JP2010232862A - Audio processing apparatus, audio processing method, and program - Google Patents

Abstract

An audio processing apparatus, an audio processing method, and a program for suitably reducing noise of a sound signal input from a microphone in a plurality of environments.
An audio processing apparatus includes: a position pattern detecting unit that detects an index of a relative position between a sound source and a plurality of microphones; and an audio process for a sound signal input from each of the microphones. A processing determining unit 103 that determines based on the index, and a signal processing unit 104 that executes the determined audio processing on the sound signal.
[Selection] Figure 1

Description

  The present invention relates to a sound processing apparatus, and more particularly to a sound processing apparatus that can obtain a target sound having a high SNR by classifying the positions of a sound source and a microphone into N patterns and performing processing corresponding to each position pattern.

  Conventionally, a technique for estimating a target sound source direction and extracting a signal from a target sound source with high SNR by collecting sound using a plurality of microphones called a microphone array and performing signal processing on the collected sound. It has been known.

  For example, in Non-Patent Document 1, a target sound is received by a microphone array, the arrival time difference of the target sound to each microphone is corrected for each signal received by each microphone, and these signals are added together. A method of obtaining a signal in which the target sound is emphasized by taking the sum is shown. The invention disclosed in Non-Patent Document 1 is based on the premise that a signal mixed with target sound and noise is input to any microphone.

  Further, as a method of using a plurality of microphones, two microphones are used, one is a noise collecting microphone, the other is a microphone that collects a target sound mixed with noise, and noise from the microphone signal collecting the target sound is detected. There is known a method for reducing noise by subtracting the output of a collecting noise microphone and extracting a target sound more clearly.

  As an example, in Japanese Patent Application Laid-Open No. 2004-226656 (Patent Document 1), two microphones are used, and the distance between the lip and a preselected reference microphone is set to the signal level of the reference microphone and the other microphone. A technique such as a speaker distance detection device is disclosed that calculates from the difference between the two and adjusts the subtraction amount when subtracting the signal of the other microphone from the signal of the reference microphone according to the distance.

  The invention disclosed in Patent Document 1 is based on the premise that one microphone contains a signal in which target sound and noise are mixed, but the other microphone has only noise or a relatively small amount of target sound even if mixed. ing.

JP 2004-226656 A

J. et al. L. Flaganan, J. et al. D. Johnston, R.A. Zahn and G.M. W. Elko, "Computer-steered microphone arrays for sound production in large rooms," J. Acoustic. Soc. Am. , Vol. 78, no. 5, pp. 1508-1518, 1985

  However, when processing sound signals generated using multiple microphones, the environment in which the target sound and noise are mixed in all microphones, the target sound and noise are input to one microphone, and the noise is input to the other microphones. If the same processing is performed in an environment where the main sound enters, the target sound may not be appropriately processed. In the inventions disclosed in Non-Patent Document 1 and Patent Document 1, this is not taken into consideration.

  The present invention has been invented in order to solve these problems in view of the above points, and has an object to suitably reduce noise of a sound signal input from a microphone in a plurality of environments.

  In order to solve the above-described problem and achieve the object, one aspect of the present invention is input from a position pattern detection unit that detects an index of a relative position between a sound source and a plurality of microphones, and each of the plurality of microphones. A sound signal processing unit that determines sound processing for the sound signal based on the relative position index; and a signal processing unit that executes the sound processing determined for the sound signal. .

  ADVANTAGE OF THE INVENTION According to this invention, it becomes possible to reduce suitably the noise of the sound signal input from a microphone in a some environment.

