresses a signal component of noise included in the main signal;
A reproducing unit that reproduces a main signal in which a noise signal component is suppressed by the noise suppressing unit ;
The noise suppression unit calculates a filter coefficient of a noise suppression filter for suppressing signal components other than the target sound signal from the main signal based on the main signal and the noise reference signal after subtraction by the subtraction unit. A noise suppression filter coefficient calculation unit for
An audio reproduction device comprising: a time-varying coefficient filter unit that performs filtering on the main signal, reflecting the filter coefficient calculated by the noise suppression filter coefficient calculation unit . A video recording unit for recording a video signal related to the audio signal recorded in the audio recording unit;
A video reproduction unit for reproducing the video signal recorded in the video recording unit;
A direction receiving unit that receives an input of a direction in which the sound should be emphasized from a user,
The audio reproduction device according to claim 14 , wherein the signal generation unit generates a main signal and a noise reference signal with a direction received by the direction reception unit as a target sound direction.

Images