Acoustic signal selection device and program

Abstract

Even when the relationship between the acoustic sensor and the external environment is unknown, it suppresses the intrusion of noise based on the external environment. [Solution] When the periodicity of a reference acoustic signal corresponding to at least one of the multiple acoustic signals obtained from multiple acoustic sensors exceeds a reference value, information corresponding to one or more first acoustic signals extracted from the multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals is output. When the periodicity of the reference acoustic signal does not exceed the reference value, information corresponding to one or more second acoustic signals extracted from the multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals is output.

Description

[Technical Field]

[0001] This invention relates to a technique for selecting a specific acoustic signal from multiple acoustic signals based on observation signals obtained from multiple acoustic sensors. [Background technology]

[0002] Acoustic signal processing techniques are known that enhance desired sounds or suppress unwanted noise based on observation signals obtained from multiple acoustic sensors (e.g., microphones) (see, for example, Non-Patent Documents 1, 2, 3, etc.). For example, beamforming is an acoustic signal processing technique that enhances or suppresses sound components emitted from a sound source in a specific direction using a linear filter. Such processing is called linear processing. Other acoustic signal processing techniques include the Wiener filter method and the spectral subtraction (SS) method, which are based on short-time spectral amplitude (STSA) analysis. These methods use nonlinear filters. Such processing is called nonlinear processing. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Osamu Hojuyama, "2-1 Beamforming", [online], November 2011, Institute of Electronics, Information and Communication Engineers, Group 2, Article 6, Acoustic Signal Processing, Chapter 2, Sound Source Separation, [Retrieved April 23, 2024], Internet <https: / / www.ieice-hbkb.org / files / ad_base / view_pdf.html?p= / files / 02 / 02gun_06hen_02.pdf#page=2> [Non-Patent Document 2] Akihiko Sugiyama, "4-1 Noise Suppression," [online], May 2014, Institute of Electronics, Information and Communication Engineers, Group 2, Article 6, Acoustic Signal Processing, Chapter 4 Noise and Reverberation Suppression, [Retrieved April 23, 2024], Internet <https: / / www.ieice-hbkb.org / files / ad_base / view_pdf.html?p= / files / 02 / 02gun_06hen_04.pdf#page=2> [Non-Patent Document 3] Takashi Oba, Kazunori Kobayashi, Takashi Uematsu, Taichi Asami, Kenta Niwa, Kira Kamadoshi, Tomoko Kawase, Takaaki Hori, "Media Processing Technology to Realize Support in Business Scenes," [online], February 2015, NTT Technical Journal, Media Technologies to Create Attractive User Experiences, pp. 26-30, [Retrieved June 5, 2024], Internet<https: / / journal.ntt.co.jp / backnumber2 / 1502 / files / jn201502026.pdf> [Overview of the initiative] [Problems that the invention aims to solve]

[0004] The amount of noise mixed in the observed signal may vary depending on the relationship between the acoustic sensor and the external environment. For example, if wind is directly hitting only a specific acoustic sensor, or if an object such as a body part or clothing is in contact with it, the observed signal obtained from that specific acoustic sensor may contain more noise than the observed signals obtained from other acoustic sensors.

[0005] However, conventionally, it has been difficult to predict which acoustic sensors will contain a lot of noise in their observed signals without knowing the relationship between the acoustic sensors and the external environment. Therefore, it has also been difficult to suppress noise contamination based on the external environment when the relationship between the acoustic sensors and the external environment is unknown.

[0006] This invention provides a technology that can suppress the intrusion of noise based on the external environment, even when the relationship between the acoustic sensor and the external environment is unknown. [Means for solving the problem]

[0007] The acoustic signal selection device outputs information corresponding to one or more first acoustic signals extracted from the multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals obtained from multiple acoustic sensors, when the height of the periodicity of a reference acoustic signal corresponding to at least one of the multiple acoustic signals exceeds a reference value, and outputs information corresponding to one or more second acoustic signals extracted from the multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals, when the height of the periodicity of the reference acoustic signal does not exceed the reference value. [Effects of the Invention]

[0008] This makes it possible to suppress the intrusion of noise based on the external environment, even when the relationship between the acoustic sensor and the external environment is unknown. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 is a block diagram illustrating the functional configuration of the acoustic signal processing system according to the first embodiment. [Figure 2] Figure 2 is a block diagram illustrating the functional configuration of the acoustic signal selection device according to the first embodiment. [Figure 3] Figure 3 is a block diagram illustrating the functional configuration of the selection control unit in the first embodiment. [Figure 4] Figure 4 is a block diagram illustrating the functional configuration of the acoustic signal processing system according to the second embodiment. [Figure 5] Figure 5 is a block diagram illustrating the functional configuration of the acoustic signal selection device according to the second embodiment. [Figure 6] Figure 6 is a block diagram illustrating the functional configuration of the selection control unit of the second embodiment. [Figure 7] Figure 7 is a block diagram illustrating the functional configuration of the acoustic signal processing device according to the second embodiment. [Figure 8] Figures 8A and 8B are conceptual diagrams illustrating the arrangement of acoustic sensors in the embodiment. [Figure 9] Figure 9 is a block diagram illustrating the hardware configuration of this embodiment. [Modes for carrying out the invention]

[0010] Embodiments of the present invention will be described below with reference to the drawings. [First Embodiment] In the first embodiment, if the periodicity of a reference acoustic signal corresponding to at least one of the multiple acoustic signals obtained from multiple acoustic sensors exceeds a reference value (reference), information corresponding to one first acoustic signal extracted from the multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals is output. If the periodicity of the reference acoustic signal does not exceed the reference value, information corresponding to one second acoustic signal extracted from the multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals is output.

[0011] <Structure> As illustrated in Figure 1, the acoustic signal processing system 1 of this embodiment includes N (multiple) acoustic sensors 10-1, ..., 10-N, an AD converter 11, an acoustic signal selection device 12, and an acoustic signal processing device 13. Here, each of the acoustic sensors 10-1, ..., 10-N is referred to as acoustic sensor 10-n, and acoustic sensors 10-1, ..., 10-N are collectively referred to as acoustic sensor 10 (the same applies to other symbols). n is an integer index corresponding to acoustic sensor 10-n, where n = 1, ..., N, and N is an integer greater than or equal to 2.

[0012] <Acoustic sensor 10> The acoustic sensor 10 in this embodiment is a device that detects (observes) sound waves or acoustic energy, converts them into an observed signal (e.g., an analog signal), and outputs it. The observed signal obtained by acoustic sensor 10-n is denoted as α(n). Furthermore, observed signals α(1), ..., α(N) are collectively referred to as observed signal α (the same applies to other symbols). The observed signal α is a time-series signal. Specific examples of acoustic sensor 10 include microphones, ultrasonic sensors, hydrophones, etc. Specific examples of sound waves or acoustic energy include sound, ultrasound, underwater sound waves, etc. The observed signal α obtained by acoustic sensor 10 may, for example, contain only highly periodic components, or only low-periodic components, or a mixture of highly periodic and low-periodic components. The ratio of highly periodic to low-periodic components in the observed signal α may change depending on the external environment of acoustic sensor 10. An index representing the degree of periodicity may be a value that does not monotonically decrease with respect to the degree of periodicity (for example, a value that monotonically increases), or a value that does not monotonically increase with respect to the degree of periodicity (for example, a value that monotonically decreases). For example, in the case of the former index, components with high periodicity are those in which the index representing the degree of periodicity exceeds a periodicity reference value (periodicity reference value), and components with low periodicity are those in which the index representing the degree of periodicity does not exceed the said periodicity reference value. Examples of the former index include pitch correlation value and autocorrelation function value. For example, in the case of the latter index, components with high periodicity are those in which the index representing the degree of periodicity does not exceed a periodicity reference value (periodicity reference value), and components with low periodicity are those in which the index representing the degree of periodicity exceeds the said periodicity reference value. Examples of the latter index include the reciprocal of the pitch correlation value and the reciprocal of the autocorrelation function value. Components with high periodicity are, for example, components of the target signal such as speech or music. Components with low periodicity are, for example, noise components such as white noise (non-periodic noise). The noise includes components corresponding to the external environment of the acoustic sensor 10. In other words, the components of the noise observed by the acoustic sensor 10 may change depending on the external environment of the acoustic sensor 10.For example, noise may be observed when wind directly hits the acoustic sensor 10, when an object such as a body part or clothing comes into contact with the acoustic sensor 10, when the acoustic sensor 10 is facing a specific direction (for example, the direction of the noise source), or when a device operates in the vicinity of the acoustic sensor 10.

[0013] The number, position, and / or orientation of the acoustic sensors 10 can be arbitrarily set according to the application. However, it is desirable that at least one of the N acoustic sensors 10-1, ..., 10-N is positioned and / or oriented differently from the other acoustic sensors 10. All acoustic sensors 10 may be positioned and / or oriented differently from the other acoustic sensors 10, or some acoustic sensors 10 may be positioned and / or oriented differently from the other acoustic sensors 10. If the acoustic characteristics (e.g., directivity) of at least one of the N acoustic sensors 10-1, ..., 10-N are different from those of the other acoustic sensors 10, then all acoustic sensors 10 may be positioned and / or oriented in the same position and / or orientation as the other acoustic sensors 10. In other words, the relative relationships of the noises contained in the observed signals of the N acoustic sensors 10-1, ..., 10-N should change in response to some change in the external environment of the N acoustic sensors 10-1, ..., 10-N. The acoustic sensors 10-1, ..., 10-N may or may not be fixed to the housing. The relative positions and / or attitudes and / or orientations of the acoustic sensors 10-1, ..., 10-N may or may not be fixed.

[0014] Figures 8A and 8B illustrate how the acoustic sensor 10 is fixed to the housing. In the example of Figure 8A, the acoustic sensors 10-1, 10-2, and 10-3 are arranged in the same direction (forward) at different positions on the front of the housing 1110. In the example of Figure 8B, the acoustic sensor 10-1 is arranged upward on the upper surface of the housing 1210, and the acoustic sensor 10-2 is arranged downward on the lower surface of the housing 1210. However, these are just examples and do not limit the present invention.

[0015] <AD conversion device 11> The AD conversion device 11 of this embodiment is a device that obtains and outputs an acoustic signal (for example, a digital signal) based on the input observation signal α. The acoustic signal based on the observation signal α(n) is denoted as x(n) (n = 1,..., N). The acoustic signals x(1),..., x(N) are collectively denoted as the acoustic signal x (the same applies to other symbols). The acoustic signal x is a time-series signal. The acoustic signal x of this embodiment is a signal obtained by digitizing the observation signal α at the sampling frequency and dividing it at regular time lengths (regular time intervals) called frames. The sampling frequency is, for example, 8 kHz, 16 kHz, 32 kHz, 48 kHz, etc. The time length of the frame is, for example, 8 milliseconds, 10 milliseconds, 20 milliseconds, etc. Additionally, for example, the acoustic signal x may be a signal obtained by digitizing the observation signal α at the sampling frequency, dividing it for each frame, and further converting it to the time-frequency domain.

[0016] <Acoustic signal selection device 12> The acoustic signal selection device 12 of this embodiment outputs information corresponding to the acoustic signal x(s) (first acoustic signal) extracted from the plurality of acoustic signals x(1),..., x(N) based on the periodicity of at least one of the plurality of acoustic signals x(1),..., x(N) when the height of the periodicity of the reference acoustic signal x(n r ) exceeds the reference value (reference) TH1. On the other hand, the acoustic signal selection device 12 is the reference acoustic signal x(n rWhen the periodicity of the reference acoustic signal x(n) does not exceed the reference value TH1, information corresponding to the acoustic signal x(s) (second acoustic signal) extracted from the multiple acoustic signals x(1), ..., x(N) is output based on at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N). In this embodiment, the reference acoustic signal x(n) r This shows an example where ) is an acoustic signal included in the acoustic signals x(1), ..., x(N).

[0017] As illustrated in Figure 2, the acoustic signal selection device 12 of this embodiment includes buffer units 121-1, ..., 121-N, periodicity calculation units 122-1, ..., 122-N, magnitude calculation units 123-1, ..., 123-N, periodicity comparison unit 124, magnitude comparison unit 125, delay units 126-1, ..., 126-N, selection control unit 127, and selection unit 128. As illustrated in Figure 3, the selection control unit 127 includes a periodicity determination unit 127a and a switching control unit 127b.

[0018] <Acoustic signal processing device 13> The acoustic signal processing device 13 (Figure 1) performs acoustic signal processing on the input acoustic signal x(s) and outputs the acoustic signal processing result y(s). The acoustic signal processing can be of any type. The acoustic signal processing can be linear or nonlinear. For example, this acoustic signal processing may be a process that emphasizes a desired sound or suppresses unwanted noise using a Wiener filter or spectral subtraction method based on short-time spectral amplitude analysis. In addition, for example, this acoustic signal processing may be speech recognition processing, signal separation processing, or other processing.

