This application is a National Stage of International Application No. PCT/JP2016/087968 filed Dec. 20, 2016, claiming priority based on Japanese patent application No. 2016-003076, filed on Jan. 8, 2016, the disclosure of which is incorporated herein in its entirety by reference.
The present invention relates to a technique of amplifying or attenuating a signal.
In the above technical field, patent literature 1 describes an automatic gain adjustment apparatus capable of listening to only target speech without distortion with an appropriate volume regardless of a variation in voice volume of a speaker and a variation in a distance to a microphone.
This technique suppresses background noise included in an input signal, and determines a threshold as a compression/decompression boundary based on effective values of residual noise of frames determined as noise frames, thereby smoothing the effective values of past frames. A compression ratio is calculated from the smoothed effective value, and a necessary gain is obtained from the threshold and a target average effective value. The obtained gain is applied to the input signal after the background noise is suppressed, thereby automatically adjusting the gain.
In this technique, however, since an output signal is obtained by applying a gain to a signal after background noise is suppressed, noise (components other than speech) included in the input signal is suppressed and only the speech is output. Therefore, this technique cannot cope with a case in which an environmental sound needs to be held in recording a natural sound or a case in which an environmental sound needs to be slightly suppressed.
That is, with respect to a signal in which a desired signal and another signal are mixed, it is impossible to amplify or attenuate the desired signal and the other signal at different ratios.
The present invention enables to provide a technique of solving the above-described problem.
One example aspect of the present invention provides a signal processing apparatus comprising:
a unit that obtains an estimated first signal and an estimated second signal by inputting a mixed signal in which a first signal and a second signal are mixed; and
a unit that obtains a gain-adjusted mixed signal based on the estimated first signal and the estimated second signal.
Another example aspect of the present invention provides a gain adjustment method comprising:
obtaining an estimated first signal and an estimated second signal by inputting a mixed signal in which a first signal and a second signal are mixed; and
obtaining a gain-adjusted mixed signal based on the estimated first signal and the estimated second signal.
Still other example aspect of the present invention provides a gain adjustment program for causing a computer to execute a method, comprising:
obtaining an estimated first signal and an estimated second signal by inputting a mixed signal in which a first signal and a second signal are mixed; and
obtaining a gain-adjusted mixed signal based on the estimated first signal and the estimated second signal.
According to the present invention, with respect to a signal in which a desired signal and another signal are mixed, it is possible to amplify or attenuate the desired signal and the other signal at different ratios.
Example embodiments of the present invention will now be described in detail with reference to the drawings. It should be noted that the relative arrangement of the components, the numerical expressions and numerical values set forth in these example embodiments do not limit the scope of the present invention unless it is specifically stated otherwise. Note that “speech signal” in the following explanation indicates a direct electrical change that occurs in accordance with speech or another sound. The speech signal transmits speech or another sound and is not limited to speech. Some example embodiments in which the number of input mixed signals is four will be described. However, these are merely examples, and the same applies to an arbitrary number of two or more signals.
A gain adjustment apparatus 100 according to the first example embodiment of the present invention will be described with reference to
With this arrangement, the gain adjustment apparatus 100 can adjust the gains of the first and second signals included in the mixed signal. Therefore, with respect to a signal in which a desired signal and another signal are mixed, it is possible to amplify or attenuate the desired signal and the other signal at different ratios.
A gain adjustment apparatus according to the second example embodiment of the present invention will be described with reference to
As shown in
With this arrangement, the gain adjustment apparatus can generate a gain-adjusted mixed signal by applying different gains to the first and second signals included in a mixed signal and then adding the signals, in addition to the effect of the first example embodiment.
A gain adjustment apparatus according to the third example embodiment of the present invention will be described with reference to
As shown in
The gain calculator 332 determines the second gain Gd so that the gain-adjusted estimated second signal Do becomes equal to a target value Dt of the second signal. That is, Gd=|Dt|/|De|.
The gain calculator 331 determines the first gain Gs so that the gain-adjusted estimated first signal So becomes equal to a target value St. However, for example, the non-stationarity of the first signal such as speech is generally higher than that of the second signal such as noise, and the non-stationarity of the value of the gain is also high. Therefore, the first gain Gs is preferably obtained by serial processing by appropriate control. The gain calculator 331 sequentially calculates the first gain Gs using the estimated first signal Se, the gain-adjusted estimated first signal So, the target value St of the first signal, and a step size μ. At this time, the maximum or minimum value of the upper limit value of the first gain Gs may be limited using the estimated second signal De. By limiting the upper limit value of the first gain Gs, it is possible to prevent unnatural signal attenuation and distortion caused by excessive amplification when the first signal is small.
