JP5494085B2 - Sound processor - Google Patents

Description

  The present invention relates to a technique for emphasizing a component (hereinafter referred to as a “localization component”) in which a sound image is localized at a target position (direction) among acoustic signals of stereo right and left channels.

  Conventionally, a technique for suppressing a predetermined localization component from an acoustic signal recorded on a recording medium such as a CD has been proposed. For example, Patent Document 1 discloses a technique for suppressing a component (typically a singing sound) in which a sound image is localized in the center direction by subtracting one of stereo audio signals from the other (reverse phase addition). ing.

JP 2007-079413 A

  If the acoustic signal in which the localization component is suppressed by the technique of Patent Document 1 is subtracted from the original acoustic signal, an acoustic signal in which the localization component is emphasized can be generated. However, under the technique of Patent Document 1, there is a problem that the processed acoustic signal is in a monaural format. In view of the above circumstances, an object of the present invention is to generate a stereo-type acoustic signal in which a localization component is emphasized.

  In order to solve the above problems, an acoustic processing device of the present invention suppresses a localization component in a specific direction and sum component generation means for generating a sum component of a stereo first audio signal and a second audio signal. Difference component generation means for generating a difference component by subtraction between the first acoustic signal and the second acoustic signal, determination means for determining the presence / absence of a localization component, and a coefficient value for each frequency so that the localization component is emphasized Is subtracted between the sum component and the difference component (for example, calculation of Equation (5A), Equation (7A), Equation (5C) described later) A first generation means generated by the subtraction processing including the second processing coefficient sequence (the distribution of coefficient values at each frequency is level compared to the first processing coefficient sequence generated by the subtraction processing when no localization component is present ( For example, the second generation means for generating the processing coefficient sequence GB), the first acoustic signal and the second acoustic signal Each coefficient value of the first processing coefficient sequence is applied when the determination unit determines that a localization component is present for each frequency component of the second component, and the second process is performed when the determination unit determines that no localization component exists. Signal processing means for operating each coefficient value of the coefficient sequence.

  In the above configuration, since the processing coefficient sequence is applied to each of the first acoustic signal and the second acoustic signal, it is possible to generate a stereo acoustic signal in which the localization component is emphasized. When the localization component does not exist, the second processing coefficient sequence in which the distribution of the coefficient values is leveled compared to the first processing coefficient sequence is applied to the processing in the signal processing means. Therefore, compared to a configuration in which the first processing coefficient sequence is applied regardless of the presence or absence of the localization component, noise (musical noise) due to the application of the first processing coefficient sequence even when the localization component does not exist is reduced. Is possible.

  When the localization component is present, the coefficient values of the first processing coefficient sequence are distributed over a wide range, and when the localization component is not present, the coefficient values of the first processing coefficient sequence tend to be unevenly distributed in the vicinity of the predetermined value. . Therefore, the determination unit according to a preferred aspect of the present invention is configured when the average value or the degree of dispersion (for example, variance or standard deviation) of the plurality of coefficient values of the first processing coefficient sequence generated by the first generation unit exceeds a threshold value. Determines that the localization component is present, and determines that the localization component does not exist if the localization component is below the threshold. In the above aspect, since the first processing coefficient sequence that can be applied to the processing in the signal processing means is used for the determination of the presence / absence of the localization component, it is compared with the configuration in which the independent element for determining the presence / absence of the localization component is installed. Therefore, there is an advantage that the configuration and operation of the sound processing apparatus are simplified. However, the configuration using the first processing coefficient sequence for the determination of the presence or absence of the localization component is not essential in the present invention.

  The method of generating the second processing coefficient sequence by the second generation means is arbitrary. For example, a second processing coefficient sequence in which each coefficient value is set to a predetermined value is generated (for example, a second processing coefficient prepared in advance). A method of acquiring a sequence from a storage unit) and a method of generating a second processing coefficient sequence by smoothing the distribution of coefficient values of the first processing coefficient sequence generated by the first generation unit. According to the former configuration, an effect that the processing and configuration for generating the second processing coefficient sequence is simplified is realized, and according to the latter configuration, the characteristics of the first acoustic signal and the second acoustic signal are improved. The effect of being able to generate an acoustically natural acoustic signal is realized by being reflected in the second processing coefficient sequence.

  In a preferred aspect of the present invention, the determination means determines the presence / absence of a localization component for each unit period, and the signal processing means performs the second operation when the presence of the localization component is continuously denied over a predetermined number of unit periods. Start applying the processing coefficient sequence. In the above aspect, since the application of the second processing coefficient sequence is started when the presence of the localization component is continuously denied over a predetermined number of unit periods, the processing coefficient applied to the processing in the signal processing means The frequency with which the column is changed from one of the first processing coefficient column and the second processing coefficient column to the other decreases. Therefore, there is an advantage that an acoustic signal with an acoustically natural impression can be generated.

  The sound processing apparatus according to a preferred aspect of the present invention includes variable setting means for variably setting a localization variable indicating a specific position (direction), and the difference component generation means is responsive to the localization variable set by the variable setting means. The other is subtracted from one of the first acoustic signal and the second acoustic signal at a predetermined ratio. In the above aspect, the ratio of the first acoustic signal and the second acoustic signal at the time of subtraction by the difference component generation means is variably set according to the localization variable. Therefore, there is an advantage that the position (direction) of the localization component to be emphasized can be variably controlled (not limited to the central direction).

