JP2011205324A - Voice processor, voice processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice processor and a voice processing method capable of easily outputting voice signals in which sound volumes of a plurality of voices mixed with the voice signals are corrected.SOLUTION: The voice processor includes a voice signal outputting section 110 which outputs an arbitrary frequency band component among the voice signals obtained by a plurality of nondirectional microphones arranged apart from one another, an arrival direction discriminating section 120 which discriminates an arrival direction of a voice based on a phase difference between frequency band components of the voice signals, a voice correction amount deriving section 130 which derives correction amounts of sound volumes for the voice signals according to the arrival direction of the voice, and a sound volume correction executing section 140 which corrects the sound volumes of the voice signals by using the derived correction amount.

Description

  The present invention relates to a voice processing device, a voice processing method, and a program.

Conventionally, the sound reproduced based on the sound signal on which the sounds from a plurality of speakers are superimposed has a problem that the sound from a specific speaker is small and difficult to hear.
In response to these issues, the speech signal with the speech from multiple speakers is separated into speech for each speaker using a technique such as independent component analysis, and then the volume is corrected for each speech. Is known to do.

  However, in order to separate voices for each speaker using the above-described method such as independent component analysis, a high-level, complicated and large amount of computation is required. In the case of mounting on a device having such a voice processing function, there is a problem that the cost increases and the power consumption increases.

In order to solve such a problem, a technique for appropriately correcting the volume of a sound signal acquired by a sound collecting device that acquires sound signals from a specific position using a plurality of microphones is known (Patent Document 1).
In the technique described in Patent Document 1, one sound collecting device is configured by using three microphones including two directional microphones and one omnidirectional microphone, and the sound collecting direction of the sound by the sound collecting device is determined. This is a technique for identifying and adjusting the level of collected sound for each direction of sound collection.

JP 2009-17343 A

However, the technique of Patent Document 1 requires two directional microphones and one omnidirectional microphone, and there is a problem that it is difficult to easily realize these microphones because there are complicated restrictions on the arrangement positions of these microphones. there were.
Moreover, since the technique of patent document 1 uses three microphones in total, including two directional microphones and one omnidirectional microphone, as the configuration of one sound collection device, the number of microphones used is high and the cost is high. In addition, there is a problem that a sufficient area must be secured in the apparatus for mounting the microphone.
Furthermore, since the sound collection direction specified by the sound collection device is limited to a predetermined direction in the technique of Patent Document 1, the direction of the sound source cannot be specified, for example, when a plurality of sound sources are close to the sound collection device. In some cases, there was a problem that fine control was not possible.

  Therefore, in order to solve the above-described problem, the present invention collects sound by using a plurality of omnidirectional microphones for a sound signal in which a plurality of sounds are mixed, and determines the direction of arrival of each sound. An object of the present invention is to provide an audio processing device and an audio processing method that easily output an audio signal in which the volume of a plurality of audio signals mixed with the signal is corrected.

  In order to achieve the above-described object, the present invention provides an audio signal output unit that outputs an arbitrary frequency band component among audio signals respectively acquired by a plurality of omnidirectional microphones arranged apart from each other; An arrival direction discriminating unit that discriminates the arrival direction of the voice collected by the microphone based on the phase difference between the frequency band components of the audio signal output from the audio signal output unit, and the arrival direction discriminating unit. A sound correction amount deriving unit for deriving a sound volume correction amount for the sound signal according to the arrival direction, and a sound volume correction for correcting the sound signal volume using the correction amount derived by the sound volume correction amount deriving unit. And an execution unit.

  According to the present invention, the direction of arrival of the voice collected by the microphone is determined based on the phase difference between the predetermined frequency band components in the voice collected by the omnidirectional microphone of the voice signal output unit, and collected by the microphone. By deriving the corrected amount of the sound according to the direction of arrival, the volume of the sound collected by the microphone can be corrected according to the direction of arrival.