[0019] <Processing> The acoustic sensors 10-1, ..., 10-N each detect sound waves or acoustic energy and output the observed signals α(1), ..., α(N) obtained thereby. The observed signals α(1), ..., α(N) are sent to the AD converter 11 (step S10).

[0020] As described above, the AD converter 11 obtains and outputs acoustic signals x(1), ..., x(N) based on the input observation signals α(1), ..., α(N). The acoustic signals x(1), ..., x(N) are sent to the acoustic signal selection device 12 (step S11).

[0021] The acoustic signal selection device 12 (Figure 2) receives the acoustic signals x(1), ..., x(N) of each frame as input. The acoustic signal selection device 12 selects a reference acoustic signal x(n) that is included in the multiple acoustic signals x(1), ..., x(N). r When the periodicity of the reference acoustic signal (corresponding to at least one of the multiple acoustic signals x(1), ..., x(N)) exceeds the reference value TH1, information corresponding to the acoustic signal x(s) (first acoustic signal) extracted from the multiple acoustic signals x(1), ..., x(N) is output based on the periodicity of at least one of the multiple acoustic signals x(1), ..., x(N). On the other hand, the acoustic signal selection device 12 outputs information corresponding to the reference acoustic signal x(n r If the periodicity of the above does not exceed the reference value TH1, information corresponding to the acoustic signal x(s) (second acoustic signal) extracted from the multiple acoustic signals x(1), ..., x(N) is output based on at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N) (step S12). This will be explained in detail below.

[0022] <Detailed example of step S12> The acoustic signal x(n) (n=1,…,N) for each frame is input to the buffer units 121-n and delay units 126-n of the acoustic signal selection device 12 (Figure 2). Each buffer unit 121-n buffers (temporarily stores) the acoustic signals x(n) for M frames. Here, M is an integer of 1 or more that represents the number of frames of acoustic signals x required for processing in the periodicity calculation unit 122-n and the magnitude calculation unit 123-n. Increasing M improves processing performance, but conversely, it increases processing delay and reduces real-time performance. An example of M is a value such that the length of M frames is between 30 milliseconds and 40 milliseconds. The buffered acoustic signals x(n) for M frames are sent one by one to the periodicity calculation unit 122-n and the magnitude calculation unit 123-n. The acoustic signals x(n) for M frames sent previously and the acoustic signals x(n) for M frames sent this time may or may not overlap with each other in some respects. The M-frame acoustic signal x(n) may be an M-frame acoustic signal x(n) that is continuous in time, or an M-frame acoustic signal x(n) that is discrete in time. The M-frame acoustic signal x(n) may or may not include the latest acoustic signal x(n) stored in the buffer unit 121-n. For example, each buffer unit 121-n is buffered with the latest M-frame acoustic signal x(n) that is continuous in time (step S121).

[0023] The M frames of acoustic signals x(n) buffered in buffer unit 121-n are input one by one to each periodicity calculation unit 122-n (n=1,…,N). Each periodicity calculation unit 122-n uses the M frames of acoustic signals x(n) to calculate and output an index r(n) that represents the degree of periodicity of the acoustic signals x(n). For example, each periodicity calculation unit 122-n applies pitch analysis to the M frames of acoustic signals x(n) to obtain a pitch correlation value, and outputs this pitch correlation value as an index r(n) that represents the degree of periodicity. Pitch is the fundamental frequency component contained in speech, and is an acoustic feature quantity generated by the vibration of the vocal cords. Pitch analysis is a method of analyzing an acoustic signal to estimate the pitch frequency, periodic features, etc. The pitch correlation value is a value that represents the strength of periodicity obtained by calculating the autocorrelation function of the acoustic signal. Generally, the pitch correlation value is often expressed as a normalized value between 0.0 and 1.0. In sections containing only voiceless consonants or background noise (non-speech sections), the pitch correlation value will be close to 0, while in sections with voiced sounds (speech sections), it will be close to 1. For pitch analysis, methods such as those described in references 1 and 2 below can be used. Reference 1: Japanese Patent Application Publication No. 10-232697 Reference 2: ITU-T G.729, “Coding of speech at 8 kbit / s using conjugate-structure algebraic-code-excited linear prediction (CS-ACELP)”, 3.4 Open-loop pitch analysis The index r(n) representing the periodicity of the acoustic signal x(n) for M frames may be calculated for each frame or for multiple frames. The index r(n) representing the periodicity of each frame is sent to the periodicity comparison unit 124 (step S122).

[0024] The periodicity comparison unit 124 receives indices r(1), ..., r(N) representing the height of periodicity sent from the periodicity calculation units 122-1, ..., N. In this embodiment, the periodicity comparison unit 124 selects the index r(n) representing the highest periodicity from among the indices r(1), ..., r(N) representing the height of periodicity. rSelect it. The index r(n r ) of this embodiment represents the degree of periodicity of the acoustic signal x(n r )(reference acoustic signal). For example, when the index r representing the degree of periodicity is a value that is monotonically non-decreasing with respect to the degree of periodicity (e.g., a monotonically increasing value, pitch correlation value, etc.), the periodicity comparison unit 124 selects the maximum index r(n r )(e.g., the maximum pitch correlation value, etc.) among the indices r(1),..., r(N) representing the degree of periodicity. On the other hand, for example, when the index r representing the degree of periodicity is a value that is monotonically non-increasing with respect to the degree of periodicity (e.g., a monotonically decreasing value), the periodicity comparison unit 124 selects the minimum index r(n r ) among the indices r(1),..., r(N) representing the degree of periodicity. The index r(n r ) may be selected for each frame or for a plurality of frames. The periodicity comparison unit 124 sends the selected index r(n r ) representing the degree of periodicity and the corresponding integer index n r ∈{1,..., N} to the selection control unit 127 (step S124).

[0025] ​​The M frames of acoustic signals x(n) buffered in buffer unit 121-n are input one by one to each magnitude calculation unit 123-n (n=1,…,N). Each magnitude calculation unit 123-n calculates and outputs an index p(n) representing the magnitude of the M frames of acoustic signals x(n). The index p(n) representing the magnitude of the M frames of acoustic signals x(n) may be the average value, maximum value, intermediate value, minimum value, representative value, or statistical value of the index representing the magnitude of acoustic signals x(n) in that M frame. The index representing the magnitude of each acoustic signal x(n) may be a value that does not decrease monotonically with respect to the magnitude of the acoustic signal x(n) (for example, a value that increases monotonically) or a value that does not increase monotonically with respect to the magnitude of the acoustic signal x(n) (for example, a value that decreases monotonically). Examples of the former include the power, amplitude, monotonically decreasing function value of power (for example, a monotonically increasing function value), and monotonically decreasing function value of amplitude of the acoustic signal x(n). Examples of the latter include the monotonically non-increasing function value of power (e.g., the monotonically decreasing function value), the monotonically non-increasing function value of amplitude, etc. The index p(n) may be calculated for each frame or for multiple frames. The index p(n) representing the magnitude of each acoustic signal x(n) is sent to the magnitude comparison unit 125 (step S123).

[0026] The size comparison unit 125 receives indexes p(1), ..., p(N) representing size, sent from the size calculation units 123-1, ..., 123-N. In this embodiment, the size comparison unit 125 uses the index p(n) representing the smallest size among the indexes p(1), ..., p(N). p ) is selected. For example, if the index representing the size is a value that does not decrease monotonically with respect to the size (for example, a value that increases monotonically), the size comparison unit 125 selects the smallest index p(n) among the indexes representing the size p(1), ..., p(N). p ) is selected. On the other hand, for example, if the index representing the size is a value that does not increase monotonically with respect to the size (for example, a value that decreases monotonically), the size comparison unit 125 selects the largest index p(n) among the indexes representing the size p(1), ..., p(N). p Select the index p(n). pThe size may be selected for each frame or for multiple frames. The size comparison unit 125 uses an index p(n) representing the selected size. p ) corresponding integer index n p Send ∈{1,…,N} to the selection control unit 127 (step S125).

[0027] The selection control unit 127 has an index r(n) that represents the height of the selected periodicity. r ), and the corresponding integer index n r ∈{1,…,N}, and an index p(n) representing the selected size. p ) corresponding integer index n p The input is ∈{1,…,N}. The selection control unit 127 determines the index r(n r If the periodicity represented by ) exceeds the reference value TH1 (when the periodicity of the reference acoustic signal exceeds the reference value), the index r(n r ) corresponding integer index n r The integer index s (information corresponding to the first acoustic signal extracted from multiple acoustic signals based on at least one of the periodicities of the multiple acoustic signals) is output. Meanwhile, the selection control unit 127 outputs the index r(n r If the periodicity represented by ) does not exceed the reference value TH1 (if the periodicity of the reference acoustic signal does not exceed the reference value), then index p(n p ) corresponding integer index n p Outputs the integer index s (outputs information corresponding to the second acoustic signal extracted from multiple acoustic signals based on at least one of the magnitudes of the multiple acoustic signals). Integer index s ∈ {n r ,n p The} may be output per frame or per multiple frames. The integer index s is sent to the selection unit 128 (step S127).

[0028] <Detailed example of step S127> For example, as illustrated in Figure 3, the index r(n rThe index r(n) is input to the periodicity determination unit 127a of the selection control unit 127. The periodicity determination unit 127a determines the index r(n) r It is determined whether the degree of periodicity represented by ) exceeds the reference value TH1. For example, the index r(n r If the index r(n) is a value that does not monotonically decrease with respect to the degree of periodicity (for example, a monotonically increasing value, a pitch correlation value, etc.), the periodicity determination unit 127a determines that the index r(n) r ) determines whether it exceeds the threshold corresponding to the reference value TH1 (for example, reference value TH1). On the other hand, for example, index r(n r If the index r(n) is a value that does not monotonically increase with respect to the degree of periodicity (for example, a value that monotonically decreases), the periodicity determination unit 127a determines that the index r(n) r The periodicity determination unit 127a determines whether the index r(n r The high degree of periodicity represented by ) exceeds the standard value TH1, which is the result of the judgment d r , or index r(n r The judgment result d indicates that the degree of periodicity represented by ) does not exceed the standard value TH1. p The following will be output. Judgment result d r or judgment result d p This is sent to the switching control unit 127b (step S127a).

[0029] The switching control unit 127b has an index r(n r ) corresponding integer index n r , index p(n p ) corresponding integer index n p , and judgment result d r or judgment result d p The input is d. The switching control unit 127b determines the result d r If this is input (when the periodicity of the reference acoustic signal exceeds the reference value), integer index n rThe integer index s is output (information corresponding to the first acoustic signal extracted from multiple acoustic signals based on at least one of the periodicities of the multiple acoustic signals is output). Meanwhile, the switching control unit 127b determines the result d p If this is input (i.e., the periodicity of the reference acoustic signal does not exceed the reference value), integer index n p Outputs the integer index s (outputs information corresponding to the second acoustic signal extracted from multiple acoustic signals based on at least one of the magnitudes of the multiple acoustic signals). Integer index s ∈ {n r ,n p The result is sent to the selection unit 128 (step S127b).

[0030] The delay units 126-1, ..., 126-N of the acoustic signal selection device 12 (Figure 2) receive the acoustic signals x(1), ..., x(N) of each frame as input. Each delay unit 126-n (n=1, ..., N) delays the input acoustic signal x(n) by a delay time τ and sends the delayed acoustic signal x(n) to the selection unit 128. The delay time τ is determined based on the time during which the acoustic signal x(n) is temporarily stored in the buffer units 121-n for processing in steps S122 to S125 and S127. For example, the delay time τ is determined so that the frame corresponding to the acoustic signal x(n) sent to the selection unit 128 matches or approximates the frame corresponding to the integer index s sent to the selection unit 128. For example, the delay time τ is set to the time from when an acoustic signal x(n) of a certain frame is stored in buffer 121-n in step S121 until the integer index s obtained in steps S122-S125, S127 using acoustic signals x(n) for M frames including this frame is sent to the selection unit 128, or an approximate time. The delay time τ may be a time set individually for each delay unit 126-n, or it may be a time common to all delay units 126-1, ..., 126-N (step S126).

[0031] The selection unit 128 receives an integer index s (step S127) sent from the selection control unit 127, and delayed acoustic signals x(1), ..., x(N) (step S126) sent from the delay units 126-1, ..., 126-N. The selection unit 128 extracts and outputs the acoustic signal x(s) corresponding to the input integer index s from the input acoustic signals x(1), ..., x(N). In other words, the selection unit 128 switches the output acoustic signal x(s) based on the integer index s. The acoustic signal x(s) may be switched frame by frame, or every multiple frames. The acoustic signal x(s) is sent to the acoustic signal processing device 13 (Figure 1) (step S128).

[0032] The acoustic signal processing device 13 (Figure 1) receives an acoustic signal x(s) as input. The acoustic signal processing device 13 performs acoustic signal processing on the acoustic signal x(s) and outputs the resulting acoustic signal processing result y(s). For example, the acoustic signal processing device 13 performs processing on the acoustic signal x(s) using a Wiener filter or spectral subtraction method based on time spectral amplitude analysis to emphasize desired sounds or suppress unwanted noise, and outputs the resulting acoustic signal processing result y(s) (step S13).