  The acoustic processing device according to each of the above aspects is realized by hardware (electronic circuit) such as a DSP (Digital Signal Processor) dedicated to processing of an acoustic signal, or a general-purpose calculation such as a CPU (Central Processing Unit). It is also realized by cooperation between the processing device and a program (software). The program of the present invention includes a sum component generation process for generating a sum component of a stereo first sound signal and a second sound signal, and a difference component obtained by suppressing a localization component in a specific direction as a first sound signal and a second sound. A difference component generation process generated by subtraction with a signal, a determination process for determining the presence or absence of a localization component, and a first processing coefficient sequence in which a coefficient value for each frequency is set so that the localization component is emphasized, The coefficient value of each frequency compared with the first generation process generated by the subtraction process including subtraction between the sum component and the difference component and the first process coefficient sequence generated by the subtraction process when no localization component exists Second generation processing for generating a second processing coefficient sequence with a uniform distribution of the first and second frequency components when the determination component determines that a localization component exists for each frequency component of the first acoustic signal and the second acoustic signal. Each coefficient value of one processing coefficient sequence is applied, and the localization component exists. Not the determination process when it is determined the signal processing and the action of each coefficient value of the second processing coefficient sequence to be executed by a computer in. According to the above program, the same operation and effect as the sound processing apparatus of the present invention are realized. The program of the present invention is provided to a user in a form stored in a computer-readable recording medium and installed in the computer, or provided from a server device in a form of distribution via a communication network and installed in the computer. Is done.

1 is a block diagram of a sound processing apparatus according to a first embodiment. It is a block diagram of a coefficient setting part. It is explanatory drawing of operation | movement of a coefficient setting part. It is explanatory drawing of operation | movement of the 2nd production | generation part in 2nd Embodiment.

<A: First Embodiment>
FIG. 1 is a block diagram of a sound processing apparatus 100 according to the first embodiment of the present invention. A signal supply device 12, a sound emission device 14, and an input device 16 are connected to the sound processing device 100. The signal supply device 12 supplies the sound processing device 100 with time-domain sound signals SIN (SIN_L, SIN_R) representing the waveform of sound (speech and music).

  The sound signal SIN_L of the left channel and the sound signal SIN_R of the right channel are localized at positions where the sound images of a plurality of sound sources generating sound are different (that is, the sound amplitude and phase differ depending on the position of each sound source). In this way, a stereo signal is collected or processed. A sound collection device (stereo microphone) that collects ambient sound to generate an acoustic signal SIN, a playback device that acquires the acoustic signal SIN from a portable or built-in recording medium, and outputs it to the acoustic processing device 100; A communication device that receives the acoustic signal SIN from the communication network and outputs the acoustic signal SIN to the acoustic processing device 100 may be employed as the signal supply device 12.

  The acoustic processing device 100 generates an acoustic signal SOUT (SOUT_L, SOUT_R) from the acoustic signal SIN supplied by the signal supply device 12. The left-channel acoustic signal SOUT_L and the right-channel acoustic signal SOUT_R are stereo-format signals in which the localization component in which the sound image is localized at a target position (direction) among the sounds represented by the acoustic signal SIN. The sound emitting device 14 (for example, a stereo speaker or a stereo headphone) emits a sound wave corresponding to the acoustic signal SOUT (SOUT_L, SOUT_R) generated by the acoustic processing device 100.

  The input device 16 is a device (for example, a mouse or a keyboard) for a user to input an instruction to the sound processing device 100. The user can arbitrarily instruct the position (direction) of the localization component to the sound processing device 100 by appropriately operating the input device 16.

  As shown in FIG. 1, the sound processing device 100 is realized by a computer system including an arithmetic processing device 22 and a storage device 24. The storage device 24 stores a program PG executed by the arithmetic processing device 22 and data used by the arithmetic processing device 22. A known recording medium such as a semiconductor recording medium or a magnetic recording medium or a combination of a plurality of types of recording media is arbitrarily adopted as the storage device 24. A configuration in which the acoustic signal SIN (SIN_L, SIN_R) is stored in the storage device 24 (therefore, the signal supply device 12 can be omitted) is also suitable.

  The arithmetic processing unit 22 executes a program PG stored in the storage device 24 to thereby generate a plurality of functions (variable setting unit 32, frequency analysis unit 34, coefficient setting unit) for generating the acoustic signal SOUT from the acoustic signal SIN. 36, a signal processing unit 38, a waveform synthesis unit 40, and a determination unit 42). A configuration in which each function of the arithmetic processing unit 22 is distributed over a plurality of integrated circuits, or a configuration in which a dedicated electronic circuit (DSP) realizes each function may be employed.

  The variable setting unit 32 sets the localization variable α to be variable. The localization variable α is a variable that designates the position (direction) of the sound image of the localization component. The variable setting unit 32 of the present embodiment variably sets the localization variable α (0 ≦ α ≦ 1) in accordance with an instruction from the user to the input device 16. For example, when the localization variable α is 0.5 (median value), the center (front) direction is indicated. Further, as the localization variable α is closer to 1, a rightward direction is indicated, and as the localization variable α is closer to 0, a leftward direction is indicated. The user can change the localization variable α by operating the input device 16 at any time while listening to the reproduced sound from the sound emitting device 14.