It is a block diagram which shows the audio | voice processing apparatus concerning 1st Embodiment. It is a figure which shows the flowchart which shows the process in the audio | voice processing apparatus which concerns on 1st Embodiment. It is a figure (the 1) which shows a position pattern. It is a figure (the 2) which shows a position pattern. It is a figure (the 3) which shows a position pattern. It is a figure which shows the example of a speech processing unit at the time of classifying with a position pattern (2). It is a figure which shows the example of a speech processing unit at the time of classifying with a position pattern (1). It is a block diagram which shows the audio processing apparatus concerning 2nd Embodiment. It is a figure which shows the flowchart which shows the operation | movement of the audio processing apparatus concerning 2nd Embodiment. It is a block diagram which shows the audio processing apparatus concerning 3rd Embodiment. It is a figure which shows the flowchart which shows the operation | movement of the speech processing unit concerning 3rd Embodiment. It is a figure explaining the example which provided the angle sensor in the mobile telephone. It is explanatory drawing which shows the hardware constitutions of the audio processing apparatus concerning this Embodiment.

(First embodiment)
FIG. 1 is a block diagram showing a speech processing apparatus according to the first embodiment. The sound processing apparatus 100 according to the first embodiment collates the input sound signal with a position pattern of a sound source held in advance, and executes sound processing corresponding to each position pattern. The sound processing apparatus 100 includes a sound input unit 101, a position pattern detection unit 102, a processing determination unit 103, a signal processing unit 104, and a pattern database (hereinafter referred to as “pattern DB”) 109.

  The sound input unit 101 converts input sounds from a plurality of microphones into digitized sound signals, and detects the start and end of the sound. The position pattern detection unit 102 detects a position pattern index between the sound source and the microphone from the sound signal. The process determining unit 103 determines a process to be performed on the sound signal by collating the position pattern index with a previously held position pattern. The signal processing unit 104 performs processing according to the determination of the processing determination unit 103.

  The pattern DB 109 holds information related to position patterns of a plurality of microphones and sound sources. The position pattern represents a relative position (relative position) between the plurality of microphones and the sound source. The pattern DB 109 stores an index of a pattern of sound signals input from a plurality of microphones in association with each position pattern. The pattern stored in the pattern DB 109 is called by the process determination unit 103 and collated with the index detected by the position pattern detection unit 102.

  FIG. 2 is a flowchart showing processing in the speech processing apparatus 100 according to the first embodiment. Here, an example in which the position pattern of the sound source and the microphone is acquired using two microphones will be described. The two microphones are called microphone 1 and microphone 2, respectively.

  In step S101, the sound input unit 101 uses an AD converter to convert sound input to the microphone from an analog signal to a digital signal.

  In step S102, voice section detection using, for example, the number of zero crossings is performed in order to detect the start and end of the voice to be subjected to noise processing. This voice section detection is performed using the microphone outputs of the microphone 1 and the microphone 2.

  More specifically, the number of zero crossings is calculated with respect to the sound signal acquired by the microphone 1 and the sound signal acquired by the microphone 2, and if it is determined that a voice section has been detected by either microphone, from the detection point Treat sound as speech.

  Here, the start end information detected by the microphone 1 and the microphone 2 is held as S1 and S2, respectively. After the voice termination is confirmed in the output microphone in which the voice start edge is detected latest, the processing in FIG. Note that the section detection method is not limited to this, and various section detection methods can be applied. Further, for example, a section detection method unique to a plurality of microphones may be applied.

  In step S103, the position pattern detection unit 102 detects an index of the position pattern of the sound source and the microphone using the audio signal detected by the sound input unit 101. As the index, for example, a difference in voice arrival time between microphones and a signal level ratio are used.

  More specifically, for example, when the microphone 1 is used as a reference, the sound source is closer to the microphone 1 as the difference in the voice arrival time to the microphone 2 increases. Further, the sound source is closer to the microphone 1 as the signal level on the microphone 1 side is larger than that of the microphone 2.

  When calculating these two indexes, the initial speech section of the target speech is used. The initial voice is a voice in a certain section after the voice is detected. The difference in voice arrival time between microphones is calculated by using cross-correlation. The time when the start of the microphone whose ear start is detected earliest is detected as time 0, the sound signal input to the microphone 1 is set as x1, and the correlation calculation section of x1 is set as time te from time ts (S1 <ts <te). Let x1 'be the waveform normalized in power in the interval. Further, the audio signal input to the microphone 2 is set to x2, and the time T of the section [0-T] for obtaining the audio arrival time difference is set at least from ts to td from the time of detection of the start end of the microphone in which the audio start end is detected latest Set to take more than. For example, when S1 <S2, T is set by the following equation (1).