Therefore, since the sound volume from different directions of arrival can be corrected so that the sound volume between the plurality of sounds is equalized from the sound signal in which the sounds from the plurality of directions of arrival are superimposed, it is superimposed on the sound signal. It is possible to reproduce each of the recorded voices in an easy-to-understand manner.
In addition, since the sound is corrected for each direction of arrival of the plurality of sounds superimposed on the sound signal without separating the plurality of sounds superimposed on the sound signal, the sound signal can be easily converted into the sound signal with a low amount of computation. It is possible to correct the volume of each superimposed voice.

It is a block diagram which shows the structure of the audio | voice processing apparatus concerning the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the audio | voice processing apparatus concerning the 1st Embodiment of this invention. It is a block diagram which shows the structure of the audio processing apparatus concerning the 2nd Embodiment of this invention. It is a figure which illustrates notionally the positional relationship between the audio | voice processing apparatus in this invention, and several audio | voices acquired. It is a figure which shows notionally the relationship between the position of a sound source, and the audio | voice signal picked up by the omnidirectional microphone. It is a figure which shows notionally the relationship between a sound collection area | region and the volume of the acquired audio | voice signal. It is a figure which shows notionally the relationship between a sound collection area | region and the corrected amount with respect to the acquired audio | voice signal. It is a flowchart which shows operation | movement of the audio processing apparatus concerning the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the audio processing apparatus concerning the 3rd Embodiment of this invention. It is a figure which illustrates notionally the determination of the division | segmentation position information in the speech processing unit concerning the 3rd Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[First Embodiment]
The speech processing apparatus according to the first embodiment of the present invention acquires a speech signal on which a plurality of speeches from different directions of arrival are superimposed by using a plurality of microphones, and an arbitrary frequency in the speech signal obtained by these microphones. This is a sound processing apparatus that corrects the volume of a sound signal acquired according to the direction of arrival of sound determined based on the phase difference between band components.

As shown in FIG. 1, the audio processing apparatus 10 according to the present embodiment includes an audio signal output unit 110, an arrival direction determination unit 120, a volume correction amount derivation unit 130, and a volume correction execution unit 140.
The audio signal output unit 110 outputs an arbitrary frequency band component among the audio signals respectively acquired by a plurality of omnidirectional microphones arranged apart from each other.
The arrival direction discriminating unit 120 discriminates the arrival direction of the voice collected by the microphone of the voice signal output unit 110 based on the phase difference between arbitrary frequency bands of the voice signal output from the voice signal output unit 110.

The sound volume correction amount deriving unit 130 derives a sound volume correction amount for the sound signal acquired by the microphone according to the sound arrival direction collected by the microphone of the sound signal output unit 110 determined by the arrival direction determining unit 120. To do.
The volume correction execution unit 140 corrects the volume of the audio signal acquired by the microphone of the audio signal output unit 110 using the correction amount derived by the volume correction amount deriving unit 130.

  Each component of the speech processing apparatus 10 according to the present embodiment is realized by installing a computer program (software) in a computer including a CPU, a memory, and an interface, and various functions of the speech processing apparatus 10 described above. Is realized by the cooperation of various hardware resources of the computer and the computer program.

Next, the operation of the speech processing apparatus 10 according to the present embodiment will be described with reference to FIG.
As shown in FIG. 2, the audio processing device 10 according to the present exemplary embodiment transmits an arbitrary frequency band component to the audio signal output unit 110 out of the audio signals respectively acquired by the plurality of microphones of the audio signal output unit 110. Each is output (S101).

  When each arbitrary frequency band component in the audio signal acquired by each of the plurality of microphones is output from the audio signal output unit 110, the arrival direction determination unit 120 determines whether each frequency band component output from the audio signal output unit 110 is between the frequency band components. Based on the phase difference, the direction of arrival of the voice collected by the microphone is determined (S102).

When the arrival direction determination unit 120 determines the arrival direction of the sound collected by the microphone, the sound correction amount deriving unit 130 derives the sound volume correction amount for the sound signal acquired by the microphone according to the sound arrival direction ( S103).
When the volume correction amount for the audio signal is derived, the volume correction execution unit 140 corrects the volume of the audio signal using the derived correction amount (S103).