[0033] The processes in steps S10 and S11 described above are performed on the time-series signals of the observed signals α(1), ..., α(N), and the processes in steps S12 and S13 are repeatedly performed for each frame. As a result, the acoustic signal processing device 13 outputs time-series information of the acoustic signal processing result y(s).

[0034] <Features of this form> The highly periodic components in an acoustic signal x are, for example, components of the target signal, such as speech or music. Therefore, time intervals with high periodicity in the acoustic signal x(n) can be presumed to be target signal intervals (e.g., speech intervals) where the target signal component is present or dominant. On the other hand, time intervals with low periodicity in the acoustic signal x(n) can be presumed to be non-target signal intervals (e.g., non-speech intervals) where the target signal component is absent or not dominant.

[0035] Typically, the components of the target signal are observed by all acoustic sensors 10-1, ..., 10-N. Therefore, a time interval in which the acoustic signal x(n) based on the observed signal α(n) obtained by any acoustic sensor 10-n has high periodicity is the target signal interval for all acoustic signals x(1), ..., x(N). Similarly, a time interval in which the acoustic signal x(n) based on the observed signal α(n) obtained by any acoustic sensor 10-n has low periodicity is a non-target signal interval for all acoustic signals x(1), ..., x(N).

[0036] In contrast, noise components based on the external environment are not necessarily observed simultaneously in all acoustic sensors 10-1, ..., 10-N, and the observed noise components often differ for each acoustic sensor 10-n. For example, if wind directly hits a particular acoustic sensor 10 or an object comes into contact with it, a large noise component may be observed in that particular acoustic sensor 10, but no large noise component may be observed in the other acoustic sensors 10.

[0037] Utilizing these characteristics, the acoustic signal selection device 12 in this embodiment selects the acoustic signal x(n) with the maximum periodicity. r Let ) (reference acoustic signal) be the acoustic signal x(n r If the periodicity of the acoustic signal x(n) exceeds the reference value TH1, r Let ) be the acoustic signal x(s) (first acoustic signal), and the acoustic signal x(n r When the periodicity of ) does not exceed the reference value TH1, the smallest acoustic signal x(n p This was defined as the acoustic signal x(s) (second acoustic signal). This allows switching to an acoustic signal x(s) with less noise based on the external environment on a frame-by-frame or multi-frame basis. As a result, even when the relationship between the acoustic sensors 10-1, ..., 10-N and the external environment is unknown, the inclusion of noise based on the external environment into the acoustic signal x(s) can be suppressed.

[0038] That is, the acoustic signal x(n rA time interval in which the periodicity of the acoustic signal x(n) exceeds the reference value TH1 can be estimated to be the target signal interval. In the target signal interval, the acoustic signal x(n) may contain both the components of the target signal and the components of noise based on the external environment. Here, the higher the ratio of the noise component to the components of the target signal, the lower the periodicity of the acoustic signal x(n). In other words, the higher the periodicity of the acoustic signal x(n), the lower the ratio of the noise component to the components of the target signal. Therefore, in the target signal interval, the acoustic signal x(n) with the highest periodicity is estimated to be the target signal interval. r By outputting this as an acoustic signal x(s) (first acoustic signal), the inclusion of noise based on the external environment into the acoustic signal x(s) can be suppressed. This process does not require information about the relationship between the acoustic sensors 10-1, ..., 10-N and the external environment, such as their position, orientation, and posture, or external environment such as wind and objects.

[0039] On the other hand, the acoustic signal x(n r A time interval in which the periodicity of ) does not exceed the reference value TH1 can be estimated to be a non-target signal interval. In a non-target signal interval, the acoustic signal x(n) may contain noise components based on the external environment, but it rarely contains components of the target signal (at least, the components of the target signal are not dominant). Here, the more of the noise components contained in the acoustic signal x(n), the larger the magnitude of the acoustic signal x(n). In other words, the smaller the magnitude of the acoustic signal x(n), the less of the noise components contained in the acoustic signal x(n). Therefore, in a non-target signal interval, the acoustic signal x(n) with the smallest magnitude is... p By outputting this as an acoustic signal x(s) (second acoustic signal), the inclusion of noise based on the external environment into the acoustic signal x(s) can be suppressed. This process does not require information about the relationship between the acoustic sensors 10-1, ..., 10-N and the external environment, such as their position, orientation, and posture, or external environment such as wind and objects.

[0040] In this configuration, among the acoustic signals x(1), ..., x(N), the one with the greatest periodicity is selected as the acoustic signal x(n). r) (reference acoustic signal). However, other acoustic signals x(n) included in acoustic signals x(1), ..., x(N) are considered as acoustic signal x(n r )(reference acoustic signal) may also be used. That is, in this embodiment, the acoustic signal x(n r Whether the periodicity of the acoustic signal x(n) exceeds the reference value TH1 is used to estimate whether it is a target signal interval or a non-target signal interval. As described above, a time interval in which the periodicity of the acoustic signal x(n) based on the observed signal α(n) obtained from any of the acoustic sensors 10-n is high is a target signal interval for all acoustic signals x(1),...,x(N). Therefore, among the acoustic signals x(1),...,x(N), the one with the highest periodicity is selected as the acoustic signal x(n) r Even without specifying ), it is possible to estimate whether each time interval is a target signal interval or a non-target signal interval. However, if the acoustic signal x(n) has high periodicity, r The more you set it to ), the higher the accuracy of this estimation becomes. Therefore, among the acoustic signals x(1), ..., x(N), the acoustic signal x(n) with high periodicity is selected as the acoustic signal x(n r It is desirable to use (reference acoustic signal) as the acoustic signal x(n) with the maximum periodicity. r It is more desirable to use (reference acoustic signal). Furthermore, among the acoustic signals x(1), ..., x(N), any acoustic signal x(n) whose periodicity exceeds the average (the average of the periodicity of acoustic signals x(1), ..., x(N)) should be set to acoustic signal x(n r ) (reference acoustic signal) may be used. Alternatively, any acoustic signal contained in acoustic signals x(1), ..., x(N) may be used as acoustic signal x(n rInstead of using (reference acoustic signal), other signals corresponding to multiple acoustic signals included in acoustic signals x(1),...,x(N) may be used as the reference acoustic signal. For example, the reference acoustic signal may be a signal corresponding to multiple acoustic signals from acoustic signals x(1),...,x(N) whose periodicity exceeds the standard, or the reference acoustic signal may be a signal corresponding to multiple acoustic signals selected in order of decreasing periodicity. The signal corresponding to multiple acoustic signals may be, for example, a signal obtained by adding the multiple acoustic signals, a signal obtained by weighted addition, or a signal obtained by averaging. Alternatively, the periodicity based on the periodicity of the multiple acoustic signals included in acoustic signals x(1),...,x(N) may be used as the periodicity of the reference acoustic signal. For example, a statistical value of the periodicity of the multiple acoustic signals included in acoustic signals x(1),...,x(N) (e.g., sum, weighted sum, mean, etc.) may be used as the periodicity of the reference acoustic signal.

[0041] In this configuration, the acoustic signal x(n r If the periodicity of ) exceeds the reference value TH1 (in the case of the target signal section), the acoustic signal x(n) with the maximum periodicity will be considered. r ) was output as the acoustic signal x(s) (first acoustic signal). However, the acoustic signal x(n r If the periodicity of ) exceeds the reference value TH1, a specific acoustic signal x(n) can be extracted based on at least one of the periodicities of acoustic signals x(1), ..., x(N), and this acoustic signal x(n) can be output as acoustic signal x(s). For example, acoustic signal x(n r If the periodicity of the acoustic signals x(1), ..., x(N) exceeds the reference value TH1, any acoustic signal x(n) whose periodicity exceeds the average may be output as acoustic signal x(s). Even in such cases, the inclusion of noise based on the external environment into the acoustic signal x(s) can be suppressed.

[0042] In this configuration, the acoustic signal x(n r If the periodicity of ) does not exceed the reference value TH1 (in the case of a non-target signal section), the smallest acoustic signal x(n p) was output as the acoustic signal x(s) (second acoustic signal). However, the acoustic signal x(n r If the periodicity of the acoustic signals x(1), ..., x(N) does not exceed the reference value TH1, a specific acoustic signal x(n) can be extracted based on at least one of the magnitudes of the acoustic signals x(1), ..., x(N), and this acoustic signal x(n) can be output as the acoustic signal x(s). For example, acoustic signal x(n r If the periodicity of the acoustic signals x(1), ..., x(N) does not exceed the reference value TH1, any acoustic signal x(n) whose magnitude is below average may be output as acoustic signal x(s). Even in such cases, the inclusion of noise from the external environment into the acoustic signal x(s) can be suppressed.

[0043] In this configuration, the acoustic signal x(n r When the periodicity of ) exceeds the reference value TH1, and when it does not exceed the reference value TH1, in either case, one of the acoustic signals x(n) included in acoustic signals x(1),...,x(N) is output as acoustic signal x(s). However, multiple acoustic signals included in acoustic signals x(1),...,x(N) may be output as acoustic signal x(s).

[0044] In this embodiment, the acoustic signal selection device 12 outputs an acoustic signal x(s). However, an integer index s may be output from the acoustic signal x(s), or other information for identifying the integer index s or the acoustic signal x(s) may be output. In short, it is sufficient for the acoustic signal selection device 12 to output information corresponding to the acoustic signal x(s) (for example, the acoustic signal x(s), the integer index s, or information for identifying either of these).

[0045] In other words, the acoustic signal selection device 12 should output information corresponding to a first acoustic signal extracted from multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals obtained from multiple acoustic sensors if the periodicity of the reference acoustic signal corresponding to at least one of the multiple acoustic signals exceeds a reference value, and output information corresponding to a second acoustic signal extracted from multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals if the periodicity of the reference acoustic signal does not exceed a reference value. This makes it possible to suppress the intrusion of noise based on the external environment, even when the relationship between the acoustic sensor and the external environment is unknown.

[0046] [Second Embodiment] The acoustic signal selection device of the first embodiment outputs information corresponding to one first acoustic signal extracted from the multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals obtained from multiple acoustic sensors when the periodicity of a reference acoustic signal corresponding to at least one of the multiple acoustic signals exceeds a reference value, and outputs information corresponding to one second acoustic signal extracted from the multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals when the periodicity of the reference acoustic signal does not exceed a reference value. However, information corresponding to one or more first acoustic signals may be output when the periodicity of the reference acoustic signal exceeds a reference value, or information corresponding to one or more second acoustic signals may be output when the periodicity of the reference acoustic signal does not exceed a reference value. The acoustic signal selection device in this embodiment outputs information corresponding to one or more first acoustic signals extracted from multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals obtained from multiple observation signals obtained from multiple acoustic sensors, when the height of the periodicity of a reference acoustic signal corresponding to at least one of the multiple acoustic signals exceeds a reference value, and outputs information corresponding to one or more second acoustic signals extracted from multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals, when the height of the periodicity of the reference acoustic signal does not exceed a reference value. From here on, the explanation will focus on the differences from the matters described so far, and the explanations of the matters described so far will be simplified by reusing the same reference numbers.

[0047] <Structure> As illustrated in Figure 4, the acoustic signal processing system 2 of this embodiment includes N (or more) acoustic sensors 10-1, ..., 10-N, an AD converter 11, an acoustic signal selection device 22, and an acoustic signal processing device 23. In this embodiment, N ≥ 3.

[0048] <Acoustic signal selection device 22> The acoustic signal selection device 22 in this embodiment selects a reference acoustic signal x(n) corresponding to at least one of the multiple acoustic signals x(1), ..., x(N) based on multiple observation signals α(1), ..., α(N) obtained from multiple acoustic sensors 10-1, ..., 10-N. r When the periodicity of the reference acoustic signal x(n) exceeds the reference value (reference) TH1, the system outputs information corresponding to one or more acoustic signals x(s) (first acoustic signals) extracted from the multiple acoustic signals x(1), ..., x(N) based on the periodicity of at least one of the multiple acoustic signals x(1), ..., x(N). On the other hand, the acoustic signal selection device 22 outputs information corresponding to the reference acoustic signal x(n) r If the periodicity of the multiple acoustic signals x(1), ..., x(N) does not exceed the reference value TH1, information corresponding to one or more acoustic signals x(s) (second acoustic signals) extracted from the multiple acoustic signals x(1), ..., x(N) is output based on at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N). In this embodiment, an example is described in which an integer index s∈{1, ..., N} is output as information corresponding to the acoustic signal x(s). The number of integer indices s corresponding to each frame is one or more, and the number of acoustic signals x(s) in each frame is one or more. In other words, the number of acoustic signals x(s) in each frame may be singular or plural.