In Expression (1), the outgoing voice arrival time difference td can be expressed by Expression (5) using the following Expressions (2) to (4).

This arrival time difference td is one of the position pattern determination indexes between the sound source and each microphone. In the case of two microphones, if the arrival time difference td from the microphone 1 to the microphone 2 is obtained, the arrival time difference from the microphone 2 to the microphone 1 is obtained by reversing the positive and negative signs.
The signal level ratio dd between the microphone 1 and the microphone 2 can be obtained by the following equation using td obtained previously.

The signal level ratio dd in Expression (6) is one of the position pattern determination indexes between the sound source and each microphone. Note that the position pattern determination index is not limited to that described above, and various standards can be applied. For example, the maximum value of the correlation calculated earlier is included in this. If the maximum correlation value is higher than a certain reference, the sound source and the two microphones are equidistant. If the maximum correlation value is lower than a certain reference, the sound source is close to one microphone, and the sound source is close to one microphone. The position pattern that is far can be derived. The maximum correlation value r _max is calculated by the following equation (7).

  In step S104, the processing determining unit 103 uses the index for determining the position pattern calculated by the position pattern detecting unit 102 to check which of the following three position patterns (1) to (3) belongs to. To do. 3 to 5 are diagrams showing three position patterns.

(1) A sound source is approaching the microphone 1 (FIG. 3).
(2) A sound source is approaching the microphone 2 (FIG. 4).
(3) The sound source is not close to either microphone (FIG. 5).

When the arrival time difference determination threshold value tthr and the signal level difference determination threshold _{values dd_thre1} and _{dd_thre2} are constants (where _{t_thre} > 0, _{dd_thre1} > _{dd_thre2} > 0), when td> 0, the following equation (8) And when Expression (9) holds, the position pattern is classified into (1).

（１）及び（２）の何れにも分類されなければ、位置パターンは（３）に分類される。 Further, when td <= 0, when the following expressions (10) and (11) hold, the position pattern is classified into (2).

If it is not classified into either (1) or (2), the position pattern is classified into (3).

  In step S105, the signal processing unit 104 performs a predetermined process according to the classified position pattern. 6 and 7 are diagrams illustrating an operation when the signal processing unit 104 switches the signal processing. FIG. 6 is a diagram illustrating an example of the voice processing device when classified as the position pattern (2), and FIG. 7 illustrates an example of the voice processing device when classified as the position pattern (1). FIG.

  Hereinafter, switching of signal processing will be described.

When the position pattern is classified as (1), the sound input to the microphone 1 is processed as the target sound and the sound input to the microphone 2 is processed as noise. Specifically, the output speech o of the speech processing apparatus 100 can be expressed by the following equation (12) using the delay time td calculated earlier, where α is a constant (0 ≦ α).

  At this time, spectrum subtraction may be performed by converting the signal into the frequency domain. Alternatively, a method of removing a noise component from x1 using x2 as a reference signal using an adaptive filter often used in an echo canceller or the like is also possible.

When the position pattern is classified as (2), the sound input to the microphone 2 is processed as the target sound, and the sound input to the microphone 1 is processed as noise. The specific process is the same as when the microphone 1 and the microphone 2 in the process for the position pattern (1) are switched. At this time, the output sound o is expressed by the following equation (13).

Other processing can be considered when the sound source is classified as a position pattern such that the sound source is approaching a specific microphone. For example, α may be a function of the maximum correlation value and the subtraction amount may be adjusted. At this time, the value of α can be controlled by the following equation (14) by a linear function having a and b as constants.

  By expressing α as in Expression (14), the subtraction amount can be reduced when the maximum correlation value is high, and the subtraction amount can be increased when the maximum correlation value is low.

When the position pattern is classified into (3), the delay sum array process is performed using both the voices input to the microphone 1 and the microphone 2. When the delay sum array is used, the output sound o is expressed by the following equation (15).