As described above, according to the sound processing device according to the present embodiment, the arrival of the sound collected by the microphone based on the phase difference between the predetermined frequency band components in the sound collected by the microphone of the sound signal output unit. By determining the direction and deriving the correction amount of the voice collected by the microphone according to the direction of arrival, the volume of the voice collected by the microphone can be corrected according to the direction of arrival.
Therefore, since the sound volume from different directions of arrival can be corrected so that the sound volume between the plurality of sounds is equalized from the sound signal in which the sounds from the plurality of directions of arrival are superimposed, it is superimposed on the sound signal. It is possible to reproduce each of the recorded voices so that they can be easily heard.

[Second Embodiment]
FIG. 3 is a block diagram showing the configuration of the speech processing apparatus according to the second embodiment of the present invention. The sound processing apparatus according to the present embodiment performs sound volume correction on a sound signal on which a plurality of sounds are superimposed according to the direction of arrival of the sound. This is a voice processing device that determines the direction of arrival of a plurality of voices superimposed on the voice signal based on the voice signal acquired by the directional microphone and corrects the volume of the voice signal.
In addition, about the component of the speech processing device concerning this Embodiment, the same code | symbol is attached | subjected to what has the structure and function similar to the component of the speech processing device 10 demonstrated in 1st Embodiment. Detailed description thereof will be omitted.

As shown in FIG. 3, the audio processing device 20 according to the present embodiment includes an audio signal output unit 210, an arrival direction determination unit 220, a volume correction amount derivation unit 230, and a volume correction execution unit 240.
The audio signal output unit 210 includes a sound collection unit 211 having two omnidirectional microphones 211-a and 221-b, and a frequency analysis unit that converts the audio signal acquired by the sound collection unit 211 into an arbitrary frequency band component. 212.

The sound collection unit 211 of the audio signal output 210 arranges the omnidirectional microphones 211-a and 211-b apart from each other, and the audio signals acquired by the omnidirectional microphones 211-a and 221-b, respectively. Output.
The frequency analysis unit 212 generates and outputs a predetermined frequency band component for the audio signals acquired by the omnidirectional microphones 211-a and 211-b output from the sound collection unit 211. Here, the frequency band in the frequency band component of the audio signal output from the frequency analysis unit 212 can be a preset frequency band. For example, it may be a frequency band component in an audio signal for each frequency band set in advance within the range of the human voice frequency band.

  The arrival direction determination unit 220 determines the arrival direction of the audio signal acquired by each of the omnidirectional microphones 211-a and 211-b output from the frequency analysis unit 212, between the predetermined frequency band components of the audio signal. It discriminate | determines for every sound collection area | region specified based on a phase difference.

<Determination of direction of arrival of audio signal>
Here, the function of determining the arrival direction of the audio signal by the arrival direction determination unit 220 will be specifically described.
FIG. 4 shows a terminal A on which the sound processing apparatus according to the present embodiment is mounted, and sounds (sound source 1 to sound source 3) collected by the two omnidirectional microphones 211-a and 211-b of the terminal A. It is a figure which shows an example of the positional relationship of no. The terminal A picks up sounds coming from different positions by different sound sources (sound sources 1 to 3) by the omnidirectional microphones 211-a and 211-b, and collects all the sounds picked up by these microphones, that is, the sound sources 1 to 1. The audio signal including the audio of 3 is output.

  As shown in FIG. 4, when collecting sounds coming from different positions of different sound sources, the sound signals including sound sources 1 to 3 collected by the omnidirectional microphone 211-a of the terminal A and omnidirectionality There is a difference in sound collection time between the sound signals including the sound sources 1 to 3 collected by the microphone 211-b according to the position of the sound source. Here, FIG. 5 shows a conceptual diagram of the relationship between the position of the sound source and the audio signal picked up by the two omnidirectional microphones.