  The frequency analysis unit 34 sequentially generates a frequency spectrum (complex spectrum) LA of the acoustic signal SIN_L and a frequency spectrum (complex spectrum) RA of the acoustic signal SIN_R for each unit period (frame) on the time axis. The frequency spectrum LA is a series (LA [1] to LA [K]) of frequency components LA [k] corresponding to each of K frequencies (frequency bands) f1 to fK (k = 1 to K). Similarly, the frequency spectrum RA is a series of frequency components RA [k] (RA [1] to RA [K]) corresponding to each of the frequencies f1 to fK. For the generation of the frequency spectrum LA and the frequency spectrum RA, a known frequency analysis such as a short-time Fourier transform may be arbitrarily employed. A filter bank composed of K band pass filters having different pass bands can also be adopted as the frequency analysis unit 34.

  The coefficient setting unit 36 in FIG. 1 generates a processing coefficient sequence G (GA, GB) used for emphasizing the localization component of the acoustic signal SIN (SIN_L, SIN_R). The processing coefficient sequence G is a series of g [k] (g [1] to g [K]) corresponding to each of the frequencies f1 to fK, and uses the frequency spectrum LA and the frequency spectrum RA for each unit period. Are generated sequentially. The coefficient value g [k] corresponds to a gain (spectral gain) for the frequency component LA [k] of the acoustic signal SIN_L and the frequency component RA [k] of the acoustic signal SIN_R, and is set within a range of 0 or more and 1 or less. (0 ≦ g [k] ≦ 1). The distinction between the processing coefficient sequence GA and the processing coefficient sequence GB will be described later.

  The signal processing unit 38 in FIG. 1 applies the processing coefficient sequence G (GA, GB) generated by the coefficient setting unit 36 to each of the frequency spectrum LA and the frequency spectrum RA, so that the frequency spectrum LB and the frequency spectrum RB Are sequentially generated for each unit period. The frequency spectrum LB is a series of frequency components LB [k] corresponding to each frequency fk (LB [1] to LB [K]), and the frequency spectrum RB is a frequency component RB [k] corresponding to each frequency fk. (RB [1] to RB [K]).

Specifically, as shown in the following formulas (1A) and (1B), the signal processing unit 38 applies a coefficient setting unit for each unit period to the frequency spectrum LA and the frequency spectrum RA of each unit period. The common processing coefficient sequence G (GA, GB) generated by 36 is multiplied. That is, the frequency component LB [k] of each frequency fk of the frequency spectrum LB is set to the product of the frequency component LA [k] of the frequency fk and the coefficient value g [k] (formula (1A)), and the frequency Each frequency component RB [k] of the spectrum RB is set to a product of the frequency component RA [k] and the coefficient value g [k] (formula (1B)).

  The waveform synthesizer 40 in FIG. 1 generates a stereo acoustic signal SOUT_L and an acoustic signal SOUT_R from the frequency spectrum LB and the frequency spectrum RB processed by the signal processor 38. Specifically, the waveform synthesizer 40 generates the acoustic signal SOUT_L by converting the frequency spectrum LB for each unit period into a time domain signal by inverse Fourier transform and connecting the unit periods before and after. Similarly, the waveform synthesizer 40 generates an acoustic signal SOUT_R from each frequency spectrum RB. The acoustic signals SOUT (SOUT_L, SOUT_R) generated by the waveform synthesis unit 40 are supplied to the sound emitting device 14 and reproduced as sound waves.

  As described above, in the first embodiment, since the processing coefficient sequence G is individually applied to each of the acoustic signal SIN_L (frequency spectrum LA) and the acoustic signal SIN_R (frequency spectrum RA), the localization component is emphasized. The generated acoustic signal SOUT (SOUT_L, SOUT_R) can be generated in a stereo format.

Next, the details of the coefficient setting unit 36 will be described with reference to FIG. As shown in FIG. 2, the coefficient setting unit 36 includes a sum component generation unit 52, a difference component generation unit 54, and a coefficient sequence generation unit 60. The sum component generation unit 52 adds the frequency components LA [1] to LA [K] of the acoustic signal SIN_L and the frequency components RA [1] to RA [K] of the acoustic signal SIN_R sequentially for each unit period. (Complex spectrum) M is generated. The sum component M is a sequence (M [1] to M [K]) of frequency components M [k] corresponding to the frequencies f1 to fK. As shown in Equation (2), the frequency component M [k] of the frequency fk of the sum component M corresponds to the addition of the frequency component LA [k] and the frequency component RA [k] of the frequency fk. Therefore, the sum component M corresponds to a monaural signal in which sounds from all sound sources are mixed. Note that a configuration in which a weighted sum or arithmetic average of the frequency component LA [k] and the frequency component RA [k] is calculated as the sum component M may be employed.

The difference component generator 54 in FIG. 2 performs subtraction between the frequency components LA [1] to LA [K] of the acoustic signal SIN_L and the frequency components RA [1] to RA [K] of the acoustic signal SIN_R for each unit period. The difference component (complex spectrum) S is generated sequentially. The difference component S is a series (S [1] to S [K]) of frequency components S [k] corresponding to the frequencies f1 to fK. The difference component generator 54 calculates the frequency components S [1] to S [K] by the calculation (weighted subtraction) of Expression (3) using the localization variable α set by the variable setting unit 32.