Note that the second gain is obtained using the absolute values of the estimated second signal De and the target value Dt of the gain-adjusted estimated second signal Do. However, the gain may be obtained using powers. Similarly, the gain calculator 331 may calculate the first gain using the absolute values of the estimated first signal Se, the gain-adjusted estimated first signal So, and the target value St of the first signal or using powers. Furthermore, the second gain Gd may be obtained by the same method as that for the first gain Gs using the gain calculator 331 instead of the gain calculator 332. Conversely, the first gain Gs may be obtained by the same method as that for the second gain Gd using the gain calculator 332 instead of the gain calculator 331.
With this arrangement, it is possible to determine different gains in accordance with the target values of the first and second signals, in addition to the effect of the first example embodiment.
A gain adjustment apparatus according to the fourth example embodiment of the present invention will be described with reference to
As shown in
The second example embodiment shown in
With this arrangement, the gain adjustment apparatus can generate a gain-adjusted mixed signal by applying a gain to the second signal included in a mixed signal to set a ratio with respect to the first signal, and applying another gain to a result of adding the obtained second signal and the estimated value of the first signal.
A gain adjustment apparatus according to the fifth example embodiment of the preset invention will be described with reference to
As shown in
In the fourth example embodiment, an equivalent gain for an estimated first signal is Gm and an equivalent gain for an estimated second signal is Gm·Gd. In this example embodiment, however, an equivalent gain for an estimated first signal is Gm and an equivalent gain for an estimated second signal is 1. Therefore, by appropriately controlling only the third gain Gm, it is possible to amplify or attenuate a desired signal at an arbitrary ratio and set another signal level invariable.
With this arrangement, the gain adjustment apparatus can appropriately amplify or attenuate the first signal without changing the level of the second signal by setting an appropriate gain for the first signal included in a mixed signal. In addition, since only one gain is controlled, this example embodiment can be implemented by simple processing. Therefore, by only controlling a single gain for a signal in which a desired signal and another signal are mixed, it is possible to amplify or attenuate the desired signal at an arbitrary ratio and set the level of the other signal invariable.
A gain adjustment apparatus according to the sixth example embodiment of the present invention will be described with reference to
As shown in
With this arrangement, by setting a target value for the first signal included in a mixed signal, the gain adjustment apparatus can amplify or attenuate the first signal to match the target value without changing the level of the second signal. In addition, since only one gain is controlled, this example embodiment can be implemented by simple processing. Therefore, by only controlling a single gain for a signal in which a desired signal and another signal are mixed, it is possible to amplify or attenuate the desired signal at an arbitrary ratio according to the target value and set the level of the other signal invariable.
A gain adjustment apparatus according to the seventh example embodiment of the present invention will be described with reference to
As shown in
Based on the mixed signal and the estimated first signal, the estimator 722 obtains an estimated second signal that is an estimated value of the second signal. Assuming that the mixed signal is the sum of the first and second signals and the first and second signals are uncorrelated, the power of the mixed signal is the sum of the powers of the first and second signals. Therefore, the estimator 722 obtains the power of the mixed signal and that of the estimated first signal, and subtracting the latter from the former, thereby obtaining the power of the estimated second signal. The estimator 722 obtains the estimated second signal by combining the obtained subtraction result with the phase of the mixed signal. The processing of the estimator 722 may be performed in a time domain or in a frequency domain after converting the signals into the frequency domain using Fourier transform or the like. If the processing is executed in the frequency domain, the power and the phase are combined and then converted into a time domain signal.
The subtracter 802 obtains the power of the supplied mixed signal, and subtracts the power of the second signal from the obtained power, thereby obtaining a subtraction signal. The subtracter 802 outputs, as an estimated first signal, a result of combining the power of the subtraction signal and the phase of the mixed signal. That is, since the subtracter 802 obtains the estimated first signal, the subtracter 802 can be regarded as an estimator. The processes in the estimator 801 and the subtracter 802 may be performed for absolute values instead of the powers. The processing of the enhancer 721 may be performed in the time domain or in the frequency domain after converting the signals into the frequency domain using Fourier transform or the like. If the processing is executed in the frequency domain, the power and the phase are combined and then converted into a time domain signal.