For example, as shown in FIG. 5, there is a difference in sound collection time between a sound waveform obtained by collecting sound from the sound source 1 by the omnidirectional microphone 211-a and a sound waveform collected by the omnidirectional microphone 211-b. There is a sound collection time difference c between a sound waveform obtained when a sound is collected from the sound source 3 by the omnidirectional microphone 211-a and a sound waveform obtained by the omnidirectional microphone 211-b. .
Such a sound collection time difference occurs according to the positional relationship between the omnidirectional microphones 211-a and 211-b and the sound source. That is, among the sounds collected by the omnidirectional microphone 211-a, the omnidirectional microphone 211-a is closer to the sound source 1 than the omnidirectional microphone 211-b for the sound from the sound source 1. Therefore, the sound is picked up earlier by time a than the sound picked up by the omnidirectional microphone 211-b, and the sound from the sound source 3 is delayed by time c because it is located farther than the omnidirectional microphone 211-b. Sound is collected.

On the other hand, the sound from the sound source 2 is collected at the same time because the omnidirectional microphones 211-a and 211-b are at equidistant positions, and there is no deviation in the sound collection time.
In this way, there is a time difference in the timing of the sound collected by the two omnidirectional microphones depending on the position of the sound source, and the time difference is acquired by the two omnidirectional microphones based on the difference in the sound collection timing. The position of the sound source can be specified by the phase difference between the sound signals.
The voice arrival direction determination unit 220 is output from the frequency analysis unit 212 among the voice signals acquired by the omnidirectional microphones 211-a and 211-b as described with reference to FIG. 5 described above. By extracting the phase difference between the frequency components, the arrival direction of the sound collected by the sound collection unit 211 is determined.

Specifically, for example, the voice arrival direction determination unit 220 stores in advance information on the phase difference of the voice signal associated with the sound collection area so as to be divided into three areas. The voice arrival direction determination unit 220 stores in advance the phase difference extracted between the frequency components output from the frequency analysis unit 212 out of the voice signals acquired by the omnidirectional microphones 211-a and 211-b, respectively. The information of the phase difference associated with each collected sound area is compared, and the sound collection area of the frequency component having the extracted phase difference is determined.
That is, the arrival direction determination unit 220 extracts the phase difference for all frequency components of the audio signal acquired by the sound collection unit 211, and associates the extracted phase difference with each pre-stored sound collection region. The direction of arrival of the sound collected by the sound collection unit 2211 is determined by comparing the information with the phase difference information.

The volume correction amount deriving unit 230 derives a volume correction coefficient that defines a volume correction amount for each direction of arrival of the audio signal acquired by the sound collection unit 211.
The volume correction amount deriving unit 230 estimates the volume level of the audio signal for each sound collection area based on the volume level of the frequency component for which the sound collection area has been determined by the arrival direction determination unit 220, and the volume estimation. Correction coefficient derivation for deriving a volume correction coefficient for each sound collection area indicating the amount of correction for the sound volume of the sound signal acquired by the sound collection section 211 based on the volume level of each sound collection area of the sound signal estimated by the section 231 Part 232.

  Here, a sound volume level estimation function for each sound collection region of the sound signal by the sound volume estimation unit 231 of the sound volume correction amount deriving unit 230 and a sound volume indicating the correction amount for the sound signal volume by the correction coefficient deriving unit 232 for each sound collection region. The correction coefficient derivation function will be specifically described.

<Estimation of audio signal volume>
The volume estimation unit 231 calculates a volume level for each frequency component of the audio signal acquired by the sound collection unit 211 output from the frequency analysis unit 212. For example, the volume estimation unit 231 can derive the volume level of the frequency component based on the spectrum energy in the frequency component of the audio signal output from the frequency analysis unit 212, and The volume level of this frequency component may be derived based on the amplitude value (voltage value).

The volume estimation unit 231 calculates the volume level of each frequency component of the audio signal output from the frequency analysis unit 212, associates the calculated volume level with the arrival direction of the frequency component having the volume level, and The relationship between the direction of arrival and the volume of each frequency component is derived. That is, the sound volume estimation unit 231 derives the relationship between the phase difference and the sound volume level in an arbitrary frequency component of the audio signal for each sound collection area determined by the arrival direction determination unit 220.
The sound volume estimation unit 231 estimates the maximum sound volume level in the sound collection area as the sound signal volume in the sound collection area based on the relationship between the derived sound collection area and the sound volume level in each frequency component of the sound signal.