Note that this array processing adaptation unit is not limited to the above-mentioned method. For example, by applying Griffiths-Jim type array processing, a dead angle of noise is formed with respect to a certain angle by two microphones, and the sound in the range is used. It is possible to extract the target speech o having a high SNR.
Returning to FIG. 2, in step S <b> 106, the sound input unit 101 detects the end, and the sound processing ends.

  As described above, the embodiment of the present invention has been described by taking two microphones as an example. However, in implementing the present invention, it is not essential that there are two microphones, and the present invention is divided into three or more microphones. It is also possible to expand. In the case of three microphones, if the microphones are microphone 1, microphone 2, and microphone 3, the following seven position patterns are prepared.

(1 ′) The microphone 1 is approaching.
(2 ′) Approaching the microphone 2.
(3 ′) Approaching the microphone 3.
(4 ′) Close to microphones 1 and 2.
(5 ′) Close to microphones 2 and 3.
(6 ′) Close to microphones 1 and 3.
(7 ') No microphone is approaching.

The position pattern to be classified is determined by using the arrival time difference and the signal level difference described above for the input sound signal. More specifically, the arrival time difference from the microphone 1 to the microphone 2 is set to td _12, and the calculation is performed by the equations (2) to (5). Similarly, the arrival time differences between the other microphones are td ₁₃ , td ₂₁ , td ₂₃ , td ₃₁ , td _32, and these are calculated. The signal level difference for the microphone 2 microphones 1 and _{dd 12,} is calculated by the equation (6). Similarly, the signal level differences between the other microphones are calculated as dd ₁₃ , dd ₂₁ , dd ₂₁ , dd ₂₃ , dd ₃₁ , and dd ₃₂ .

  At this time, the arrival time difference between the microphone n1 closest to the sound source and the other two microphones is a positive value. The microphone n2, which is the second closest to the sound source, has a positive arrival time difference with respect to the remaining microphone and is negative with respect to the microphone n1. The microphone n3 farthest from the sound source has a negative arrival time difference from the other two microphones. Therefore, based on this characteristic, first, which microphone is to be the microphone n1, the microphone n2, or the microphone n3 is determined.

Mike n1 and arrival time difference _td between the microphone n2 _N1N2, the arrival time difference threshold _{td thre1,} microphone n1 and the signal level difference _{dd N1N2} with microphone n2, the threshold of the signal level difference is _{dd thre1,} the following equation When (16) and Expression (17) are satisfied, when the microphone 1 is the microphone n1 (1 ′), when the microphone 2 is the microphone n1 (2 ′), and when the microphone 3 is the microphone n1 (3) ') Position pattern.

Next, the arrival time difference between the microphone n1 and the microphone n2 is td _n1n2 , the arrival time difference threshold is td _thre1 , the arrival time difference between the microphone n2 and the microphone n3 is td _n2n3 , the arrival time difference threshold is td _thre2 , and the microphone n1 The signal level difference with the microphone n2 is dd _n1n2 , the signal level difference threshold is dd _thre1 , the signal level difference between the microphone n2 and the microphone n3 is dd _n2n3 , the signal level difference threshold is dd _thre2 , and the following formula ( 18) to the expression (21), when the microphone 3 is the microphone n3 (4 '), when the microphone 1 is the microphone n3 (5'), and when the microphone 2 is the microphone n3 (6 ') The position pattern is classified.

  If the position pattern is not classified into any of the position patterns (1 ') to (6'), it is regarded that the distance of the sound source with respect to all the microphones is long, and the position pattern is classified into the position pattern (7 ').

但し、α１、α２は定数であり、α１≧０、α２≧０である。 After the classification, the process is switched according to each pattern. More specifically, in the case of (1 ′), (2 ′), and (3 ′), the processing according to Expression (22) is performed, and noise is subtracted from the target sound of the microphone close to the sound source.

However, α1 and α2 are constants, and α1 ≧ 0 and α2 ≧ 0.