  As understood from Equation (3), the difference component is obtained by subtracting the frequency component RA [k] from the frequency component LA [k] by a variable ratio (weight value) corresponding to the localization variable α (reverse phase addition). Each frequency component S [k] of S is generated. Therefore, the difference component S is a signal in which the localization component in the position (direction) corresponding to the localization variable α in the acoustic signal SIN (SIN_L, SIN_R) is relatively suppressed with respect to other components (that is, other than the localization component). Relatively emphasized signal). For example, when the localization variable α is 0.5 (median value), a difference component S in which the localization component in the central direction is suppressed is generated. Further, as the localization variable α exceeds 0.5, the localization component closer to the center direction is suppressed by the difference component S, and as the localization variable α is less than 0.5, the localization component toward the left side relative to the center direction is suppressed. However, the difference component S is suppressed.

  The symbol max (α, 1-α) in Equation (3) means the maximum value of the localization variable α or the variable (1-α). Since the localization variable α is set to a numerical value of 1 or less, there is a possibility that the power (amplitude) of the frequency component S [k] is insufficient only by the calculation of the numerator of Equation (3). Dividing by the maximum value max (α, 1−α) as in Equation (3) maintains the power of the frequency component S [k] equal to the frequency component LA [k] and the frequency component RA [k]. Because.

  The coefficient sequence generation unit 60 in FIG. 2 uses the sum component M generated by the sum component generation unit 52 and the difference component S generated by the difference component generation unit 54 to generate the processing coefficient sequence G (GA, GB) as a unit period. Generate every time. As shown in FIG. 2, the coefficient sequence generation unit 60 includes a first generation unit 62 that generates a processing coefficient sequence GA and a second generation unit 64 that generates a processing coefficient sequence GB.

The processing coefficient sequence GA is a numerical sequence for emphasizing the localization component of the acoustic signal SIN (SIN_L, SIN_R). The first generation unit 62 calculates each coefficient value g [k] of the processing coefficient sequence GA by the calculation of the following formula (4). Note that in equation (4), the amplitude P [k] ^1/2 is divided by the amplitude | M [k] | of the frequency component M [k] to make each coefficient value g [k] of the processing coefficient sequence GA 1 This is for normalization to the following numerical values (0 ≦ g [k] ≦ 1).

数式(5A)から理解されるように、周波数成分Ｍ[k]のパワー|Ｍ[k]|^２が周波数成分Ｓ[k]のパワー|Ｓ[k]|^２を上回る周波数ｆkでのパワーＰ[k]は、和成分Ｍのパワー|Ｍ[k]|^２から差成分Ｓのパワー|Ｓ[k]|^２を減算した数値に設定される。すなわち、パワースペクトルＰは、和成分Ｍと差成分Ｓとの間のパワースペクトルの減算（スペクトル減算）で生成される。他方、パワー|Ｍ[k]|^２がパワー|Ｓ[k]|^２を下回る周波数ｆkでのパワーＰ[k]は、和成分Ｍのパワー|Ｍ[k]|^２と所定の係数（フロアリング係数）βとの乗算値に設定される。 The symbol P [k] in Equation (4) means the power at the frequency fk in the power spectrum P in which the localization component is emphasized. The first generation unit 62 calculates the power P [k] of the equation (4) by an operation expressed by the following equations (5A) and (5B) (hereinafter referred to as “subtraction process”).

As understood from the equation (5A), the power P at the frequency fk in which the power | M [k] | ² of the frequency component M [k] exceeds the power | S [k] | ² of the frequency component S [k]. [k] is set to a value obtained by subtracting the power | S [k] | ² of the difference component S from the power | M [k] | ² of the sum component M. That is, the power spectrum P is generated by subtraction (spectrum subtraction) of the power spectrum between the sum component M and the difference component S. On the other hand, the power | M [k] | ² is the power | S [k] | Power P [k] of ² at below frequencies fk, the power of the sum component M | M [k] | ² and a predetermined coefficient (Floor It is set to the product of the ring coefficient (β).

As can be understood from the above description, the subtraction processing expressed by the equations (5A) and (5B) assumes that the sum component M is a signal component (target sound) and the difference component S is a noise component. This corresponds to spectral subtraction (SS) for suppressing noise components. Note that the configuration and to control the coefficient β in Equation (5B) variable, the power of the equation (5B) | is also employed ² was replaced by construction | M [k] | ² the difference component S of the power | S [k] obtain.

By the way, in the acoustic signal SIN (SIN_L, SIN_R), there are a section where the localization component exists and a section where the localization component does not exist (section where the sound generation of the localization component stops). Part (A) of FIG. 3 is a schematic diagram of the processing coefficient sequence GA generated by the first generation unit 62 for the unit period in which the localization component exists, and part (B) of FIG. 3 is a unit in which the localization component does not exist. It is a mimetic diagram of processing coefficient sequence GA which the 1st generating part 62 generates about a period. The predetermined value b of the part (A) and the part (B) in FIG. 3 is a predetermined numerical value (for example, b = β ^1/2 , 0 ≦ b ≦ 1) corresponding to the coefficient β of the equation (5B).