  Specifically, for example, the volume for each sound collection region estimated by the sound volume estimation unit 231 is indicated by a solid line in FIG. 6 which is an example of the relationship between the sound collection region and the volume level of each frequency component of the audio signal. Thus, among the frequency components of the audio signal in the three sound collection areas (areas 1 to 3), the maximum volume level of the frequency component in area 1 is set as the volume of the audio signal in area 1, and these are similarly applied to areas 2 and 3 as well. The maximum volume level of the frequency component in the region is estimated as the volume of the audio signal in these regions.

<Derivation of volume correction coefficient>
The correction coefficient deriving unit 232 determines whether or not a desired audio signal exists for each sound collection region based on the sound volume of the sound signal for each sound collection region estimated by the sound volume estimation unit 231, and the desired sound signal A correction amount is calculated so as to make the volume of the sound signal of each sound collection region where the signal exists uniform, and a sound volume correction coefficient that determines this correction amount for each sound collection region is derived.

  Specifically, for example, as illustrated in FIG. 6, the correction coefficient deriving unit 232 sets a predetermined threshold in advance for the sound volume for each sound collection region of the audio signal estimated by the sound volume estimation unit 231. An audio signal having a volume larger than the threshold value can be determined as a desired audio signal acquired by the sound collection unit 211. This threshold may be set in advance as a specified value, or may be determined according to the amount of ambient noise by appropriately calculating the volume of ambient noise included in the sound collected by the sound collection unit 211.

  In the example illustrated in FIG. 6, among the sound volume for each sound collection area of the audio signal estimated by the sound volume estimation unit 231, the sound volume of the area 1 and the area 3 exceeds the threshold value, and the sound volume of the area 2 is set to the threshold value. It is below. That is, the correction coefficient estimator 232 indicates that the desired sound signal is present in the sound signal from the region 1 and the sound region 3 and the desired sound signal is not present in the sound signal from the region 2. Can be determined.

When it is determined whether or not a desired sound signal exists for each sound collection area, the correction coefficient estimation unit 232 collects the sound signal having the highest volume in the sound collection area where the desired sound signal exists. A volume correction amount is calculated for each sound collection area so that the sound signal volume of the sound area is the same as that of the other sound collection areas.
For example, in the example illustrated in FIG. 6, the correction coefficient deriving unit 232 adjusts the sound of the region 3 so as to match the sound volume of the sound signal of the region 1 having a high sound volume with respect to the regions 1 and 3 where the desired sound signal exists. A correction amount for the volume of the signal is calculated. Further, the sound signal in the region 2 is not corrected because the sound volume is lower than the threshold value.

  That is, the correction coefficient deriving unit 232 does not perform correction on the volume of the audio signal in the areas 1 and 2, but corrects the volume of the audio signal in the area 3 to be the same as the volume of the audio signal in the area 1. A predetermined volume correction coefficient is derived. For example, the correction amount of the area where correction is not performed can be 1, and the correction amount s of the area where correction is performed can be s = (maximum volume of the area) / (volume of the correction area). The correction amount s with respect to the volume of the audio signal in the region 3 in the example shown in FIG. 6 is calculated by s = V1 / V3.

The correction coefficient deriving unit 232 derives a volume correction coefficient in which the correction amount with respect to the volume of the audio signals in the areas 1 and 2 shown in FIG.
The volume correction coefficient derived by the correction coefficient deriving unit 232 may limit the corrected volume so that the corrected volume does not become excessive. For example, the volume correction coefficient may be adjusted so that the corrected volume does not exceed a preset limit value, and α is appropriately set with s · α (0 <α <1) as the derived volume correction coefficient. Thus, the corrected volume may be adjusted.