[0049] As illustrated in Figure 5, the acoustic signal selection device 22 of this embodiment includes buffer units 121-1, ..., 121-N, periodicity calculation units 122-1, ..., 122-N, magnitude calculation units 123-1, ..., 123-N, periodicity comparison unit 224, magnitude comparison unit 225, and selection control unit 227.

[0050] <Acoustic signal processing device 23> The acoustic signal processing device 23 performs acoustic signal processing on the acoustic signal x(s) extracted from the acoustic signals x(1), ..., x(N) based on an integer index s, and outputs the acoustic signal processing result y(s). The acoustic signal processing can be of any type. The acoustic signal processing can be linear or nonlinear. For example, this acoustic signal processing may be beamforming, which uses a linear filter to emphasize or suppress sound components emitted from a sound source in a specific direction. Figure 7 illustrates the configuration of the acoustic signal processing device 23 that performs beamforming. The acoustic signal processing device 23 illustrated in Figure 7 includes a selection control unit 231, a filter providing unit 232, and a beamforming processing unit 233. In addition, for example, this acoustic signal processing may be speech recognition processing, signal separation processing, or other processing.

[0051] <Processing> The acoustic sensors 10-1, ..., 10-N (Figure 4) each detect sound waves or acoustic energy and output the observed signals α(1), ..., α(N) obtained thereby. The observed signals α(1), ..., α(N) are sent to the AD converter 11 (step S10).

[0052] As illustrated in the first embodiment, the AD converter 11 obtains and outputs acoustic signals x(1), ..., x(N) based on the input observation signals α(1), ..., α(N). The acoustic signals x(1), ..., x(N) for each frame are sent to the acoustic signal selection device 22 and the acoustic signal processing device 23 (step S11).

[0053] The acoustic signal selection device 22 (Figure 5) receives the acoustic signals x(1), ..., x(N) for each frame. The acoustic signal selection device 22 selects the reference acoustic signal x(n rWhen the periodicity of a signal exceeds the reference value TH1, the system outputs information s∈{1,…,N} (integer index s) corresponding to one or more acoustic signals x(s) (first acoustic signals) extracted from multiple acoustic signals x(1),…,x(N) based on at least one of the periodicities of the multiple acoustic signals x(1),…,x(N). In this case, the extracted acoustic signal x(s) is an acoustic signal that is judged to have high periodicity among the acoustic signals x(1),…,x(N). For example, the acoustic signal x(s) may be extracted based on a relative evaluation of at least one of the periodicities of multiple acoustic signals x(1),…,x(N) with respect to a certain periodicity criterion (relative evaluation of periodicity), or it may be extracted based on an absolute evaluation of at least one of the periodicities of multiple acoustic signals x(1),…,x(N) (absolute evaluation of periodicity), or it may be extracted based on both the relative evaluation and the absolute evaluation of periodicity. Here, the relative evaluation of periodicity may be an evaluation based on the difference in periodicity, or an evaluation based on the ratio of periodicity. Furthermore, the periodicity criterion is a criterion based on one of the periodicities of multiple acoustic signals x(1), ..., x(N). The periodicity criterion may be one of the periodicities of acoustic signals x(1), ..., x(N) (for example, the maximum or minimum value of the periodicity, etc.), or it may be a statistical value of the periodicity of acoustic signals x(1), ..., x(N) (for example, the mean or median value of the periodicity, etc.). On the other hand, the acoustic signal selection device 22 uses the reference acoustic signal x(n rIf the periodicity of the multiple acoustic signals x(1), ..., x(N) does not exceed the reference value TH1, information s corresponding to one or more acoustic signals x(s) (second acoustic signals) extracted from the multiple acoustic signals x(1), ..., x(N) based on at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N) is output. In this case, the extracted acoustic signal x(s) is an acoustic signal that is judged to have a small magnitude among the acoustic signals x(1), ..., x(N). For example, the acoustic signal x(s) may be extracted based on a relative evaluation of at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N) with respect to a certain magnitude reference (relative evaluation of magnitude), or it may be extracted based on an absolute evaluation of at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N) (absolute evaluation of magnitude), or it may be extracted based on both the relative evaluation and the absolute evaluation of magnitude. Here, the relative evaluation of magnitude may be an evaluation based on the difference in magnitude, or an evaluation based on the ratio of magnitude. Furthermore, the magnitude criterion is a criterion based on the magnitude of one of the multiple acoustic signals x(1), ..., x(N). The magnitude criterion may be one of the magnitudes of the acoustic signals x(1), ..., x(N) (for example, the maximum or minimum magnitude), or it may be a statistical value of the magnitudes of the acoustic signals x(1), ..., x(N) (for example, the mean or median magnitude) (step S22). This will be explained in detail below.

[0054] <Detailed example of step S22> The acoustic signal x(n) (n=1,…,N) of each frame is input to the buffer unit 121-n of the acoustic signal selection device 22 (Figure 5). Each buffer unit 121-n buffers the acoustic signals x(n) of M frames, as illustrated in the first embodiment (step S121).

[0055] The acoustic signal x(n) for M frames buffered in the buffer unit 121-n is sequentially input to each periodicity calculation unit 122-n (n = 1, …, N). As exemplified in the first embodiment, each periodicity calculation unit 122-n calculates and outputs an index r(n) representing the degree of periodicity of the acoustic signal x(n) using the acoustic signal x(n) for M frames. The index r(n) representing each degree of periodicity is sent to the periodicity comparison unit 224 (step S222).

[0056] Indices r(1), …, r(N) representing the degree of periodicity are input to the periodicity comparison unit 224. Using the indices r(1), …, r(N), the periodicity comparison unit 224 extracts one or more acoustic signals x(Φ(1)), …, x(Φ(K)) from the plurality of acoustic signals x(1), …, x(N), and determines a set (the first set) set r ={Φ(1), …, Φ(K)}, and outputs information representing the set set r and information representing the degree of periodicity of the acoustic signals x(Φ(1)), …, x(Φ(K)). Here, the acoustic signals x(Φ(1)), …, x(Φ(K)) are acoustic signals determined to have high periodicity among the acoustic signals x(1), …, x(N). {x(Φ(1)), …, x(Φ(K))} ⊆ {x(1), …, x(N)}, {Φ(1), …, Φ(K)} ⊆ {1, …, N}, Φ(k) ∈ {1, …, N} is a function value of k, k = 1, …, K, and K is an integer of 1 or more (for example, K is an integer of 2 or more). The periodicity comparison unit 224 may determine the set set r using the indices r(1), …, r(N) based on the relative evaluation of periodicity described above, or may determine the set set r using the indices r(1), …, r(N) based on the absolute evaluation of periodicity described above, or may determine the set set r using both the relative evaluation of periodicity and the absolute evaluation of periodicity. Specific examples thereof are shown below.

[0057] <Example of determining the set set r based on the relative evaluation of periodicity> Example of determining the set set rWhen identifying, the periodicity comparison unit 224 uses indices r(1), ..., r(N) to extract acoustic signals x(Φ(1)), ..., x(Φ(K)) based on a relative evaluation of the periodicity of at least one of the multiple acoustic signals x(1), ..., x(N) with respect to a periodicity criterion, and sets them into a set. r Identify {Φ(1),…,Φ(K)}. As mentioned above, the relative evaluation of periodicity may be based on the difference in periodicity or on the ratio of periodicity. Furthermore, the periodicity criterion is a criterion based on one of the periodicities of multiple acoustic signals x(1),…,x(N), and may be one of the periodicities of acoustic signals x(1),…,x(N) (for example, the maximum or minimum value of periodicity, etc.), or it may be a statistical value of the periodicity of acoustic signals x(1),…,x(N) (for example, the mean or median value of periodicity, etc.). Also, if acoustic signals x(1),…,x(N) contain almost no noise components, all acoustic signals x(1),…,x(N) are extracted as acoustic signals x(Φ(1)),…,x(Φ(K)), and set r The values ​​{1, ..., N} may also be set. Specific examples are listed below.

[0058] Specific example (Ar-1): (1) The relative evaluation of periodicity is based on the difference in periodicity. (2) The periodicity criterion is periodicity criterion RB21 which represents the periodicity that exceeds the average of the periodicity heights of multiple acoustic signals x(1),...,x(N). For example, periodicity criterion RB21 represents the maximum value among the periodicity heights of multiple acoustic signals x(1),...,x(N). For example, if the index r(n) representing the periodicity height is a value that does not decrease monotonically with respect to the periodicity height (for example, a value that increases monotonically) (for example, a pitch correlation value), then periodicity criterion RB21 is a value that exceeds the average of the index r(1),...,r(N), for example, the maximum value r(max) of the index r(1),...,r(N). For example, if the index r(n) representing the periodicity height is a value that does not increase monotonically with respect to the periodicity height (for example, a value that decreases monotonically), then periodicity criterion RB21 is a value that falls below the average of the index r(1),...,r(N), for example, the minimum value r(min) of the index r(1),...,r(N). (3) setr It is a set of information corresponding to an acoustic signal x in which the relative value dr of the periodicity height is smaller than the relative reference value RTH21. The relative value dr of the periodicity height of the acoustic signal x in this example is the difference between the periodicity reference RB21 and the periodicity height of the acoustic signal x(n). For example, the relative value dr(n) of the periodicity height of the acoustic signal x(n) in this example is the difference between the periodicity reference RB21 and the index r(n) representing the periodicity height of the acoustic signal x(n). The relative reference value RTH21 is a predetermined value. For example, the relative reference value RTH21 is a value lower than the average value of the relative value dr. Hereinafter, an example will be illustrated in which the index r(n) is a value that is monotonically non-decreasing with respect to the periodicity height (for example, a pitch correlation value), the periodicity reference RB21 is the maximum value r(max) of the indices r(1),..., r(N), and the information for specifying the acoustic signal x is the integer index n of the acoustic signal x(n). The periodicity comparison unit 224 in this example, for example, arranges the indices r(1),..., r(N) in descending order, sets the largest index r(max) ∈ {r(1),..., r(N)} as the periodicity reference RB21 (RB21 = r(max)), and calculates the relative value dr(n) of the periodicity height of the acoustic signal x(n) as in the following formula (1). dr(n)=r(max)-r(n)(n = 1,..., N) (1) Next, the periodicity comparison unit 224 sets the set of integer indices n corresponding to the relative value dr(n) that is less than the relative reference value RTH21 as the set set r ⊆{1,..., N}. In this example, the smaller dr(n) is, the larger r(n) is, and the higher the periodicity is. That is, the frame of the acoustic signal x(n) corresponding to the information belonging to the set set r in this example is estimated to be the target signal section.

[0059] Specific example (A-r-2): (1) The relative evaluation of periodicity is based on the difference in periodicity. (2) The periodicity criterion is the periodicity criterion RB22, which represents the periodicity below the average of the periodicity heights of multiple acoustic signals x(1),...,x(N). For example, the periodicity criterion RB22 represents the minimum (minimum value) of the periodicity heights of multiple acoustic signals x(1),...,x(N). For example, if the index r(n) representing the periodicity height is a value that does not decrease monotonically with respect to the periodicity height (for example, a value that increases monotonically) (for example, a pitch correlation value), then the periodicity criterion RB22 is a value that is below the average of the indices r(1),...,r(N), for example, the minimum value r(min) of the indices r(1),...,r(N). For example, if the index r(n) representing the periodicity height is a value that does not increase monotonically with respect to the periodicity height (for example, a value that decreases monotonically), then the periodicity criterion RB22 is a value that exceeds the average of the indices r(1),...,r(N), for example, the maximum value r(max) of the indices r(1),...,r(N). (3) set r This is a set of information corresponding to an acoustic signal x whose relative periodicity value dr is greater than the relative reference value RTH22. In this example, the relative periodicity value dr of acoustic signal x is the difference between the periodicity reference RB22 and the periodicity of acoustic signal x. For example, in this example, the relative periodicity value dr(n) of acoustic signal x(n) is the difference between the periodicity reference RB22 and the index r(n) that represents the periodicity of acoustic signal x(n). The relative reference value RTH22 is a predetermined value. For example, the relative reference value RTH22 is a value that is greater than the average value of the relative value dr. The following example illustrates a case where the index r(n) is a value that does not monotonically decrease with respect to the height of periodicity (for example, the pitch correlation value), the periodicity criterion RB22 is the minimum value r(min) of the indices r(1),...,r(N), and the information for identifying the acoustic signal x is the integer index n of the acoustic signal x(n). In this example, the periodicity comparison unit 224 arranges the indices r(1),...,r(N) in ascending order, sets the smallest index r(min)∈{r(1),...,r(N)} as the periodicity criterion RB22 (RB22=r(min)), and calculates the relative value dr(n) of the height of periodicity of the acoustic signal x(n) as shown in equation (2) below. dr(n)=r(n)-r(min)(n=1,…,N) (2) Next, the periodicity comparison unit 224 selects the set of integer indices n corresponding to relative values ​​dr(n) that exceed the relative reference value RTH22. r Let ⊆{1,…,N}. In this example, the larger dr(n), the larger r(n) and the higher the periodicity. That is, the set in this example r The frame of the acoustic signal x(n) corresponding to the information belonging to is presumed to be the target signal segment.