As shown in
With this arrangement, in addition to the effect of the first example embodiment, the separator can be implemented by a simple arrangement and it is thus possible to provide a low-end, high-performance gain adjustment apparatus.
A gain adjustment apparatus according to the eighth example embodiment of the present invention will be described with reference to
As shown in
Referring to
With this arrangement, in addition to the effect of the first example embodiment, the separator can be implemented by a simpler arrangement and it is thus possible to provide a low-end, high-performance gain adjustment apparatus.
A gain adjustment apparatus according to the ninth example embodiment of the present invention will be described with reference to
As shown in
The estimator 1122 receives the plurality of mixed signals and the estimated first signal, and obtains an estimated second signal that is an estimated value of the second signal. The estimator 1122 is different from the estimator 722 in that the estimator 1122 receives the plurality of mixed signals and integrates these signals into a single mixed signal.
As the single mixed signal, an arbitrary one of the plurality of mixed signals can be selected and used. Alternatively, a statistic value of these signals may be used. Examples of the statistic value are an average value, a maximum value, a minimum value, and a median. Each of the average value and the median provides a signal in a virtual sensor existing at the center of the plurality of sensors. The maximum value provides a signal in a sensor whose distance to a signal is shortest when the signal arrives from a direction other than the front direction. The minimum value provides a signal in a sensor whose distance to a signal is longest when the signal arrives from a direction other than the front direction. Furthermore, simple addition of these signals can be used. Alternatively, one of array signal processes described in non-patent literature 5 may be applied. The array signal processes include a delay-sum beamformer, filter-sum beamformer, MSNR (Maximum Signal-to-Noise Ratio) beamformer, MMSE (Minimum Mean Square Error) beamformer, LCMV (Linearly Constrained Minimum Variance) beamformer, and a nested beamformer. The present invention, however, is not limited to them. The thus calculated value is set as a single mixed signal.
The estimator 1122 receives the integrated single mixed signal and the estimated first signal, and obtains an estimated second signal that is an estimated value of the second signal by the same method as that in the estimator 722. The processing of the estimator 1122 may be performed in the frequency domain after converting the signals into the frequency domain using Fourier transform or the like. If the processing is executed in the frequency domain, the power and the phase are combined and then converted into a time domain signal.
The fixed beamformer 1201 forms a beam having high sensitivity to the direction of arrival of the first signal, and enhances the first signal, thereby obtaining an enhanced first signal. That is, the fixed beamformer 1201 functions as an enhancer for the first signal. The enhanced first signal is supplied to the blocking matrix 1202 and the multiple-input canceller 1203. As the operation of the fixed beamformer, one of the array signal processes described in non-patent literature 5 can be applied.
The blocking matrix 1202 receives the plurality of mixed signal and the enhanced first signal, and removes components correlated with the enhanced first signal from each mixed signal, thereby obtaining a plurality of pseudo second signals. That is, the blocking matrix 1202 can be regarded as an estimator for the second signal. The plurality of pseudo second signals are supplied to the multiple-input canceller 1203.
The multiple-input canceller 1203 receives the enhanced first signal and the plurality of pseudo second signals, and removes components correlated with the plurality of pseudo second signals from the enhanced first signal, thereby obtaining an estimated first signal. That is, the multiple-input canceller 1203 can be regarded as an estimator for the first signal.
As the enhancer 1121, filtering based on a phase difference described in non-patent literature 6 may be applied.
With this arrangement, in addition to the effect of the first example embodiment, the separator separates the second signal after enhancing the first signal using directivity, and it is thus possible to provide a gain adjustment apparatus having high performance, especially, with respect to a mixed signal including a signal arriving from a specific direction.
A gain adjustment apparatus according to the 10th example embodiment of the present invention will be described with reference to
As shown in
The integrator 1401 integrates a plurality of pseudo second signals output from a blocking matrix 1202, and outputs the thus obtained signal as an estimated second signal that is an estimated value of the second signal.