For frequencies fk that localization component is present in a significant power, the power of the sum component M | arithmetic operation formula with ² to exceed (5A) | S [k] | M [k] | power of ² is the difference component S Likely to be executed. The difference component S applied to the equation (5A) is a component in which the localization component is relatively suppressed (a component in which other than the localization component is emphasized with respect to the localization component), and thus the power P calculated by the equation (5A) The series [k] corresponds to a power spectrum in which the localization component of the acoustic signal SIN (SIN_L, SIN_R) is emphasized relative to other components. Therefore, as shown in part (A) of FIG. 3, when there is a localization component, each coefficient value g [k] of the processing coefficient sequence GA calculated by equation (4) is a frequency at which the localization component power is large. On the frequency axis, the coefficient value g [k] of fk is closer to 1, and the coefficient value g [k] of the frequency fk having a smaller localization component power is closer to a predetermined value b (a value closer to zero). Distributed by. Accordingly, the signal processing unit 38 multiplies each of the frequency spectrum LA and the frequency spectrum RA by the processing coefficient sequence GA, thereby generating a stereo-type acoustic signal SOUT (SOUT_L, SOUT_R) in which the localization component is emphasized.

On the other hand, in the unit period in which the localization component does not exist, the amount of suppression in the calculation of the difference component S in Equation (3) is small, so the power | M [k] | ^{2 of} the sum component M is the power | S [ The number of frequencies fk exceeding k] | ² decreases compared to the case where the localization component is present. That is, the frequency fk at which the power P [k] is set in Expression (5B) instead of Expression (5A) increases. Therefore, as shown in part (B) of FIG. 3, each coefficient value g [k] of the processing coefficient sequence GA calculated when no localization component is present is a numerical value close to the predetermined value b over the entire range of the frequency axis. The distribution is flat so that (a value close to zero).

However, even when there is no localization component, the frequency fk in which the power | M [k] | ^{2 of} the sum component M exceeds the power | S [k] | ² of the difference component S (that is, the frequency to which the formula (5A) is applied). fk) may exist, and therefore the local peak where the coefficient value g [k] is a high numerical value (the numerical value calculated by Equation (5A)) is present in the distribution of the processing coefficient sequence GA when no localization component is present. p is scattered on the frequency axis. Therefore, if the processing coefficient sequence GA is applied to the acoustic signal SIN (SIN_L, SIN_R) even when the localization component does not exist, a high-intensity component corresponding to the peak p of the processing coefficient sequence GA is detected as the acoustic signal SOUT (SOUT_L). , SOUT_R) are scattered on the time axis and the frequency axis, and a problem arises that the listener can perceive them as artificial and annoying musical noise (a noise component called Hulu Hulu). Note that a local peak p of the coefficient value g [k] may also occur in the processing coefficient sequence GA when the localization component is present, but depending on the localization component after enhancement included in the acoustic signal SOUT (SOUT_L, SOUT_R). Since it is masked (auditory masking effect), almost no musical noise is perceived in the reproduced sound of the acoustic signal SOUT.

  In consideration of the above tendency, in the first embodiment, when the localization component is present in the acoustic signal SIN (SIN_L, SIN_R), the processing coefficient sequence GA (part (A) in FIG. 3) is processed by the signal processing unit 38. On the other hand, when the localization component does not exist in the acoustic signal SIN, as shown in the part (C) of FIG. 3, the variation (especially the local peak p) of each coefficient value g [k] The suppressed processing coefficient sequence GB is applied to the processing in the signal processing unit 38. That is, the processing coefficient sequence GB is compared with the processing coefficient sequence GA (part (B) in FIG. 3) generated by the subtraction processing of the equations (5A) and (5B) when no localization component is present. This is a numerical sequence in which the distribution of numerical values g [k] is level (that is, fluctuations are suppressed).

  The determination unit 42 in FIG. 1 determines the presence or absence of a localization component in the acoustic signal SIN (SIN_L, SIN_R). As described above with reference to part (A) and part (B) in FIG. 3, the distribution mode of the coefficient values g [1] to g [K] of the processing coefficient sequence GA is such that a localization component is present. And the case where it does not exist. Therefore, the determination unit 42 determines the presence / absence of a localization component using the processing coefficient sequence GA generated by the first generation unit 62 as illustrated below.

  Specifically, when there is no localization component (part (B) in FIG. 3), many coefficient values g [k] of the processing coefficient sequence GA are numerical values in the vicinity of the predetermined value b, whereas localization is performed. When the component exists (part (A) in FIG. 3), the coefficient value g [k] corresponding to the localization component exceeds the predetermined value b. Therefore, the average value μ of the coefficient values g [1] to g [K] of the processing coefficient sequence GA tends to be larger when the localization component is present than when the localization component is not present. is there. Considering the above tendency, the determination unit 42 of the first embodiment determines that the average value μ of the coefficient values g [1] to g [K] of the processing coefficient sequence GA generated by the first generation unit 62 is a predetermined threshold value. If it exceeds μTH, it is determined that the localization component is present in the acoustic signal SIN, and if the average value μ is less than the threshold value μTH, it is determined that there is no localization component in the acoustic signal SIN. The determination of the presence / absence of a localization component is executed sequentially for each unit period, for example.