[0060] Specific example (Br): (1) The relative evaluation of periodicity is based on the ratio of periodicity. (2) The periodicity criterion is a predetermined periodicity criterion RB23. The periodicity criterion RB23 may be any value. For example, the periodicity criterion RB23 may be a value that represents the average of the periodicity levels of multiple acoustic signals x(1), ..., x(N), a value that exceeds the average, a maximum value, a value that falls below the average, or a minimum value. For example, the periodicity criterion RB23 may be the average value of the indices r(1), ..., r(N), a value that exceeds the average, a maximum value, a value that falls below the average, or a minimum value. (3) set r This is a set of information corresponding to an acoustic signal x whose relative periodicity level dr exceeds the periodicity criterion RTH23. In this example, the relative periodicity level dr of acoustic signal x is the ratio of the periodicity level of acoustic signal x to the periodicity criterion RB23 (periodicity level of acoustic signal x / periodicity criterion RB23). For example, in this example, the relative periodicity level dr(n) of acoustic signal x(n) is the ratio of the index r(n) representing the periodicity level of acoustic signal x(n) to the periodicity criterion RB23. The following example illustrates a case where the index r(n) is a value that does not monotonically decrease with respect to the degree of periodicity (for example, the pitch correlation value), the periodicity criterion RB23 is the maximum value r(max) of the indices r(1), ..., r(N), and the information for identifying the acoustic signal x is the integer index n of the acoustic signal x(n). In this example, the periodicity comparison unit 224 arranges the indices r(1), ..., r(N) in descending order, sets the largest index r(max) as the periodicity criterion RB23 (RB23 = r(max)), and calculates the relative value dr(n) of the degree of periodicity of the acoustic signal x(n) as shown in equation (3) below. dr(n)=r(n) / r(max)(n=1,…,N) (3) Next, the periodicity comparison unit 224 selects the set of integer indices n corresponding to relative values ​​dr(n) that exceed the relative reference value RTH23. r Let ⊆{1,…,N}. In this example, the larger dr(n), the larger r(n) and the higher the periodicity. That is, the set in this example r The frame of the acoustic signal x(n) corresponding to the information belonging to is presumed to be the target signal segment.

[0061] Specific example (AB-r): In specific examples (Ar-1), (Ar-2), and (Br), instead of evaluating all acoustic signals x(1), ..., x(N) (as subjects for relative evaluation of periodicity), only a portion of the acoustic signals x(1), ..., x(N) may be evaluated.

[0062] <Set based on absolute evaluation of periodicity> r Examples of identifying specific individuals > Set based on absolute evaluation of periodicity r When identifying, the periodicity comparison unit 224 uses indices r(1), ..., r(N) to extract acoustic signals x(Φ(1)), ..., x(Φ(K)) based on an absolute evaluation of the periodicity of at least one of the multiple acoustic signals x(1), ..., x(N) relative to the periodicity criterion, and sets them into a set. r Identify {Φ(1),…,Φ(K)}.

[0063] Specific example (Cr-1): The periodicity comparison unit 224 may extract a predetermined number (K, where K is a constant such that 2 ≦ K < N in this example) of acoustic signals x(Φ(1)),..., x(Φ(K)) in descending order of the periodicity represented by the index r(n). For example, when the index r is a value that is monotonically non-decreasing with respect to the degree of periodicity (e.g., a monotonically increasing value, pitch correlation value, etc.), the periodicity comparison unit 224 may arrange the index r in descending order and extract the acoustic signals x(Φ(1)),..., x(Φ(K)) corresponding to a predetermined number (K) of indices r(Φ(1)),..., r(Φ(K)) selected in descending order. On the other hand, for example, when the index r is a value that is monotonically non-increasing with respect to the degree of periodicity (e.g., a monotonically decreasing value), the periodicity comparison unit 224 may arrange the index r in ascending order and extract the acoustic signals x(Φ(1)),..., x(Φ(K)) corresponding to a predetermined number (K) of indices r(Φ(1)),..., r(Φ(K)) selected in ascending order.

[0064] Specific example (C-r-2): The periodicity comparison unit 224 may extract acoustic signals x(Φ(1)),..., x(Φ(K)) whose degree of periodicity represented by the index r(n) exceeds the absolute reference value ATH21. For example, when the index r is a value that is monotonically non-decreasing with respect to the degree of periodicity (e.g., a monotonically increasing value, pitch correlation value, etc.), the periodicity comparison unit 224 may extract the acoustic signals x(Φ(1)),..., x(Φ(K)) corresponding to the indices r(Φ(1)),..., r(Φ(K)) that exceed the absolute reference value ATH21. On the other hand, for example, when the index r is a value that is monotonically non-increasing with respect to the degree of periodicity (e.g., a monotonically decreasing value), the periodicity comparison unit 224 may extract the acoustic signals x(Φ(1)),..., x(Φ(K)) corresponding to the indices r(Φ(1)),..., r(Φ(K)) that are below the absolute reference value ATH21. The absolute reference value ATH21 is a predetermined value. For example, the absolute reference value ATH21 is a value representing a height that exceeds the average of the degree of periodicity of the acoustic signal represented by the index r(n).

[0065] Specific example (C-r): In specific examples (C-r-1) and (C-r-2), instead of using all the acoustic signals x(1), …, x(N) as the evaluation targets (targets for absolute evaluation of periodicity), only a part of the acoustic signals x(1), …, x(N) may be used as the evaluation targets.

[0066] <Example of specifying set r based on relative and absolute evaluations of periodicity> The set r may be specified by combining relative and absolute evaluations of periodicity. Specific examples of relative and absolute evaluations of periodicity are as described above.

[0067] Specific example (D-r-1): The periodicity comparison unit 224 may use the indices r(1), …, r(N) to specify some elements of the set r by relative evaluation of periodicity and specify the remaining elements of the set r by absolute evaluation of periodicity. Hereinafter, cases where specific examples (A-r-1) and (AB-r) are used for relative evaluation of periodicity and specific examples (C-r-1) and (C-r) are used for absolute evaluation of periodicity will be exemplified. Here, the index r(n) is a value that is monotonically non-decreasing with respect to the height of periodicity (for example, a pitch correlation value), the periodicity criterion RB21 is the maximum value r(max) of the indices r(1), …, r(N), and the information for specifying the acoustic signal x is the integer index n of the acoustic signal x(n). The periodicity comparison unit 224 in this example, for example, arranges the indices r(1), …, r(N) in descending order to obtain the rearranged indices r(θ(1)), …, r(θ(N)). The largest index r(θ(1)) is used as the periodicity criterion RB21 (RB21 = r(θ(1))), and as shown in the following formula (4), the relative value dr(n’) of the height of periodicity of the acoustic signal x(n’) is calculated. dr(n’)=r(θ(1))-r(n’)(n’=μ,…,N) (4) Here, μ is an integer satisfying 2 ≦ μ < N. For example, μ = 3. Next, the periodicity comparison unit 224 calculates the union of the set of integer indices n' corresponding to relative values ​​dr(n') that are below the relative reference value RTH21, and the set of integer indices θ(1), ..., θ(μ-1) corresponding to indices r(θ(1)), ..., r(θ(μ-1)), and the set set. r Let {Φ(1),…,Φ(K)} ⊆ {1,…,N}. This set r Of the elements, the frame of the acoustic signal x(n') corresponding to the integer index n' was estimated to be the target signal interval based on the relative evaluation of periodicity, and the frames of the acoustic signals x(θ(1)),…,x(θ(μ-1)) corresponding to the integer index θ(1),…,θ(μ-1) were estimated to be the target signal interval based on the absolute evaluation of periodicity.

[0068] Specific example (Dr-2): The periodicity comparison unit 224 uses indices r(1), ..., r(N) to perform a comprehensive evaluation of the relative and absolute evaluations of periodicity, thereby creating a set. r The elements may be identified. Below are examples of cases where specific examples (Ar-1) and (AB-r) are used for the relative evaluation of periodicity, and specific examples (Cr-1) and (Cr) are used for the absolute evaluation of periodicity. Here, the index r(n) is a value that does not decrease monotonically with respect to the height of periodicity (for example, the pitch correlation value), the periodicity criterion RB21 is the maximum value r(max) of the indices r(1),...,r(N), and the information for identifying the acoustic signal x is the integer index n of the acoustic signal x(n). In this example, the periodicity comparison unit 224 sorts the indices r(1),...,r(N) in descending order, for example, and obtains the sorted indices r(θ(1)),...,r(θ(N)). The largest index r(θ(1)) is taken as the periodicity criterion RB21 (RB21=r(θ(1))), and the relative value dr(n) of the height of periodicity of the acoustic signal x(n') is calculated as shown in equation (5) below. dr(n)=r(θ(1))-r(n)(n=1,…,N) (5) Next, the periodicity comparison unit 224 sets the product set of the set of integer indices n corresponding to the relative value dr(n) that is less than the relative reference value RTH21 and the set of integer indices θ(1), …, θ(ω) corresponding to the indicators r(θ(1)), …, r(θ(ω)) as the set set r = {Φ(1), …, Φ(K)} ⊆ {1, …, N}. Here, ω is an integer satisfying 2 ≤ ω < N. For the acoustic signals x corresponding to the elements of this set set r it is estimated that they are the target signal sections in both the relative evaluation and the absolute evaluation of periodicity.

[0069] The set set r = {Φ(1), …, Φ(K)} and the information representing the periodicity heights of the acoustic signals x(Φ(1)), …, x(Φ(K)) are sent to the selection control unit 227 of the acoustic signal selection device 22 (FIG. 5). In this embodiment, as the information representing the periodicity heights of the acoustic signals x(Φ(1)), …, x(Φ(K)), the indicators r(Φ(1)), …, r(Φ(K)), that is, the indicators r(w) (w ∈ set r ) are sent (step S224).

[0070] The acoustic signals x(n) for M frames buffered in the buffer unit 121-n are also input one by one to each magnitude calculation unit 123-n (n = 1, …, N). Each magnitude calculation unit 123-n calculates and outputs an indicator p(n) representing the magnitude of the acoustic signals x(n) for M frames as exemplified in the first embodiment. The indicator p(n) representing the magnitude of each acoustic signal x(n) is sent to the magnitude comparison unit 225 (step S223).

[0071] Indicators p(1), …, p(N) representing magnitudes are input to the magnitude comparison unit 225. Using the indicators p(1), …, p(N), the magnitude comparison unit 225 extracts one or more acoustic signals x(Θ(1)), …, x(Θ(Q)) from the plurality of acoustic signals x(1), …, x(N), specifies the set (second set) set p = {Θ(1), …, Θ(Q)}, and the set set pThe output represents the set. Here, acoustic signals x(Θ(1)),...,x(Θ(Q)) are acoustic signals that are judged to have a small magnitude among acoustic signals x(1),...,x(N). {x(Θ(1)),...,x(Θ(Q))}⊆{x(1),...,x(N)}, {Θ(1),...,Θ(Q)}⊆{1,...,N}, Θ(q)∈{1,...,N} is a function value of k, q=1,...,Q, and Q is an integer greater than or equal to 1 (for example, Q is an integer greater than or equal to 2). The magnitude comparison unit 225 uses the indices p(1),...,p(N) and the set based on the relative magnitude evaluation described above. p You can specify the set, or you can use the absolute size evaluation mentioned above to determine the set. p You may specify the size, or you may set a set based on the relative evaluation and the absolute evaluation of the size. p You may specify these. Specific examples are given below.

[0072] <set based on relative size evaluation> p Examples of identifying specific individuals > Set based on relative size p When identifying, the size comparison unit 225 uses indices p(1), ..., p(N) to extract acoustic signals x(Θ(1)), ..., x(Θ(Q)) based on a relative evaluation of the magnitude of at least one of the multiple acoustic signals x(1), ..., x(N) with respect to a size criterion, and sets them into a set. p Identify {Θ(1),…,Θ(Q)}. As mentioned above, the relative evaluation of magnitude may be based on the difference in magnitude or on the ratio of magnitude. Furthermore, the magnitude criterion is a criterion based on the magnitude of one of the multiple acoustic signals x(1),…,x(N), and may be one of the magnitudes of the acoustic signals x(1),…,x(N) (for example, the maximum or minimum magnitude), or it may be a statistical value of the magnitude of the acoustic signals x(1),…,x(N) (for example, the mean or median magnitude). Specific examples are listed below.