In the case of (4 ′), (5 ′), and (6 ′), the processing according to Expression (23) is performed. As a result, the two microphones close to the sound source are emphasized by the delay sum array, and the output of the microphone farthest from the sound source is used for noise subtraction.

このように３つのマイクロホンにも容易に拡張可能である。 In the case of (7 ′), the processing is performed according to the equation (24). As a result, voice enhancement is performed with a delay-and-sum array using all microphones.

Thus, it can be easily expanded to three microphones.

Further, the sound source position may be estimated in a three-dimensional space using three or more microphones. When the sound source position can be estimated, the distance from each microphone to the sound source can be calculated. The distances between the microphone and the sound source obtained by this processing are denoted by ld ₁ , ld ₂ , and ld ₃ , respectively.

同様に、（２’）〜（７’）の位置パターンへの分類も実現することができる。 At this time, if the following formula (25) is satisfied with the distance threshold _{ld_thre} as a constant, the distance is classified as (1 ′).

Similarly, classification into position patterns (2 ′) to (7 ′) can also be realized.

(Second Embodiment)
FIG. 8 is a block diagram showing a speech processing apparatus according to the second embodiment. The sound processing apparatus 100a according to the second embodiment selects and performs processing corresponding to each sound source position acquired by the position sensor on the sound signal. The voice processing device 100a includes a sound input unit 101, a position pattern detection unit 102a, a processing determination unit 103a, a signal processing unit 104, and a pattern DB 109a.

  The sound input unit 101 detects the start and end of the sound from the input sound. The position pattern detection unit 102a detects a position pattern index between the sound source and the microphone based on a signal from the position sensor. The process determining unit 103a determines a process to be executed by collating a position pattern index with a previously held position pattern. The signal processing unit 104 performs processing according to the determination of the processing determination unit.

  The pattern DB 109a holds a position pattern between the sound source and the microphone. In the pattern DB 109a, an index of a signal input from the position sensor is associated with each position pattern of a relative position between the sound source and the microphone. The position pattern stored in the pattern DB 109a is read from the position pattern detection unit 102a and collated with the input from the position sensor.

  FIG. 9 is a flowchart showing the operation of the speech processing apparatus according to the second embodiment. A description will be given using an example of processing target speech using two microphones (microphone 1 and microphone 2). Note that it is not essential that there are two microphones, and this is possible if there are two or more microphones. It is not an essential element that the target sound is a voice. The operation of this embodiment is the same as that of the first embodiment except for the operations of the position sensor, the position pattern detection unit 102a, and the processing determination unit 103a, and the description of the same operation part as that of the first embodiment is omitted. .

In step S203, based on the output from the distance sensor attached near each microphone, the measurement result of that sensor is used as a position pattern determination index. Specifically, the distance sensor is an infrared sensor or the like that can measure the distance from each microphone to the target object that hits the sound source, and the distance from each microphone to the sound source is measured. Two microphones are used, and the distances from the microphone 1 and the microphone 2 to the sound source are ld ₁ and ld ₂ , respectively.

In step S204, the process determination unit 103a uses the position pattern determination index calculated by the position pattern detection unit 102 to classify which of the three position patterns. Three patterns are shown below.
(1A) A sound source is approaching the microphone 1.
(2A) A sound source is approaching the microphone 2.
(3A) No sound source is approaching either microphone.

At this time, if the distance threshold _{ld_thre} is a constant, it is classified as (1A) when the following expression (26) is satisfied, and is classified as (2A) when the following expression (27) is satisfied.

If none of the above, the position pattern is classified into (3A). Since the processing of step S205 and step S206 after the position pattern classification is the same as step S105 and step S106 of FIG. 2, description thereof is omitted here.

(Third embodiment)
FIG. 10 is a block diagram showing a speech processing apparatus according to the third embodiment. The sound processing apparatus 100b according to the third embodiment detects a position pattern of a sound source based on an input from a position sensor and a sound signal, and executes sound processing corresponding to each position pattern.