  Here, FIG. 7 shows an example of the volume correction coefficient derived by the correction coefficient deriving unit 232. As shown in FIG. 7, the correction amount for the volume of the audio signal in the area 1 and area 2 shown in FIG. 6 derived by the correction coefficient deriving unit 232 is 1, and the correction amount for the volume of the audio signal in the area 3 is s. The determined volume correction coefficient is indicated by a wavy line. When the volume correction coefficient indicated by the wavy line as shown in FIG. 7 is applied to the volume correction of the audio signal in each area, the volume correction that is discontinuous with respect to the audio signal on the boundary between the area 2 and the area 3 is executed. It will be. This may cause abnormal noise. Therefore, the correction coefficient deriving unit 232 derives a sound volume correction coefficient that does not cause discontinuous sound volume correction.

  For example, as shown by a solid line in FIG. 7, linear interpolation is performed so that the phase difference at the center of each section of the region 2 and the region 3 becomes the correction amount of each region. It is possible to eliminate the discontinuous points of the correction amount in. Further, not only linear interpolation but also nonlinear interpolation such as secondary interpolation may be used, and smoothing may be performed using past correction amounts.

The volume correction execution unit 240 corrects the volume of the audio signal acquired by the sound collection unit 211 by reflecting the volume correction coefficient derived by the volume correction amount deriving unit 230 in the audio signal output from the audio signal output unit 210. To do.
Specifically, the volume of the frequency component is corrected using a predetermined frequency component of the audio signal output by the frequency analysis unit 212 and the volume correction coefficient derived by the correction coefficient deriving unit 232. For example, any frequency components X _i (f, t) Volume Volume correction coefficient C _i (f, t) of the frequency component corrected using the Y _i (f, t) and, Y _i (f, t ) = X _i (f, t) · C _i (f, t). Here, f is a frequency index and t is a time index.

  Each component of the speech processing apparatus 20 according to the present embodiment is realized by installing a computer program (software) in a computer including a CPU, a memory, and an interface, and various functions of the speech processing apparatus 20 described above. Is realized by the cooperation of various hardware resources of the computer and the computer program.

Next, the voice processing operation of the voice processing apparatus 20 according to the present embodiment will be described with reference to the flowchart shown in FIG.
As shown in FIG. 8, the sound processing device 20 according to the present embodiment collects sound by using the two omnidirectional microphones 211-a and 211-b mounted on the sound collection unit 211 (S <b> 201).

The audio signals acquired by the two omnidirectional microphones 211-a and 211-b are respectively divided into frequency components for each frequency band set in advance by the frequency analysis unit 212 and output (S202).
When each frequency component of the audio signal acquired by the two omnidirectional microphones 211-a and 211-b is output from the frequency analysis unit 212, the arrival direction determination unit 220 is output for each predetermined frequency band. The phase difference between the frequency component of the audio signal acquired by the omnidirectional microphone 211-a and the frequency component of the audio signal acquired by the omnidirectional microphone 211-b is extracted (S203).

  When the phase difference of the frequency component is extracted for each predetermined frequency band, the arrival direction determination unit 220 collects sound based on the relationship between the phase difference of the frequency component stored in advance and the sound collection area and the extracted phase difference. The arrival direction of each frequency component of the audio signal acquired by the unit 211 is determined for each preset sound collection area (S204).

  When the direction of arrival of each frequency component of the audio signal acquired by the sound collection unit 211 is determined for each sound collection region, the volume correction amount deriving unit 230 estimates the sound volume of the sound signal for each sound collection region and collects the sound signal. A volume correction coefficient that determines a correction amount for the volume of the audio signal for each sound region is derived (S205).

  When the sound volume correction coefficient is derived by the sound volume correction amount deriving unit 230, the sound volume correction executing unit 240 is configured to correct each frequency component of the audio signal output from the frequency analysis unit 212 and the sound volume correction output from the sound volume correction amount deriving unit 230. Using the coefficient, the volume of the audio signal acquired by the sound collection unit 211 is corrected for each sound collection region (S206).

As described above, the sound processing device 20 according to the present embodiment is configured to perform processing between the frequency components divided for each predetermined frequency band with respect to the sound collected by the two omnidirectional microphones that are installed apart from each other. By determining the direction of arrival of this frequency component based on the phase difference and deriving the amount of volume correction for each direction of arrival so that the volume of the sound signal from all directions of arrival is equalized, multiple sounds are superimposed. The volume of the plurality of sounds can be corrected so that the volume between the plurality of sounds becomes equal from the received signal.
Therefore, it is possible to correct the volume of a specific sound that is difficult to hear among the sound signals on which a plurality of sounds are superimposed, and it is possible to reproduce the sound signal that is easy to hear.