[0073] Specific example (Ap-1): (1) Relative evaluation of size is based on the difference in size. (2) The magnitude criterion is a magnitude criterion PB21 that represents a magnitude greater than the average magnitude of multiple acoustic signals x(1),...,x(N). For example, magnitude criterion PB21 represents the largest magnitude (maximum value) among multiple acoustic signals x(1),...,x(N). For example, if the magnitude index p(n) is a value that does not decrease monotonically with respect to magnitude (for example, a value that increases monotonically) (for example, power), then magnitude criterion PB21 is a value that exceeds the average of the indices p(1),...,p(N), for example, the maximum value p(max) of the indices p(1),...,p(N). For example, if the magnitude index p(n) is a value that does not increase monotonically with respect to magnitude (for example, a value that decreases monotonically), then magnitude criterion PB21 is a value that falls below the average of the indices p(1),...,p(N), for example, the minimum value p(min) of the indices p(1),...,p(N). (3) set p This is a set of information corresponding to an acoustic signal x whose relative magnitude value dp is greater than the relative reference value PTH21. In this example, the relative magnitude value dp of the acoustic signal x is the difference between the magnitude reference PB21 and the magnitude of the acoustic signal x(n). For example, in this example, the relative magnitude value dp(n) of the acoustic signal x(n) is the difference between the magnitude reference PB21 and the index p(n) that represents the magnitude of the acoustic signal x(n). The relative reference value PTH21 is a predetermined value. For example, the relative reference value PTH21 is a value that is greater than the average value of the relative magnitude value dp. The following example illustrates a case where the index p(n) is a value that does not decrease monotonically with respect to magnitude (e.g., power), the magnitude reference PB21 is the maximum value p(max) of the indices p(1), ..., p(N), and the information for identifying the acoustic signal x is the integer index n of the acoustic signal x(n). In this example, the magnitude comparison unit 225 arranges the indices p(1), ..., p(N) in descending order, sets the largest index p(max) ∈ {p(1), ..., p(N)} as the magnitude reference PB21 (PB21 = p(max)), and calculates the relative magnitude dp(n) of the acoustic signal x(n) as shown in equation (6) below. dp(n)=p(max)-p(n)(n=1,…,N) (6) Next, the size comparison unit 225 selects the set of integer indices n corresponding to relative values ​​dp(n) that exceed the relative reference value PTH21. p Let ⊆{1,…,N}. In this example, the larger dp(n), the smaller p(n) is, and therefore the smaller the magnitude.

[0074] Specific example (Ap-2): (1) Relative evaluation of size is based on the difference in size. (2) The magnitude criterion is a magnitude criterion PB22 that represents a magnitude below the average magnitude of multiple acoustic signals x(1), ..., x(N). For example, magnitude criterion PB22 represents the smallest magnitude among multiple acoustic signals x(1), ..., x(N). For example, if the magnitude index p(n) is a value that does not decrease monotonically with respect to magnitude (for example, a value that increases monotonically) (for example, power), then magnitude criterion PB22 is a value below the average of the indices p(1), ..., p(N), for example, the minimum value p(min) of the indices p(1), ..., p(N). For example, if the magnitude index p(n) is a value that does not increase monotonically with respect to magnitude (for example, a value that decreases monotonically), then magnitude criterion PB22 is a value that exceeds the average of the indices p(1), ..., p(N), for example, the maximum value p(max) of the indices p(1), ..., p(N). (3) set p This is a set of information corresponding to an acoustic signal x whose relative magnitude value dp is smaller than the relative reference value PTH22. In this example, the relative magnitude value dp of the acoustic signal x is the difference between the magnitude reference PB22 and the magnitude of the acoustic signal x. For example, in this example, the relative magnitude value dp(n) of the acoustic signal x(n) is the difference between the magnitude reference PB22 and the index p(n) that represents the magnitude of the acoustic signal x(n). The relative reference value PTH22 is a predetermined value. For example, the relative reference value PTH22 is a value below the average value of the relative magnitude value dp. The following example illustrates a case where the index p(n) is a value that does not decrease monotonically with respect to magnitude (e.g., power), the magnitude reference PB22 is the minimum value p(min) of the indices p(1), ..., p(N), and the information for identifying the acoustic signal x is the integer index n of the acoustic signal x(n). In this example, the magnitude comparison unit 225 arranges the indices p(1), ..., p(N) in ascending order, sets the smallest index p(min) ∈ {p(1), ..., p(N)} as the magnitude reference PB22 (PB22 = p(min)), and calculates the relative magnitude dp(n) of the acoustic signal x(n) as shown in equation (7) below. dp(n)=p(n)-p(min)(n=1,…,N) (7) Next, the size comparison unit 225 selects the set of integer indices n corresponding to relative values ​​dp(n) that are lower than the relative reference value PTH22. p Let ⊆{1,…,N}. In this example, the smaller dp(n), the smaller p(n) and the smaller the magnitude.

[0075] Specific example (Bp): (1) Relative evaluation of size is based on the ratio of size. (2) The magnitude criterion is a predetermined magnitude criterion PB23. The magnitude criterion PB23 may be any value. For example, the magnitude criterion PB23 may be a value that represents the average magnitude of multiple acoustic signals x(1), ..., x(N), a value that exceeds the average, a maximum value, a value that falls below the average, or a minimum value. For example, the magnitude criterion PB23 may be the average value of indices p(1), ..., p(N), a value that exceeds the average, a maximum value, a value that falls below the average, or a minimum value. (3) set pThis is a set of information corresponding to acoustic signals x whose relative magnitude value dp is less than the magnitude reference PTH23. In this example, the relative magnitude value dp of acoustic signal x is the ratio of the magnitude of acoustic signal x to the magnitude reference PB23 (magnitude of acoustic signal x / magnitude reference PB23). For example, in this example, the relative magnitude value dp(n) of acoustic signal x(n) is the ratio of the index p(n) representing the magnitude of acoustic signal x(n) to the magnitude reference PB23. The following example illustrates a case where the index p(n) is a value that does not decrease monotonically with respect to magnitude (e.g., power), the magnitude reference PB23 is the minimum value p(min) of the indices p(1), ..., p(N), and the information for identifying the acoustic signal x is the integer index n of the acoustic signal x(n). In this example, the magnitude comparison unit 225 arranges the indices p(1), ..., p(N) in ascending order, sets the smallest index p(min) as the magnitude reference PB23 (PB23 = p(min)), and calculates the relative magnitude dp(n) of the acoustic signal x(n) as shown in equation (8) below. dp(n)=p(n) / p(min)(n=1,…,N) (8) Next, the size comparison unit 225 selects the set of integer indices n corresponding to relative values ​​dp(n) that are lower than the relative reference value PTH23. p Let ⊆{1,…,N}. In this example, the smaller dp(n), the smaller p(n) and the smaller the magnitude.

[0076] Specific example (AB-p): In specific examples (Ap-1), (Ap-2), and (Bp), instead of evaluating all acoustic signals x(1), ..., x(N) (for relative magnitude evaluation), only a portion of the acoustic signals x(1), ..., x(N) may be evaluated.

[0077] <Set based on absolute evaluation of size p Examples of identifying specific individuals > Set based on absolute size pWhen identifying, the magnitude comparison unit 225 uses the indices p(1),..., p(N) to extract the acoustic signals x(Θ(1)),..., x(Θ(Q)) based on at least any absolute evaluation of the magnitudes of the plurality of acoustic signals x(1),..., x(N) with respect to the magnitude reference, and identifies the set set p ={Θ(1),..., Θ(Q)}.

[0078] Specific example (C-p-1): The magnitude comparison unit 225 may extract a predetermined number (Q, where Q is a constant with 2 ≤ Q < N in this example) of acoustic signals x(Θ(1)),..., x(Θ(Q)) in ascending order of the magnitude represented by the index p(n). For example, when the index p is a value that is monotonically non-decreasing with respect to the magnitude (e.g., a monotonically increasing value, power, etc.), the magnitude comparison unit 225 may arrange the index p in ascending order and extract the acoustic signals x(Θ(1)),..., x(Θ(Q)) corresponding to a predetermined number (Q) of the indices p(Θ(1)),..., p(Θ(Q)) selected in ascending order. On the other hand, for example, when the index p is a value that is monotonically non-increasing with respect to the magnitude (e.g., a monotonically decreasing value), the magnitude comparison unit 225 may arrange the index p in descending order and extract the acoustic signals x(Θ(1)),..., x(Θ(Q)) corresponding to a predetermined number (Q) of the indices p(Θ(1)),..., p(Θ(Q)) selected in descending order.

[0079] Specific example (C-p-2): The magnitude comparison unit 225 may extract acoustic signals x(Θ(1)), ..., x(Θ(Q)) whose magnitude, represented by index p(n), is below the absolute reference value ATH22. For example, if index p is a value that does not decrease monotonically with respect to magnitude (e.g., a value that increases monotonically, such as power), the magnitude comparison unit 225 may extract acoustic signals x(Θ(1)), ..., x(Θ(Q)) corresponding to index p(Θ(1)), ..., p(Θ(Q)) that are below the absolute reference value ATH22. On the other hand, for example, if index p is a value that does not increase monotonically with respect to magnitude (e.g., a value that decreases monotonically), the magnitude comparison unit 225 may extract acoustic signals x(Θ(1)), ..., x(Θ(Q)) corresponding to index p(Θ(1)), ..., p(Θ(Q)) that are above the absolute reference value ATH22. The absolute reference value ATH22 is a predetermined value. For example, the absolute reference value ATH22 is a value that represents a magnitude below the average magnitude of the acoustic signal represented by index p(n).

[0080] Specific example (Cp): In specific examples (Cp-1) and (Cp-2), instead of evaluating all acoustic signals x(1), ..., x(N) (as subjects of absolute magnitude evaluation), only a portion of the acoustic signals x(1), ..., x(N) may be evaluated.

[0081] <Set based on relative and absolute evaluation of size p Examples of identifying specific individuals > A set is created by combining relative and absolute size evaluations. p You may specify the size. Specific examples of relative and absolute evaluation of size are as described above.

[0082] Specific example (Dp-1): The size comparison unit 225 uses indices p(1), ..., p(N) to evaluate the relative size of the set. p Identify some of the elements and use absolute evaluation of their size to determine the set. pThe remaining elements may be specified. Hereinafter, specific examples (A-p-2) and (AB-p) will be used for relative evaluation of magnitude, and specific examples (C-p-1) and (C-p) will be used for absolute evaluation of magnitude. Here, the index p(n) is a monotonically non-decreasing value (e.g., power) with respect to magnitude, the magnitude reference PB22 is the minimum value p(min) of the indices p(1),..., p(N), and the information for specifying the acoustic signal x is the integer index n of the acoustic signal x(n). The magnitude comparison unit 225 in this example, for example, arranges the indices p(1),..., p(N) in ascending order to obtain the rearranged indices p(ζ(1)),..., p(ζ(N)). The smallest index p(ζ(1)) is used as the magnitude reference PB22 (PB22 = p(ζ(1))), and the relative value dp(n”) of the magnitude of the acoustic signal x(n”) is calculated as in the following formula (9). dp(n”)=p(n”)-p(ζ(1))(n”=ν,…,N) (9) Here, ν is an integer satisfying 2≦ν<N. For example, ν = 3. Next, the magnitude comparison unit 225 determines the union of the set of integer indices n” corresponding to the relative values dp(n”) that are less than the relative reference value PTH22 and the set of integer indices ζ(1),..., ζ(ν - 1) corresponding to the indices p(ζ(1)),..., p(ζ(ν - 1)) as the set set p ={Θ(1),…,Θ(Q)}⊆{1,…,N}. Among the elements of this set set p the acoustic signal x(n”) corresponding to the integer index n” is estimated to have a small magnitude based on relative evaluation of magnitude, and the acoustic signals x(ζ(1)),..., x(ζ(ν - 1)) corresponding to the integer indices ζ(1),..., ζ(ν - 1) are estimated to have a small magnitude based on absolute evaluation of magnitude.

[0083] Specific example (D-p-2): The magnitude comparison unit 225 uses the indices p(1),..., p(N) to obtain the set set based on a comprehensive evaluation of relative and absolute evaluations of magnitude pThe elements of may also be specified. Hereinafter, cases where specific examples (A-p-2) and (AB-p) are used for relative evaluation of magnitude, and specific examples (C-p-1) and (C-p) are used for absolute evaluation of magnitude will be exemplified. Here, the index p(n) is a value that is monotonically non-decreasing with respect to magnitude (for example, power), the magnitude reference PB22 is the minimum value p(min) of the indices p(1),..., p(N), and the information for specifying the acoustic signal x is the integer index n of the acoustic signal x(n). The magnitude comparison unit 225 in this example, for example, arranges the indices p(1),..., p(N) in ascending order to obtain the rearranged indices p(ζ(1)),..., p(ζ(N)). The smallest index p(ζ(1)) is used as the magnitude reference PB22 (PB22 = p(ζ(1))), and the relative value dp(n) of the magnitude of the acoustic signal x(n') is calculated as in the following equation (10). dp(n)=p(n)-p(ζ(1))(n = 1,…,N) (10) Next, the magnitude comparison unit 225 determines the product set of the set of integer indices n corresponding to the relative value dp(n) that is less than the relative reference value PTH22 and the set of integer indices ζ(1),..., ζ(κ) corresponding to the indices p(ζ(1)),..., p(ζ(κ)). p Let it be set p ={Θ(1),…,Θ(Q)}⊆{1,…,N}. Here, κ is an integer satisfying 2≦κ<N. The acoustic signal x corresponding to the elements of this set

[0084] set p ={Θ(1),…,Θ(Q)} is sent to the selection control unit 227 of the acoustic signal selection device 22 (FIG. 5) (step S225).