As the estimated second signal, an arbitrary one of the plurality of pseudo second signals can be selected and used. Alternatively, a statistic value of these signals may be used. Examples of the statistic value are an average value, a maximum value, a minimum value, and a median. Each of the average value and the median provides a signal in a virtual sensor existing at the center of a plurality of sensors. The maximum value provides a signal in a sensor whose distance to a signal is shortest when the signal arrives from a direction other than the front direction. The minimum value provides a signal in a sensor whose distance to a signal is longest when the signal arrives from a direction other than the front direction. Furthermore, simple addition of these signals can be used. Alternatively, one of the array signal processes described in non-patent literature 5 may be applied. The array signal processes include a delay-sum beamformer, filter-sum beamformer, MSNR (Maximum Signal-to-Noise Ratio) beamformer, MMSE (Minimum Mean Square Error) beamformer, LCMV (Linearly Constrained Minimum Variance) beamformer, and a nested beamformer. The present invention, however, is not limited to them. The thus calculated value is set as an estimated second signal.
According to this example embodiment, with this arrangement, the separator separates the first and second signals using directivity, and it is thus possible to provide a gain adjustment apparatus having high performance and a simple arrangement, especially, with respect to a mixed signal including a signal arriving from a specific direction.
A gain adjustment apparatus according to the 11th example embodiment of the present invention will be described with reference to
As shown in
The subtracter 1602 is also supplied with the mixed signal. The subtracter 1602 subtracts the pseudo second signal from the mixed signal, and outputs the subtraction result as an estimated first signal. That is, the subtracter 1602 functions as an estimator for the first signal. The filter coefficient is updated to minimize an expected value of the power of the subtraction result. As a coefficient update algorithm, the LMS (Least Mean Square) algorithm or normalized LMS algorithm is used most. These algorithms are described in patent literatures 6 and 7 and non-patent literature 7, and a detailed description thereof will be omitted. Another coefficient update algorithm such as the LS (Least Square) algorithm can also be used. The processing of the enhancer 1521 may be performed in the time domain or in the frequency domain after converting the signals into the frequency domain using Fourier transform or the like. If the processing is executed in the frequency domain, the obtained signal is converted into a time domain signal after enhancement processing.
According to this example embodiment, with this arrangement, the first signal is enhanced using the reference signal and then the second signal is separated. It is thus possible to provide a gain adjustment apparatus having high performance, especially, with respect to a mixed signal including a diffusible signal.
A gain adjustment apparatus according to the 12th example embodiment of the present invention will be described with reference to
As shown in
The enhancer 1721 shown in
With this arrangement, in addition to the effect of the first example embodiment, the first and second signals are separated using the reference signal, and it is thus possible to provide a gain adjustment apparatus having high performance, especially, with respect to a mixed signal including a diffusible signal.
A gain adjustment apparatus according to the 13th example embodiment of the present invention will be described with reference to
The gain calculator 1901 includes an average unit 1911, a reciprocal converter 1912, a multiplier 1913, a subtracter 1914, a multiplier 1915, an adder 1916, an average unit 1917, a limiter 1918, a storage unit 1919, and a delay unit 1920. The reciprocal converter 1912 receives an estimated first signal (or provisional gain-adjusted mixed signal Sp) Se, obtains a reciprocal 1/Se (or 1/Sp), and transmits it to the multiplier 1913. The multiplier 1913 receives a step size μ and the reciprocal 1/Se (or 1/Sp) of the estimated first signal (or provisional gain-adjusted mixed signal Sp), calculates a product μ/Se (or μ/Sp) as a normalized step size, and transmits it to the multiplier 1915.
The subtracter 1914 receives a gain-adjusted mixed signal Xo and a target value St of the first signal, obtains an error Xo−St, and transmits it to the multiplier 1915. The multiplier 1915 receives the normalized step size μ/Se (or μ/Sp) and the error Xo−St, obtains a product μ(Xo−St)/Se (or μ(Xo−St)/Sp) as a gain-adjusted signal, and transmits it to the adder 1916.
Every time a first gain Gs is updated, it is stored as a new value Gsn in the storage unit 1919. The new value Gsn of the first gain Gs read out from the storage unit 1919 is transmitted to the delay unit 1920. The delay unit 1920 delays the new value Gsn of the first gain Gs, and transmits the delayed value as a current value Gsc of the first gain Gs to the adder 1916.
The adder 1916 obtains the new value Gsn of the first gain by adding the current value Gsc of the first gain Gs supplied from the delay unit 1920 and the gain-adjusted signal μ(Xo−St)/Se supplied from the multiplier 1915, and stores the obtained value in the storage unit 1919. That is, the first gain is updated by an equation below.