In addition, the audio processing apparatus according to the present embodiment is a frequency obtained by dividing a sound signal on which a plurality of sounds acquired by an omnidirectional microphone are superimposed without being separated into individual sounds and divided into predetermined frequency bands. Since the arrival direction of the audio signal is determined by deriving the phase difference between the components and the volume of the audio signal is corrected for each arrival direction, each audio is separated after being separated using a technique such as independent component analysis. The volume of a plurality of sounds can be easily corrected with a smaller amount of computation than the process of executing the volume correction.
Therefore, the audio processing device according to the present embodiment can be mounted on a device having an audio processing function, such as a general-purpose terminal device or an audio conference device, with reduced cost and reduced power consumption.

[Third Embodiment]
FIG. 9 is a block diagram showing the configuration of the speech processing apparatus according to the third embodiment of the present invention. The sound processing apparatus according to the present embodiment appropriately sets the sound signal collection area of the sound signal according to the volume of each frequency component of the acquired sound signal in the function of the sound processing apparatus 20 described in the second embodiment. The function to be added is added.
In addition, about the structure and function of the speech processing apparatus 30 concerning this Embodiment, what has the same structure and function as the speech processing apparatus 20 demonstrated in 2nd Embodiment attaches | subjects the same code | symbol, and these The detailed description of is omitted.

  As shown in FIG. 9, the audio processing device 30 according to the present exemplary embodiment uses each frequency component of the audio collected by two omnidirectional microphones 211-a and 211-b that are installed apart from each other. The audio signal output unit 210 that outputs the sound, the arrival direction determination unit 320 that determines the arrival direction of each frequency component of the sound output from the audio signal output unit 210, and the volume of the audio signal acquired by the audio signal output unit 210 A sound volume correction amount deriving unit 230 for deriving a correction amount according to the direction of arrival of sound, and by reflecting the sound volume correction amount derived by the sound volume correction amount deriving unit 230 in the sound output by the sound signal output unit 210, A volume correction execution unit 240 that corrects the volume of audio is configured.

  Of the components of the audio processing device 30 according to the present embodiment described above, the arrival direction determination unit 320 includes the frequency band component in the phase difference between the frequency band components of the audio signal output from the audio signal output unit 210. A sound collection area specifying unit 321 that specifies a sound collection area according to the volume level is further provided.

Here, the sound collection area specifying function by the sound collection area specifying unit 321 will be described in detail.
The sound collection region specifying unit 321 includes the sound acquired by the omnidirectional microphone 211-a among the frequency components of the sound signal acquired by the sound collection unit 211 output from the frequency analysis unit 212 for each predetermined frequency band. The phase difference between the frequency component of the signal and the frequency component of the audio signal acquired by the omnidirectional microphone 211-b is extracted, and the extracted phase difference and volume level (spectrum or frequency component in the frequency component) are extracted from the frequency component of the audio signal. The relationship with the voltage value is derived.

FIG. 10 conceptually illustrates the relationship between the phase difference in the frequency component of the audio signal derived by the sound collection area specifying unit 321 and the volume level.
As shown in FIG. 10, the sound collection area specifying unit 321 stores the extracted phase difference and volume level in association with each frequency component of the audio signal acquired by the sound collection unit 211 (× shown in FIG. 10). The relationship between the phase difference in the frequency component of the audio signal and the volume level is derived.

The sound collection area specifying unit 321 specifies the sound collection area based on the relationship between the phase difference in the derived frequency component of the audio signal and the volume level.
Specifically, from the relationship between the phase difference and volume level in each frequency component of the audio signal acquired by the sound collection unit 211, the relationship between the phase difference and volume level as shown by the dashed line in FIG. The curve shown is derived. Spline interpolation may be used as an interpolation method.