[0085] The selection control unit 227 receives information representing set r ={Φ(1),…,Φ(K)}, the index r(w) (w ∈ set r )(information representing the degree of periodicity of the acoustic signals x(Φ(1)),..., x(Φ(K))), and set pInformation representing ={Θ(1),…,Θ(Q)} is input. The selection control unit 227 selects a set if the periodicity (periodicity of the reference acoustic signal) represented by index r(w) exceeds the reference value TH1. r The integer indices s=Φ(1),…,Φ(K) which are elements of the set (information corresponding to the first acoustic signal extracted from multiple acoustic signals based on at least one of the periodicities of the multiple acoustic signals) are output. On the other hand, the selection control unit 227 outputs the set if the height of the periodicity represented by the index r(w) (the height of the periodicity of the reference acoustic signal) does not exceed the reference value TH1. p The integer indices s = Θ(1), ..., Θ(Q), which are elements of the array, are output (information corresponding to a second acoustic signal extracted from multiple acoustic signals based on at least one of the magnitudes of multiple acoustic signals is output). The integer indices s may be output per frame or per multiple frames. The integer indices s are sent to the acoustic signal processing device 23 (Figure 4) (step S227).

[0086] <Detailed example of step S227> For example, as illustrated in Figure 6, the index r(w) is input to the periodicity determination unit 227a of the selection control unit 227. The periodicity determination unit 227a determines whether the height of the periodicity represented by the index r(w) exceeds the reference value TH1. The periodicity determination unit 227a determines whether any of the index r(w)(w∈set r You can also make a decision using an index r(w) that represents the highest periodicity (for example, an index r(w) that represents the highest periodicity), or you can make a decision using an index r(w) (w∈set). r) may be used as an index to determine periodicity using statistical values ​​(e.g., summation value, weighted summation value, mean value, etc.). For example, if the index r(w) is a value that does not monotonically decrease with respect to the height of periodicity (e.g., a value that monotonically increases, a pitch correlation value, etc.), the periodicity determination unit 227a determines whether the index representing periodicity exceeds the threshold corresponding to the reference value TH1 (e.g., reference value TH1). On the other hand, for example, if the index r(w) is a value that does not monotonically increase with respect to the height of periodicity (e.g., a value that monotonically decreases), the periodicity determination unit 227a determines whether the index representing periodicity is less than the threshold corresponding to the reference value TH1 (e.g., reference value TH1). The determination of whether the index exceeds the threshold and the determination of whether the index is less than the threshold may be made by comparing the index with the threshold, by comparing the value corresponding to the threshold with the index, or by comparing the value corresponding to the index with the threshold. The periodicity determination unit 227a determines that the height of periodicity represented by the index exceeds the reference value TH1. r , or the judgment result that the degree of periodicity represented by the indicator does not exceed the standard value TH1 d p The following will be output. Judgment result d r or judgment result d p This is sent to the switching control unit 227b (step S227a).

[0087] The switching control unit 227b has a set r , set p , and judgment result d r or judgment result d p The input is d. The switching control unit 227b determines the result d r If this is input (when the periodicity of the reference acoustic signal exceeds the reference value), the set r The corresponding integer index s=Φ(1),…,Φ(K) (information corresponding to the first acoustic signal extracted from multiple acoustic signals based on at least one of the periodicities of the multiple acoustic signals) is output. Meanwhile, the switching control unit 227b determines the result d p If this is input (if the periodicity of the reference acoustic signal does not exceed the reference value), the set pThe corresponding integer indices s = Θ(1), ..., Θ(Q) are output (information corresponding to the second acoustic signal extracted from multiple acoustic signals based on at least one of the magnitudes of the multiple acoustic signals is output). The integer indices s are sent to the acoustic signal processing device 23 (Figure 4) (step S227b).

[0088] The acoustic signal processing unit 23 receives acoustic signals x(1), ..., x(N) and integer index s for each frame as input. The acoustic signal processing unit 23 performs acoustic signal processing on the acoustic signal x(s) corresponding to the integer index s and outputs the acoustic signal processing result y(s) for each frame. Each frame has one or more acoustic signals x(s), and the acoustic signal processing result y(s) for each frame is the result obtained by performing acoustic signal processing on one or more acoustic signals x(s). For example, in the acoustic signal processing unit 23 illustrated in Figure 7, the selection control unit 231 receives acoustic signals x(1), ..., x(N) and integer index s for each frame as input, and the filter providing unit 232 receives integer index s as input. The selection control unit 231 uses the acoustic signals x(1), ..., x(N) and integer index s to extract and output the acoustic signal x(s) for each frame from the acoustic signals x(1), ..., x(N). The acoustic signal x(s) is sent to the beamforming processing unit 233. The filter provider 232 applies a filter W to the acoustic signal x(s) using an integer index s. s It generates and outputs (filter coefficients). In other words, the filter provider 232 provides a filter W for beamforming with the acoustic signal x(s). s Generates and outputs the following: Filter W s This is a linear filter for beamforming that emphasizes or suppresses sound components emitted from a sound source in a specific direction. Filter W s This may be generated per frame, or every multiple frames. Filter W s The signal is sent to the beamforming processing unit 233. The beamforming processing unit 233 filters the received acoustic signal x(s) with a filter W. s Apply the following to obtain and output the acoustic signal processing result y(s) for each frame (step S23).

[0089] The processes described in steps S10 and S11 are performed on the time-series signals of the observed signals α(1), ..., α(N), and the processes in steps S22 and S23 are repeatedly performed for each frame. As a result, the acoustic signal processing device 23 outputs time-series information of the acoustic signal processing result y(s).

[0090] <Features of this form> The acoustic signal selection device 22 in this embodiment selects a reference acoustic signal x(n) corresponding to at least one of the multiple acoustic signals x(1), ..., x(N) based on multiple observation signals α(1), ..., α(N) obtained from multiple acoustic sensors 10-1, ..., 10-N. r When the periodicity of the reference acoustic signal x(n) exceeds the reference value (reference) TH1, the system outputs information corresponding to one or more acoustic signals x(s) (first acoustic signals) extracted from the multiple acoustic signals x(1), ..., x(N) based on the periodicity of at least one of the multiple acoustic signals x(1), ..., x(N). On the other hand, the acoustic signal selection device 22 outputs information corresponding to the reference acoustic signal x(n) r If the periodicity of the acoustic sensors x(1), ..., x(N) does not exceed the reference value TH1, information corresponding to one or more acoustic signals x(s) (second acoustic signals) extracted from the multiple acoustic signals x(1), ..., x(N) is output based on at least one of the magnitudes of the multiple acoustic signals x(1), ..., x(N). This makes it possible to suppress the inclusion of noise based on the external environment into the acoustic signal x(s) even when the relationship between the acoustic sensors 10-1, ..., 10-N and the external environment is unknown. Information regarding the relationship between the acoustic sensors 10-1, ..., 10-N and the external environment is not required for this process.

[0091] Furthermore, in this embodiment, it is preferable to set a set based on the relative and absolute evaluation of the periodicity of the acoustic signal. r It is desirable to identify the set based on relative and absolute evaluation of periodicity. r(Example of identifying) That is, when the periodicity of the reference acoustic signal exceeds the reference value TH1, it is desirable that the acoustic signal selection device 22 outputs information corresponding to the multiple acoustic signals x(s) (first acoustic signals) extracted from the multiple acoustic signals x(1), ..., x(N) based on a relative evaluation of at least one of the periodicity references of the multiple acoustic signals x(1), ..., x(N) and an absolute evaluation of at least one of the periodicity references of the multiple acoustic signals x(1), ..., x(N). More preferably, it is desirable that the multiple acoustic signals x(s) (first acoustic signals) include acoustic signals extracted based on relative evaluation and acoustic signals extracted based on absolute evaluation. In other words, an index representing the height of periodicity, such as the pitch correlation value, becomes a larger value as the number of highly periodic components in the acoustic signal increases, but it may become an even larger value if noise components are added. For this reason, with only an absolute evaluation of periodicity, acoustic signals containing many noise components are preferentially extracted, and the set r This can sometimes result in the setting being incorrect. In relative evaluation of periodicity, the component based on noise is canceled out by the index representing the height of periodicity, so this problem does not occur. Also, if the height of periodicity of the acoustic signals x(1),...,x(N) is not very high, then using only absolute evaluation of periodicity will result in the set being incorrect. r The number of elements may become smaller in some cases, but this problem does not occur with relative evaluation of periodicity. On the other hand, if the acoustic signals x(1),...,x(N) contain almost no noise components, absolute evaluation of periodicity may be able to appropriately extract acoustic signals with higher periodicity than relative evaluation of periodicity. Therefore, based on relative and absolute evaluation of periodicity, the set r It is desirable, and more preferably, to identify a set. r It is more desirable to include both elements based on a relative evaluation of periodicity and elements based on an absolute evaluation of periodicity.

[0092] Furthermore, in this embodiment, it is preferable to set up based on the relative evaluation of the magnitude of the acoustic signal. p It is desirable to identify the set (for example, based on relative size). p(Example of identifying a specific acoustic signal). That is, when the periodicity of the reference acoustic signal does not exceed the reference value TH1, it is desirable that the acoustic signal selection device 22 outputs information corresponding to the acoustic signal x(s) (second acoustic signal) extracted from the multiple acoustic signals x(1),...,x(N) based on a relative evaluation of the magnitude of at least one of the multiple acoustic signals x(1),...,x(N) with respect to a magnitude reference based on one of the magnitudes of the multiple acoustic signals x(1),...,x(N). Relative evaluation of the magnitude of acoustic signals makes it possible to perform an evaluation that excludes ambient noise common to all acoustic signals x(1),...,x(N). In other words, relative evaluation of the magnitude of acoustic signals makes it possible to appropriately evaluate noise observed only at a specific acoustic sensor 10. Therefore, when a large amount of noise is observed only at a specific acoustic sensor 10 (for example, when wind directly hits only a specific sensor 10), the acoustic signal x corresponding to such a specific acoustic sensor 10 can be appropriately excluded.

[0093] In the second embodiment, the periodicity determination unit 227a of the selection control unit 227 (Figure 6) determined whether the height of the periodicity represented by the index r(w) exceeded the reference value TH1. However, the periodicity determination unit 227a may also be input to dr(w), and the periodicity determination unit 227a may determine whether the height of the periodicity represented by dr(w) exceeds the reference value TH1.

[0094] In the second embodiment, the switching control unit 227b of the selection control unit 227 determines the decision result d r If this is entered, the set r Output the corresponding integer indices s=Φ(1),…,Φ(K), and the judgment result d p If this is entered, the set p The corresponding integer indices s=Θ(1),…,Θ(Q) were output. However, the switching control unit 227b was output r(w) and / or dr(w)(w∈set r ) is input, and the switching control unit 227b determines the result d r If a set is entered, the set will be determined based on r(w) and / or dr(w). rThe elements Φ(1), ..., Φ(K) may be further narrowed down, and the narrowed-down elements may be output as integer index s. Similarly, the switching control unit 227b may be given p(u) and / or dp(u) (u∈set p ) is input, and the switching control unit 227b determines the result d p If a value is entered, the set is determined based on p(u) and / or dp(u). p Alternatively, we can further narrow down the elements Θ(1), ..., Θ(Q) and output the narrowed-down elements as integer index s.

[0095] In the second embodiment, the selection control unit 227 of the acoustic signal selection device 22 (Figure 5) sends an integer index s to the acoustic signal processing device 23, and the selection control unit 231 of the acoustic signal processing device 23 (Figure 7) uses the acoustic signals x(1), ..., x(N) of each frame and the integer index s to extract the acoustic signal x(s) of each frame from the acoustic signals x(1), ..., x(N). However, the acoustic signal selection device 22 (Figure 5) may have a selection control unit 231, and the selection control unit 231 of the acoustic signal selection device 22 may use the acoustic signals x(1), ..., x(N) of each frame and the integer index s to extract and output the acoustic signal x(s) of each frame from the acoustic signals x(1), ..., x(N).

[0096] In short, the acoustic signal selection device 22 should output information corresponding to a first acoustic signal extracted from multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals obtained from multiple acoustic sensors if the periodicity of the reference acoustic signal corresponding to at least one of the multiple acoustic signals exceeds a reference value, and output information corresponding to a second acoustic signal extracted from multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals if the periodicity of the reference acoustic signal does not exceed a reference value. This makes it possible to suppress the intrusion of noise based on the external environment, even when the relationship between the acoustic sensor and the external environment is unknown.

[0097] [Hardware configuration] The functions realized by the components described herein may be implemented in a circuit or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), CPUs (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to realize the functions described herein. A processor is considered a circuit or processing circuitry, including transistors and other circuits. A processor may be a programmed processor that executes a program stored in memory.