The voice processing device 100b includes a sound input unit 101, a position pattern detection unit 102b, a processing determination unit 103b, a signal processing unit 104, and a pattern DB 109b.
The sound input unit 101 converts the input sound from the microphone into a digitized sound signal and detects the start and end of the sound. The position pattern detection unit 102b detects an index of the position pattern of the sound source and the microphone from the input from the position sensor and sound. The process determining unit 103b determines a process to be executed by collating a position pattern index with a previously held position pattern. The signal processing unit 104 performs processing according to the determination of the processing determination unit.

  The pattern DB 109b holds a position pattern between the microphone and the sound source. In the pattern DB 109b, a combination of a signal index and a sound signal index input from the position sensor is associated with each position pattern of the relative positions of the microphone and the sound source. The pattern stored in the pattern DB 109b is called from the position pattern detection unit 102b and collated with the sound signal acquired by the sound input unit 101 and the input from the position sensor.

  FIG. 11 is a flowchart showing the operation of the speech processing apparatus according to the third embodiment. Here, a description will be given using an example in which target microphones are processed using two microphones (microphone 1 and microphone 2). Note that it is not essential that there are two microphones, and there may be two or more microphones. It is not an essential element that the target sound is a voice. The operation of this embodiment is the same as that of the second embodiment except that the operations of the position pattern detection unit 102b and the processing determination unit 103b are different from those of the second embodiment. , Omit the explanation.

  The voice processing device 100b may use a distance sensor as a position sensor, for example. In step S303, the position pattern detection unit 102b acquires a measurement result by the distance sensor and audio information as a position pattern determination index.

  More specifically, an infrared sensor or the like is used as a position sensor, and the distance from the device to the sound source is measured. Also, two microphones that acquire sound signals are used, and the distance from the sound processing device 100b acquired using the sensor to the sound source is ld. Similarly to the first embodiment, the voice arrival time difference td and the signal level ratio dd are also obtained.

In step S304, the process determination unit 103b uses the position pattern determination index calculated by the position pattern detection unit to classify which of the three position patterns it belongs to. Three patterns are shown below.
(1B) A sound source is approaching the microphone 1.
(2B) A sound source is approaching the microphone 2.
(3B) The sound source is not approaching either microphone.

The arrival time difference determination threshold value t _thre , the signal level difference determination threshold _values dd _thre1 , dd _thre2 , and the distance determination threshold value ld _thre are constants (where t _thre > 0, dd _thre1 > dd _thre2 > 0, ld _thre > 0). Here, when all of the following expressions (28) hold when td> 0, the position pattern is classified as (1B).

また、（１Ｂ）、（２Ｂ）の何れでもない位置パターンを（３）に分類する。この３つの位置パターン毎に、第１の実施形態の（１），（２），（３）と同様の処理を行う。 Further, when td <= 0, when the following expression (29) holds, the position pattern is classified as (2B).

Further, a position pattern that is neither (1B) nor (2B) is classified as (3). Processing similar to (1), (2), and (3) of the first embodiment is performed for each of these three position patterns.

  Also, the output from the angle sensor can be used as a position pattern determination index. FIG. 12 is a diagram illustrating an example in which an angle sensor is provided in a mobile phone. In the example of FIG. 12, the mobile phone is used in the horizontal direction during operation and in the vertical direction during a call. In such a device, an angle is detected using an angle sensor attached to the device body. An example of the detected angle θ is shown in FIG. For the angle θ, for example, the position where the line segment connecting the two microphones and the ground are horizontal is 0 degrees. Similarly to the first embodiment, the voice arrival time difference td and the signal level ratio dd are also obtained.

In the example of FIG. 12, the position pattern classification into (1B), (2B), and (3B) includes arrival time difference determination threshold t _thre , signal level difference determination thresholds dd _thre1 , dd _thre2 , and angle determination threshold θ _thre . If constants (where t _thre > 0, dd _thre1 > dd _thre2 > 0, θ _thre ≧ 0), respectively, the following equations (30) and (31) are performed.

When td> 0, when the following equation (30) holds, the position pattern is classified as (1B).