  When the determination unit 42 determines that the localization component is not present in the acoustic signal SIN, the second generation unit 64 performs K coefficient values g [1] to g [K as illustrated in a part (C) of FIG. ] Is set to the same value γ (that is, a numerical sequence in which the coefficient values g [k] are evenly distributed). The predetermined value γ is set within a range of 0 or more and 1 or less (0 ≦ γ ≦ 1). For example, the predetermined value γ is set to a predetermined value b (for example, γ = b = 0.1) corresponding to the coefficient β in the equation (5B).

  For each of the frequency spectrum LA and the frequency spectrum RA of the unit period determined by the determination unit 42 that the localization component is present, the signal processing unit 38 selects each coefficient value g [of the processing coefficient sequence GA generated by the first generation unit 62. k] is applied. On the other hand, the signal processing unit 38 applies the processing coefficient sequence GB generated by the second generation unit 64 to each of the frequency spectrum LA and the frequency spectrum RA of the unit period determined by the determination unit 42 that the localization component does not exist. . That is, since the coefficient values g [1] to g [K] of the processing coefficient sequence GB are set to small numerical values, the volume over the entire band (f1 to fK) of the acoustic signal SOUT for the unit period in which the localization component does not exist. Decreases compared to the presence of a stereotaxic component.

  In the above embodiment, when a localization component exists, the processing coefficient sequence GA in which each coefficient value g [k] is set so that the localization component is emphasized is applied to the processing in the signal processing unit 38, and the localization component is Is not applied, the processing coefficient sequence GB in which the coefficient values g [k] are evenly distributed is applied to the processing in the signal processing unit 38. Therefore, it is possible to effectively emphasize the localization component of the acoustic signal SIN and effectively reduce musical noise in each unit period in which no localization component exists in the acoustic signal SOUT.

<B: Second Embodiment>
A second embodiment of the present invention will be described. In addition, about the element which an effect | action and a function are equivalent to 1st Embodiment in each aspect illustrated below, each reference detailed in the above description is diverted and each detailed description is abbreviate | omitted suitably.

  The second generation unit 64 of the first embodiment generates a processing coefficient sequence GB in which the coefficient values g [1] to g [K] are set to a common predetermined value γ. As shown in FIG. 4, the second generation unit 64 of the second embodiment smoothes the distribution of the coefficient values g [1] to g [K] of the processing coefficient sequence GA generated by the first generation unit 62 (low-pass). The processing coefficient sequence GB is generated by performing the filtering process. For example, the coefficient value g [k] of the processing coefficient sequence GB is set to an average value of a predetermined number of coefficient values g [k1] to g [k2] including the coefficient value g [k] in the processing coefficient sequence GA. .

  With the above configuration, the same effect as that of the first embodiment is realized. In the first embodiment, each coefficient value g [k] of the processing coefficient sequence GB is set to a predetermined value γ that is not related to the acoustic signal SIN, whereas in the second embodiment, the processing coefficient sequence GB is also set to the processing coefficient sequence GB. Since the characteristics of the acoustic signal SIN are reflected, there is also an advantage that the acoustic signal SOUT having an acoustically natural impression can be generated.

<C: Modification>
Each of the above forms can be variously modified. Specific modifications are exemplified below. Two or more aspects arbitrarily selected from the following examples can be appropriately combined.

(1) Modification 1
The method by which the determination unit 42 determines the presence / absence of a localization component is arbitrary. For example, a configuration is used in which the degree of dispersion (variance σ ² and standard deviation σ) of the coefficient values g [1] to g [K] of the processing coefficient sequence GA is used to determine the presence or absence of a localization component. That is, when the localization component exists (part (A) in FIG. 3), the coefficient values g [k] of the processing coefficient sequence GA are dispersed over a wide range, whereas the localization component does not exist (in FIG. 3). In part (B), many coefficient values g [k] of the processing coefficient sequence GA are unevenly distributed in the vicinity of the predetermined value b. Therefore, the determination unit 42 determines that the localization component exists when the variance σ ² of the coefficient values g [1] to g [K] of the processing coefficient sequence GA exceeds the threshold σTH, and the variance σ ² has the threshold σTH. If it is below, it is determined that no stereotaxic component is present. It is also possible to replace the variance σ ² in the above example with the standard deviation σ.

  Note that the configuration using the processing coefficient sequence GA for determining the presence or absence of a localization component is not essential. For example, when the localization component is a voice such as a singing sound, a voice detection technique (VAD: voice signal) that divides the acoustic signal SIN into a voice section (section in which voice is present) and a non-voice section (section in which no voice is present). Voice Activity Detection) is used to determine the presence or absence of a localization component. That is, the determination unit 42 divides the acoustic signal SIN into a voice segment and a non-speech segment, affirms the presence of a localization component for each unit period in the speech segment, and is localized for each unit period in the non-speech segment. Denies the existence of the component.

  In the case where the processing coefficient sequence GA is used for the determination of the presence / absence of a localization component as in the first embodiment and the second embodiment, the first generation unit 62 sets the processing coefficient sequence GA regardless of the presence / absence of the localization component. Need to be generated. On the other hand, when the processing coefficient sequence GA is not used for the determination of the presence / absence of the localization component as in the above example using the voice detection technique, the first generation is performed only for the unit period determined by the determination unit 42 that the localization component exists. A configuration may be employed in which the first generation unit 62 does not generate the processing coefficient sequence GA for the unit period in which the unit 62 generates the processing coefficient sequence GA and the determination unit 42 determines that there is no localization component.