[0098] In this specification, circuitry, unit, and means are hardware programmed to perform or execute the functions described herein. Such hardware may be any hardware disclosed herein, or any hardware known to be programmed to perform or execute the functions described herein.

[0099] If the hardware is a processor that is considered to be a type of circuitry, then the circuitry, means, or unit is a combination of hardware and software used to constitute the hardware and / or processor.

[0100] For example, the acoustic signal selection devices 12, 22 and acoustic signal processing devices 13, 23 in each embodiment are devices configured by a general-purpose or dedicated computer equipped with a processor (hardware processor) such as a CPU (central processing unit) and memory such as RAM (random-access memory) and ROM (read-only memory) executing a predetermined program. That is, the acoustic signal selection devices 12, 22 and acoustic signal processing devices 13, 23 in each embodiment have, for example, processing circuits configured to implement each of their respective parts. This computer may have one processor and memory, or it may have multiple processors and memories. This program may be installed on the computer, or it may be pre-recorded in ROM, etc. Furthermore, some or all of the processing units may be configured using electronic circuits that realize processing functions independently, rather than electronic circuits that realize the functional configuration by loading a program, such as a CPU. Also, the electronic circuits that constitute one device may include multiple CPUs.

[0101] Figure 9 is a block diagram illustrating the hardware configuration of the acoustic signal selection devices 12, 22 and acoustic signal processing devices 13, 23 in each embodiment. As illustrated in Figure 9, the acoustic signal selection devices 12, 22 and acoustic signal processing devices 13, 23 in this example have a CPU (Central Processing Unit) 10a, an input unit 10b, an output unit 10c, a RAM (Random Access Memory) 10d, a ROM (Read Only Memory) 10e, an auxiliary storage device 10f, a communication unit 10h, and a bus 10g. The CPU 10a in this example has a control unit 10aa, an arithmetic unit 10ab, and a register 10ac, and performs various arithmetic processing according to various programs loaded into the register 10ac. The input unit 10b is an input terminal, keyboard, mouse, touch panel, etc., to which data is input. The output unit 10c is an output terminal, display, etc., to which data is output. The communication unit 10h is a LAN card, etc., controlled by the CPU 10a which has loaded a predetermined program. Furthermore, RAM 10d is an SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), etc., and has a program area 10da where a predetermined program is stored and a data area 10db where various data is stored. Furthermore, auxiliary storage device 10f is, for example, a hard disk, MO (Magneto-Optical disc), semiconductor memory, etc., and has a program area 10fa where a predetermined program is stored and a data area 10fb where various data is stored. Furthermore, bus 10g connects CPU 10a, input unit 10b, output unit 10c, RAM 10d, ROM 10e, communication unit 10h, and auxiliary storage device 10f so that information can be exchanged. CPU 10a writes the program stored in the program area 10fa of auxiliary storage device 10f to the program area 10da of RAM 10d according to the loaded OS (Operating System) program. Similarly, CPU 10a writes various data stored in the data area 10fb of auxiliary storage device 10f to the data area 10db of RAM 10d.The addresses on RAM 10d where the program and data are written are then stored in register 10ac of CPU 10a. The control unit 10aa of CPU 10a sequentially reads these addresses stored in register 10ac, reads the program and data from the area on RAM 10d indicated by the read addresses, sequentially has the arithmetic unit 10ab execute the calculations indicated by the program, and stores the calculation results in register 10ac. This configuration realizes the functional configuration of the acoustic signal selection devices 12, 22 and the acoustic signal processing devices 13, 23.

[0102] The program describing this process can be recorded on a computer-readable recording medium. Examples of computer-readable recording media are non-transitory recording media. Examples of such recording media include magnetic recording devices, optical discs, magneto-optical recording media, and semiconductor memory.

[0103] Furthermore, this program may be distributed, for example, by selling, transferring, or lending portable recording media such as DVDs or CD-ROMs on which the program is recorded. Alternatively, the program may be stored in the storage device of a server computer and distributed by transferring the program from the server computer to other computers via a network.

[0104] A computer executing such a program may, for example, first store the program recorded on a portable storage medium or a program transferred from a server computer in its own storage device. Then, when processing is to be executed, the computer reads the program stored on its own storage medium and executes the processing according to the read program. Alternatively, the computer may directly read the program from the portable storage medium and execute the processing according to that program, or it may sequentially execute the processing according to the received program each time a program is transferred to it from a server computer. Furthermore, the processing may be executed by a so-called ASP (Application Service Provider) type service, where the processing function is realized only by execution instructions and result acquisition, without transferring the program from the server computer to this computer. Furthermore, the processing may be executed using a so-called SaaS (Software as a Service) type service, where a part of the server computer is made available to the user along with the program. In this form, the program includes information used for processing by an electronic computer that is equivalent to a program (data that is not a direct instruction to the computer but has the property of defining the computer's processing).

[0105] Furthermore, in this configuration, the device is configured by executing a predetermined program on a computer, but at least a part of these processes may be implemented in hardware.

[0106] [Other variations] It should be noted that the present invention is not limited to the embodiments described above. For example, at least one of the periodicity comparison unit 124, size comparison unit 125, periodicity comparison unit 224, size comparison unit 225, and selection control units 127, 227 may obtain an output value by applying the input value to a function, rather than selecting an output value from the input value by comparison.

[0107] In the embodiment described above, the integer index corresponding to the acoustic sensor 10-n was set to n. However, instead of the integer index n corresponding to the acoustic sensor 10-n, other information corresponding to the acoustic sensor 10-n may be set to n.

[0108] In the above embodiment, the AD converter 11 output an acoustic signal x(n) based on the input observation signal α(n). However, the function of the AD converter 11 may be included in the acoustic sensor 10. For example, if digital acoustic sensors with built-in AD converters in each acoustic sensor 10-n are used, the AD converter 11 can be omitted. In this case, each acoustic sensor 10-n outputs an acoustic signal x(n). Alternatively, the function of the AD converter 11 may be included in the acoustic signal selection devices 12, 22. In such cases as well, the AD converter 11 can be omitted.

[0109] Furthermore, the various processes described above may not only be executed sequentially as described, but may also be executed in parallel or individually as needed, depending on the processing capacity of the device performing the processes. It goes without saying that other modifications can be made as appropriate without departing from the spirit of the present invention.

[0110] [Note] The features described in the above-mentioned embodiments are summarized below. [Note 1] An acoustic signal selection device that, when the height of the periodicity of a reference acoustic signal corresponding to at least one of a plurality of acoustic signals obtained from multiple acoustic sensors exceeds a reference value, outputs information corresponding to one or more first acoustic signals extracted from the plurality of acoustic signals based on the periodicity of at least one of the plurality of acoustic signals; and when the height of the periodicity of the reference acoustic signal does not exceed the reference value, outputs information corresponding to one or more second acoustic signals extracted from the plurality of acoustic signals based on the magnitude of at least one of the plurality of acoustic signals. [Note 2] The acoustic signal selection device described in Appendix 1, When the periodicity of the reference acoustic signal exceeds the reference value, information corresponding to the first acoustic signal whose periodicity exceeds the average among the plurality of acoustic signals is output. An acoustic signal selection device that outputs information corresponding to a second acoustic signal whose magnitude is below average among the plurality of acoustic signals, when the periodicity of the reference acoustic signal does not exceed the reference value. [Note 3] The acoustic signal selection device described in Appendix 2, The acoustic signal whose periodicity exceeds the average is the acoustic signal with the highest periodicity. An acoustic signal selection device in which the aforementioned acoustic signal with a magnitude below average is the acoustic signal with the smallest magnitude. [Note 4] The acoustic signal selection device described in Appendix 1, An acoustic signal selection device that, when the periodicity of the reference acoustic signal exceeds the reference value, outputs information corresponding to a plurality of first acoustic signals, extracted from the plurality of acoustic signals based on a relative evaluation of at least one of the periodicities of the plurality of acoustic signals with respect to a periodicity reference based on any one of the periodicities of the plurality of acoustic signals, and an absolute evaluation of at least one of the periodicities of the plurality of acoustic signals. [Note 5] The acoustic signal selection device described in Appendix 4, An acoustic signal selection device in which the plurality of first acoustic signals include an acoustic signal extracted based on the relative evaluation and an acoustic signal extracted based on the absolute evaluation. [Note 6] An acoustic signal selection device as specified in Appendix 4 or 5, An acoustic signal selection device that outputs information corresponding to the second acoustic signal, extracted from the plurality of acoustic signals based on a relative evaluation of at least one of the magnitudes of the plurality of acoustic signals with respect to a magnitude reference based on one of the magnitudes of the plurality of acoustic signals, when the periodicity of the reference acoustic signal does not exceed the reference value. [Note 7] An acoustic signal selection method comprising: outputting information corresponding to one or more first acoustic signals extracted from the multiple acoustic signals based on the periodicity of at least one of the multiple acoustic signals obtained from multiple acoustic sensors when the height of the periodicity of a reference acoustic signal corresponding to at least one of the multiple acoustic signals exceeds a reference value; and outputting information corresponding to one or more second acoustic signals extracted from the multiple acoustic signals based on the magnitude of at least one of the multiple acoustic signals when the height of the periodicity of the reference acoustic signal does not exceed the reference value. [Industrial applicability]

[0111] The present invention can be used, for example, in acoustic signal processing based on observed signals observed by microphones in devices such as smartphones and tablet devices. For example, smartphones and tablet devices may have multiple microphones, but the amount of noise mixed in the observed signal may differ depending on the relationship between the microphones and the external environment. For example, even if wind is blowing on the device, depending on the relative position of the wind direction and the microphones, the mounting position of the microphones in the device housing, etc., it is not necessarily the case that the same amount of noise is mixed into all microphones at the same time. For example, if microphones are arranged on the top and bottom sides of the housing, the microphone on the bottom side will be less affected by the wind if it is blowing from above, and the microphone on the top side will be less affected by the wind if it is blowing from below. Also, when the device is held by hand, it is rare for the hand to touch all microphones at the same time, and it is thought that some microphones will be affected by noise due to hand contact, while others will not be affected by noise. The type of microphone that will be affected by wind or contact noise changes constantly depending on the relationship between the microphone and the external environment, such as the relative relationship between the wind direction and the device, and how it is held, making it conventionally difficult to estimate. Therefore, when acoustic signal processing is performed based on an observation signal that is affected by wind or contact noise, problems arise such as the inclusion of unpleasant noise in the acoustic signal processing result, or incorrect estimation of the sound source, leading to the amplification or suppression of the wrong signal. In this invention, since noise contamination based on the external environment can be suppressed without using the relationship between the microphone and the external environment, such noise contamination can be suppressed. As a result, problems such as unpleasant noise being included in the acoustic signal processing result, or incorrect estimation of the sound source, leading to the amplification or suppression of the wrong signal can be solved. Furthermore, this invention can also be used as a countermeasure when some of the microphones among multiple microphones malfunction and cause noise contamination. [Explanation of Symbols]

[0112] 12,22 Acoustic signal selection device 13,23 Acoustic signal processing device

Claims

1. An acoustic signal selection device that, when the height of the periodicity of a reference acoustic signal corresponding to at least one of a plurality of acoustic signals obtained from a plurality of acoustic sensors exceeds a reference value, outputs information corresponding to one or more first acoustic signals extracted from the plurality of acoustic signals based on the periodicity of at least one of the plurality of acoustic signals; and when the height of the periodicity of the reference acoustic signal does not exceed the reference value, outputs information corresponding to one or more second acoustic signals extracted from the plurality of acoustic signals based on the magnitude of at least one of the plurality of acoustic signals.

2. An acoustic signal selection device according to claim 1, When the periodicity of the reference acoustic signal exceeds the reference value, information corresponding to the first acoustic signal whose periodicity extracted from the plurality of acoustic signals exceeds the average is output. An acoustic signal selection device that outputs information corresponding to a second acoustic signal whose magnitude is below average when the periodicity of the reference acoustic signal does not exceed the reference value.

3. An acoustic signal selection device according to claim 1, An acoustic signal selection device that, when the periodicity of the reference acoustic signal exceeds the reference value, outputs information corresponding to a plurality of first acoustic signals, extracted from the plurality of acoustic signals based on a relative evaluation of at least one of the periodicities of the plurality of acoustic signals with respect to a periodicity reference based on any one of the periodicities of the plurality of acoustic signals, and an absolute evaluation of at least one of the periodicities of the plurality of acoustic signals.

4. The acoustic signal selection device according to claim 3, An acoustic signal selection device that outputs information corresponding to the second acoustic signal, extracted from the plurality of acoustic signals based on a relative evaluation of at least one of the magnitudes of the plurality of acoustic signals with respect to a magnitude reference based on one of the magnitudes of the plurality of acoustic signals, when the periodicity of the reference acoustic signal does not exceed the reference value.

5. A program for causing a computer to function as an acoustic signal selection device according to any of claims 1 to 4.

Images