This equation is nothing but the normalized LMS algorithm for a one-tap adaptive filter. Therefore, to update the first gain, another adaptive filter coefficient update algorithm such as the LMS algorithm or LS algorithm can be used.
The average unit 1911 receives the estimated first signal Se, averages it, and supplies the averaged estimated first signal to the reciprocal converter 1912. The average unit 1917 receives the gain-adjusted mixed signal Xo, averages it, and supplies the averaged gain-adjusted mixed signal to the subtracter 1914. These average units reduce excessive variations in supplied signals by averaging, thereby contributing to avoidance of an undesired operation including instability caused by an excessively large or small value that occurs instantaneously.
The same applies to a case in which the gain calculator 1901 is replaced with the gain calculator 631, by replacing the estimated first signal Se with the provisional gain-adjusted mixed signal Sp.
If the first gain becomes excessively large, especially when the first signal is not sufficiently larger than the second signal, distortion occurs in the gain-adjusted mixed signal. The reason for this is as follows. That is, when the first signal is represented by S and the second signal is represented by N, a case in which the first signal is not sufficiently larger than the second signal indicates a state in which the SN ratio (SNR) is not sufficiently high. At this time, the component of the first signal is masked by the component of the second signal, and it is difficult to separate the first signal from the mixed signal. An estimated value of the first signal separated from the mixed signal in this state, that is, the estimated first signal includes a large error and is perceived as a distorted signal. This distortion poses a problem particularly when the first signal is small. When the first signal is small, the first gain takes a large value and the distortion is readily perceived by gain adjustment. To prevent this, it is effective to limit the first gain not to become excessively large when the first signal is not sufficiently larger than the second signal.
The limiter 1918 shown in
An example of a method of determining an upper limit value Gsmax of the first gain will be described with reference to
Since Gsmax=1, the level of the first signal is equal to St, and the SNR is St/δ2. For example, if δ2 is set to be half the target value St of the first signal, that is, δ2=0.5St is set, the SNR at Q is 3 dB. If an estimated value of the power of the second signal at P is δ1, corresponding Gsmax=G0 is given by G0=(δ1/δ2)·St=2δ1 under the condition that the SNR is constant.
Although the SNRs at P and Q are equal to each other, the second signal at P is smaller than that at Q, and thus the first signal at P is smaller than that at Q, and only the LSB side of a fixed point representation is used. That is, a resolution with respect to the first signal is lower at P than at Q. If the same first gain is applied at P and Q in this state, the first gain at P becomes excessively large and distortion in the gain-adjusted mixed signal is perceived.
To avoid the distortion from being perceived, the upper limit value Gsmax=G0 is introduced to the first gain. Since the upper limit value of the first gain depends on the minimum value of the first signal, the minimum value of the first signal is estimated and the upper limit value is determined based on the estimated minimum value. The minimum value of the first signal can be obtained by, for example, sequentially comparing the value of the estimated first signal with a provisional minimum value, and setting a smaller value as a new provisional minimum value. The first value of the estimated first signal is set as the initial value of the provisional minimum value. The upper limit value of the first gain may be obtained by assigning in advance an appropriate fixed value as the minimum value of the first signal, and reading it out from a storage device.
The minimum value can be provided for the upper limit value Gsmax of the first gain. For example, referring to
Setting of the maximum and minimum values of the upper limit value Gsmax of the first gain has been described with reference to
According to this example embodiment, with this arrangement, in addition to the effect of the first example embodiment, when calculating a gain, the upper limit value of the gain is limited to a predetermined range, and it is thus possible to avoid unnatural signal attenuation and distortion caused by excessive amplification when the power of a desired signal is particularly small.
A gain adjustment apparatus according to the 14th example embodiment of the present invention will be described with reference to
The difference from the sixth example embodiment of the present invention is that a mixed signal converted into the frequency domain by a converter 2101 is supplied to a separator 102, and a provisional gain-adjusted mixed signal output from an adder 431 is converted into the time domain by an inverter 2132 and then supplied to a multiplier 432 and a gain calculator 631. The arrangements and operations of the converter 2101 and the inverter 2132 are described in patent literature 7 and a description thereof will be omitted.
By executing separation processing in the frequency domain, it is possible to apply different processing for each frequency in accordance with the distribution states (power spectra or amplitude spectra) of the frequency components of an input mixed signal and the first and second signals as separation targets, thereby making it possible to improve the separation accuracy of the first and second signals.