(2) Modification 2
In each of the above embodiments, when the determination unit 42 determines that no localization component exists, the second generation unit 64 generates the processing coefficient sequence GB. However, regardless of the presence of the localization component, the second generation unit 64 uses the processing coefficient. A configuration for generating the column GB can also be adopted. If it is determined that the localization component does not exist, the processing coefficient sequence GB is applied to the processing in the signal processing unit 38 as in the above-described embodiments, and if it is determined that the localization component exists, the second generation unit 64. Is generated without being applied to the processing in the signal processing unit 38.

(3) Modification 3
In each of the above embodiments, one of the processing coefficient sequence GA and the processing coefficient sequence GB is selected for each unit period. However, when the determination result by the determination unit 42 is maintained over a plurality of unit periods, the signal processing unit 38 A configuration may be employed in which the processing coefficient sequence G applied to the process is changed from one of the processing coefficient sequence GA and the processing coefficient sequence GB to the other.

  For example, the signal processing unit 38 applies the processing of the mathematical formula (1A) and the mathematical formula (1B) when the existence of the localization component is denied continuously for N unit periods (N is a natural number of 2 or more). When the processing coefficient sequence G is changed from the processing coefficient sequence GA to the processing coefficient sequence GB, and the presence of the localization component is affirmed continuously over N unit periods, the processing coefficient sequence G applied to the processing is changed to the processing coefficient sequence G The GB is changed to the processing coefficient sequence GA. According to the above configuration, since the frequency with which the signal processing unit 38 changes the processing coefficient sequence G applied to the processing between the processing coefficient sequence GA and the processing coefficient sequence GB is reduced, the temporal characteristics of the characteristics of the acoustic signal SOUT are reduced. Fluctuations are mitigated. Therefore, there is an advantage that the acoustic signal SOUT having an acoustically natural impression can be generated. Note that the number N of unit periods, which is a condition for applying the processing coefficient sequence GB, can be variably set. For example, when the localization component is speech, the frequency at which it is determined that the localization component does not exist decreases within the speech interval. Therefore, a configuration in which the number N is increased for each unit period within the speech interval can be employed.

(4) Modification 4
In each of the above embodiments, as understood from the equation (3), the amplitude ratio between the acoustic signal SIN_L (frequency component LA [k]) and the acoustic signal SIN_R (frequency component RA [k]) is the localization variable α. The difference component S was calculated so as to change linearly accordingly, but as illustrated in the following embodiments, the calculation method of the sum component M and the difference component S and the relationship with the localization variable α are appropriately changed. The

<First aspect>
A configuration in which the sum component M and the difference component S are calculated from the power | LA [k] | ² of the acoustic signal SIN_L and the power | RA [k] | ^{2 of the} acoustic signal SIN_R may be employed. Specifically, the sum component generation unit 52 generates a sum component M (complex spectrum composed of frequency components M [1] to M [K]) by calculation of the following formula (6A), and generates a difference component. The unit 54 generates a difference component S (complex spectrum composed of frequency components S [1] to S [K]) by the calculation of the following mathematical formula (6B). Therefore, in the difference component S, the power ratio between the acoustic signal SIN_L and the acoustic signal SIN_R changes linearly according to the localization variable α. Symbols e ^jL in equation (6A) and formula (6B) denotes the phase angle of the frequency spectrum LA, symbol e ^jR means the phase angle of the frequency spectrum RA.

When the sum component M and the difference component S are generated by the above method, the first generation unit 62 of the coefficient sequence generation unit 60 generates a power spectrum P (power P [k]) in which the localization component is emphasized by the following formula ( The coefficient values g [k] of the processing coefficient sequence G are calculated by the calculation of the formula (8) using the power spectrum P and generated by the calculations of 7A) and (7B). With the above configuration, the same effect as that of the first embodiment is realized.

<Second aspect>
A configuration in which the position (direction) of the localization component is controlled according to the function f (α) of the localization variable α can be adopted. Specifically, the sum component generation unit 52 generates the sum component M (frequency components M [1] to M [K]) by the calculation of the following formula (9A), and the difference component generation unit 54 A difference component S (frequency components S [1] to S [K]) is generated by the calculation of Equation (9B). Therefore, the ratio of the amplitudes of the acoustic signal SIN_L and the acoustic signal SIN_R in the difference component S changes according to the function f (α) of the localization variable α. The first generation unit 62 uses the same method (Formula (4), Formula (5A), Formula (5B)) as in the first embodiment, and the sum component M of Formula (9A) and the difference component S of Formula (9B). A processing coefficient sequence GA is generated from the above.

In Formula (5A), the difference between the power | M [k] | ^{2 of the} sum component M and the power | S [k] | ² of the difference component S is calculated as the power P [k]. As shown in 5C), a configuration in which the difference between the amplitude | M [k] | of the sum component M and the amplitude | S [k] | of the difference component S is calculated as the amplitude P [k] may be employed. In the configuration in which the amplitude P [k] is calculated using the formula (5C), the first generation unit 62 calculates the coefficient value g [k] of the processing coefficient sequence GA by the calculation of the following formula (4C).