位置パターンが（１Ｂ），（２Ｂ）の何れでもない場合には、（３Ｂ）に分類される。 When td <= 0, when the following equation (31) holds, the position pattern is classified as (2B).

If the position pattern is neither (1B) nor (2B), it is classified as (3B).

(Minimum configuration realized by computer etc.)
Next, the hardware configuration of the speech processing apparatus according to this embodiment will be described with reference to FIG. FIG. 13 is an explanatory diagram showing a hardware configuration of the speech processing apparatus according to the present embodiment.

  The audio processing apparatus according to the present embodiment is connected to a control device such as a CPU (Central Processing Unit) 51, a storage device such as a ROM (Read Only Memory) 52 and a RAM (Random Access Memory) 53, and a network. A communication I / F 54 that performs communication and a bus 61 that connects each unit are provided.

  A program executed by the speech processing apparatus according to the present embodiment is provided by being incorporated in advance in the ROM 52 or the like.

  A program executed by the audio processing apparatus according to the present embodiment is a file in an installable format or an executable format, and is a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD), a CD-R (Compact). You may comprise so that it may record and provide on computer-readable recording media, such as a Disk Recordable (DVD) and DVD (Digital Versatile Disk).

  Furthermore, the program executed by the audio processing apparatus according to the present embodiment may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. The program executed by the speech processing apparatus according to the present embodiment may be provided or distributed via a network such as the Internet.

  The program executed by the speech processing apparatus according to the present embodiment has a module configuration including the above-described units. As actual hardware, the CPU 51 reads the program from the ROM 52 and executes the program so that each unit is It is loaded on the main storage device, and each unit is generated on the main storage device.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

  As described above, the sound processing apparatus according to the embodiment of the present invention is useful for noise removal, and is particularly suitable for processing a sound signal input from a microphone array.

1, 2, 3 Microphone 100, 100a, 100b Sound processing device 101 Sound input unit 102, 102a, 102b Position pattern detection unit 103, 103a, 103b Processing determination unit 104 Signal processing unit

Claims (8)

A position pattern detection unit for detecting an index of relative position between the sound source and the plurality of microphones;
A process determining unit that determines a sound process for a sound signal input from each of the plurality of microphones based on an index of the relative position;
A signal processing unit that executes the determined audio processing on the sound signal;
A speech processing apparatus comprising:   The speech processing apparatus according to claim 1, wherein the relative position index includes a difference in arrival time of sound signals input from the plurality of microphones and a difference in level of the sound signals.   The speech processing apparatus according to claim 1, wherein the relative position index includes a distance measured by a distance sensor provided at a predetermined position with respect to each of the plurality of microphones.   The speech processing apparatus according to claim 1, wherein the relative position index includes an inclination of the microphone measured by an angle sensor provided at a predetermined position with respect to each of the plurality of microphones.   The process determining unit is configured to add a greater weight to a sound signal input from a microphone whose distance from the sound source is smaller than a predetermined value than a sound signal input from a microphone whose distance from the sound source is a predetermined value or more. The speech processing apparatus according to claim 1, wherein the speech processing apparatus is determined to perform the process.   The processing determining unit determines to perform audio processing that takes a delay sum for a sound signal input from the plurality of microphones when a distance from the sound source of the plurality of microphones is equal to or greater than a predetermined value. The speech processing apparatus according to any one of claims 1 to 4, wherein the speech processing apparatus is characterized. Computer
A position pattern detection unit for detecting an index of relative position between the sound source and the plurality of microphones;
A process determining unit that determines a sound process for a sound signal input from each of the plurality of microphones based on an index of the relative position;
A signal processing unit that executes the determined audio processing on the sound signal;
Program to function as. A position pattern detection unit that detects a relative position index between the sound source and the plurality of microphones;
A process determining step in which a process determining unit determines a sound process for a sound signal input from each of the plurality of microphones based on an index of the relative position;
A signal processing step in which the signal processing unit executes the determined sound processing on the sound signal;
A voice processing method characterized by comprising:

Images