The sound collection area specifying unit 321 detects peaks and valleys of a curve indicating the relationship between the phase difference of the audio signal derived by interpolation and the volume level, and specifies the phase difference indicating the valley as the boundary position of the sound collection area. . For example, as shown in FIG. 10, the phase difference D1 is the boundary point between the sound collection region 1 and the region 2, and the phase difference D2 is specified as the boundary point between the region 2 and the region 3.
In the example of FIG. 10, valleys are detected after interpolation and specified as boundary points of the sound collection area. However, in order to reduce the amount of calculation, valleys are detected by using neighboring values without performing interpolation, and are collected. You may specify as a boundary point of a sound area.

  When the sound collection region is specified by the sound collection region specifying unit 321, the sound signal volume correction amount obtained by the sound signal output unit 210 is derived by the sound correction amount deriving unit 230 for each sound collection region, and the sound volume correction is performed. The execution unit 240 corrects the volume of the audio signal.

As described above, according to the sound processing device according to the present embodiment, by specifying the sound collection area according to the acquired sound signal and correcting the volume of the sound signal according to the sound collection area, The arrival direction of the sound source can be appropriately specified according to the arrival directions of the collected voices.
Therefore, since the volume can be corrected even for a moving sound source, it is possible to generate high-quality output sound.

  The present invention can be used for a telephone terminal and a TV conference system for performing a voice call, and a recording device such as an IC recorder having a voice recording function.

  DESCRIPTION OF SYMBOLS 10,20,30 ... Audio | voice processing apparatus, 110,210 ... Audio | voice signal output part, 120,220,320 ... Arrival direction determination part, 321 ... Sound collection area | region identification part, 130,230 ... Volume correction amount derivation | leading-out part, 231 ... Volume estimation unit, 232... Correction coefficient derivation unit, 140, 240.

Claims (6)

An audio signal output unit that outputs an arbitrary frequency band component among audio signals respectively acquired by a plurality of omnidirectional microphones arranged apart from each other;
A direction-of-arrival determination unit that determines a direction of arrival of sound collected by the microphone based on a phase difference between the frequency band components of the sound signal output from the sound signal output unit;
A volume correction amount derivation unit that derives a correction amount for the volume of the audio signal according to the direction of arrival determined by the direction of arrival determination unit;
A sound processing apparatus comprising: a sound volume correction execution unit that corrects a sound volume of the sound signal using the correction amount derived by the sound volume correction amount deriving unit. The speech processing apparatus according to claim 1,
The arrival direction determining unit determines the arrival direction of the sound collected by the microphone for each sound collection region specified according to the phase difference between the frequency band components,
The volume correction amount deriving unit derives a correction amount for the volume of the audio signal for each sound collection region based on the volume level of the frequency band component for which the sound collection region is determined by the arrival direction determination unit. A voice processing apparatus characterized by the above. The speech processing apparatus according to claim 2, wherein
The voice direction characterized in that the arrival direction determination unit specifies the sound collection region according to a volume level of the frequency band component in a phase difference between arbitrary frequency band components output from the voice signal output unit apparatus. An audio signal output step for outputting an arbitrary frequency band component among audio signals respectively acquired by a plurality of omnidirectional microphones arranged apart from each other;
A direction-of-arrival determination step of determining a direction of arrival of the sound collected by the microphone based on a phase difference between the frequency band components of the sound signal output from the sound signal output step;
A volume correction amount derivation step for deriving a correction amount for the volume of the audio signal according to the direction of arrival determined by the direction of arrival determination step;
A sound processing method comprising: a sound volume correction executing step of correcting the sound volume of the sound signal using the correction amount derived in the sound volume correction amount deriving step. The voice processing method according to claim 4,
The direction of arrival determination step determines the direction of arrival of sound collected by the microphone for each sound collection area specified according to the phase difference between the frequency band components,
The volume correction amount derivation step derives a correction amount for the volume of the audio signal for each sound collection area based on the volume level of the frequency band component in which the sound collection area is determined by the arrival direction determination step. A voice processing method characterized by the above.   A speech processing program for causing a computer to execute the speech processing method according to claim 4 or 5.

Images