(5) Modification 5
In each of the above embodiments, each coefficient value g [k] of the processing coefficient sequence GA is calculated by subtraction of power spectrum (formula (5A), formula (7A)) or amplitude spectrum (formula (5C)). A method for generating the processing coefficient sequence GA from the sum component M and the difference component S is arbitrary. For example, if the localization component included in the sum component M is assumed to be a signal component and the difference component S is assumed to be a noise component, the generation of the processing coefficient sequence GA using the sum component M and the difference component S may include a noise component (difference). A known speech enhancement technique for generating a numerical sequence (processing coefficient sequence GA) for emphasizing a signal component (localization component) by suppressing the component S) can be similarly applied.

  Examples of technologies that can be applied to the generation of the processing coefficient sequence GA include speech enhancement using a Wiener filter and speech enhancement using an MMSE-STSA method or a MAP (maximum a posteriori estimation) estimation method. Can be done. For the MMSE-STSA method, see Y. Ephraim and D. Malah, "Speech enhancement using a minimum mean-square error short-time spectral amplitude estimator", IEEE ASSP, vol.ASSP-32, no.6, p.1109- 1121, Dec. 1984, MAP estimation method is described in T. Lotter and P. Vary, "Speech enhancement by MAP spectral amplitude estimation using a Super-Gaussian speech model", EURASIP Journal on Applied Signal Processing, vol.2005 , no, 7, p.1110-1126, July 2005.

(6) Modification 6
The cycle for generating the processing coefficient sequence G (GA, GB) and the cycle for determining whether or not there is a localization component are arbitrary. For example, a configuration in which the coefficient setting unit 36 generates (updates) the processing coefficient sequence G using a plurality of unit periods as the period TA may be employed. Specifically, the processing coefficient sequence G (GA, GB) generated from the acoustic signal SIN of one or more unit periods within each period TA is applied to the processing of each unit period within the period TA. In addition, a configuration in which the determination unit 42 determines the presence / absence of a localization component using a plurality of unit periods as a cycle may be employed.

(7) Modification 7
In each of the above embodiments, the frequency spectrum (LA, RA) of the acoustic signal SIN is multiplied by the processing coefficient sequence G (GA, GB), but the content of the processing by the signal processing unit 38 is the processing coefficient sequence G (GA, GB). It will be changed as appropriate according to the definition and nature. For example, the processing coefficient sequence GA is generated so that the coefficient value g [k] is set to a smaller numerical value within a predetermined range for the frequency fk where the power of the localization component is larger, and each coefficient value g [k] of the processing coefficient sequence GB is set. ] Is set to a large value within the range, the frequency component LA [k] and the frequency component RA [k] are divided by the coefficient values g [k] of the processing coefficient sequence G (GA, GB) ( LB [k] = LA [k] / g [k], RB [k] = RA [k] / g [k]) may also be employed.

(8) Modification 8
A configuration in which localization component emphasis is executed only for components within a predetermined frequency band is also suitable. For example, the processing of each of the above forms is executed only for the frequency band (for example, 100 kHz to 8 kHz) in which the power of human voice is concentrated in the acoustic signal SIN (reproduction is performed without processing for other bands).

DESCRIPTION OF SYMBOLS 100 ... Sound processing device, 12 ... Signal supply device, 14 ... Sound emission device, 16 ... Input device, 22 ... Arithmetic processing device, 24 ... Memory | storage device, 32 ... Variable setting part, 34 ... Frequency analysis unit 36... Coefficient setting unit 38... Signal processing unit 40... Waveform synthesis unit 42 .. determination unit 52... Sum component generation unit 54. Coefficient sequence generation unit, 62 ... first generation unit, 64 ... second generation unit.

Claims (5)

Sum component generating means for generating a sum component of the first acoustic signal and the second acoustic signal in stereo format;
Difference component generating means for generating a difference component in which a localization component in a specific direction is suppressed by subtraction between the first acoustic signal and the second acoustic signal;
Determining means for determining the presence or absence of the localization component;
First generation means for generating a first processing coefficient sequence in which a coefficient value for each frequency is set so that the localization component is emphasized by subtraction processing including subtraction between the sum component and the difference component;
Second generation means for generating a second processing coefficient sequence in which the distribution of coefficient values of each frequency is level compared to the first processing coefficient sequence generated by the subtraction process when the localization component does not exist;
For each frequency component of the first acoustic signal and the second acoustic signal, when the determination unit determines that the localization component is present, each coefficient value of the first processing coefficient sequence is applied, and the localization is performed. A sound processing apparatus comprising: signal processing means for causing each coefficient value of the second processing coefficient sequence to act when the determination means determines that no component is present. The determination unit determines that the localization component is present when an average value or a degree of dispersion of the plurality of coefficient values of the first processing coefficient sequence generated by the first generation unit exceeds a threshold value, and determines the threshold value. The acoustic processing apparatus according to claim 1, wherein if it falls below, the localization component is determined not to exist. The sound processing apparatus according to claim 1, wherein the second generation unit generates the second processing coefficient sequence in which the coefficient values are set to the same value. The said 2nd production | generation means produces | generates a said 2nd process coefficient sequence by smoothing distribution of each coefficient value of the said 1st process coefficient sequence which the said 1st production | generation means produced | generated. Sound processing device. The determination means determines the presence or absence of the localization component for each unit period,
The acoustic processing according to any one of claims 1 to 4, wherein the signal processing means starts application of the second processing coefficient sequence when the presence of the localization component is continuously denied over a predetermined number of unit periods. apparatus.

Images