Note that the inverter needs to be arranged before the gain calculator. This is because if a different gain is obtained for each frequency, the shape of the power spectrum or amplitude spectrum of a signal to be applied with the gain may be destroyed.
Furthermore, the converter 2101 and the inverter 2132 can be configured to simply perform only frame division and frame composition. Since one common gain is calculated for a plurality of signal samples forming a frame, the effect of averaging works to make it possible to obtain a stable value of a gain for a signal with high non-stationarity. Therefore, it is possible to perform gain adjustment capable of executing stable gain control.
According to this example embodiment, with this arrangement, processing is performed in the frequency domain when separating a desired signal and another signal, and it is thus possible to provide a gain adjustment apparatus that can perform stable gain control for a signal with high separation accuracy and high non-stationarity.
A gain adjustment apparatus according to the 15th example embodiment of the present invention will be described with reference to
The gain adjustment apparatus 2200 includes a processor 2210, a ROM (Read Only Memory) 2220, a RAM (Random Access Memory) 2240, a storage 2250, an input/output interface 2260, an operation unit 2261, an input unit 2262, and an output unit 2263. The processor 2210 is a central processing unit, and controls the overall gain adjustment apparatus 2200 by executing various programs.
The ROM 2220 stores a boot program to be executed first by the processor 2210, various parameters, and the like. The RAM 2240 has a program load area (not shown), and an area for storing a mixed signal 2240a (input signal), an estimated first signal 2240b, an estimated second signal 2240c, a gain 2240d, a gain-adjusted mixed signal 2240e (output signal), and the like.
The storage 2250 also stores a gain adjustment program 2251. The gain adjustment program 2251 includes a signal separation module 2251a, a gain calculation module 2251b, and a multiplication module 2251c. When the processor 2210 executes the modules included in the gain adjustment program 2251, the functions of the separator 102 shown in
The gain-adjusted mixed signal 2240e as an output associated with the gain adjustment program 2251 executed by the processor 2210 is output from the output unit 2263 via the input/output interface 2260. This can individually perform gain adjustment for a desired signal and another signal included in the mixed signal 2240a input from the input unit 2262.
In step S2305, individual gains are calculated for the first and second signals. In step S2307, a gain-adjusted first signal and a gain-adjusted second signal are calculated by applying the calculated gains, respectively. In step S2309, the gain-adjusted second signal is added to the gain-adjusted first signal to generate a gain-adjusted mixed signal.
Finally, in step S2311, the sum of the gain-adjusted first signal and the gain-adjusted second signal is output as a gain-adjusted mixed signal obtained by individually performing gain adjustment for a desired signal and another signal.
An example of the processing procedure when the gain adjuster 103 with the arrangement according to the third example embodiment is implemented by software in the gain adjustment apparatus according to this example embodiment has been explained with reference to
According to this example embodiment, with the above arrangement, generation of a gain-adjusted mixed signal can be implemented by software by applying different gains to the first and second signals included in a mixed signal.
While the invention has been particularly shown and described with reference to example embodiments thereof, the invention is not limited to these example embodiments. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the claims.
The present invention is applicable to a system including a plurality of devices or a single apparatus. The present invention is also applicable even when an information processing program for implementing the functions of example embodiments is supplied to the system or apparatus directly or from a remote site. Hence, the present invention also incorporates the program installed in a computer to implement the functions of the present invention by the computer, a medium storing the program, and a WWW (World Wide Web) server that causes a user to download the program. Especially, the present invention incorporates at least a non-transitory computer readable medium storing a program that causes a computer to execute processing steps included in the above-described example embodiments.
Some or all of the above-described example embodiments can also be described as in the following supplementary notes but are not limited to the followings.
(Supplementary Note 1)
There is provided a signal processing apparatus comprising:
a separator that obtains an estimated first signal and an estimated second signal from a mixed signal in which a first signal and a second signal are mixed; and
a gain adjuster that obtains a gain-adjusted mixed signal using the estimated first signal and the estimated second signal.
(Supplementary Note 2)
There is provided the signal processing apparatus according to supplementary note 2, wherein the gain adjuster includes
a first multiplier that obtains a gain-adjusted estimated first signal by applying a first gain to the estimated first signal,
a second multiplier that obtains a gain-adjusted estimated second signal by applying a second gain to the estimated second signal, and
a first adder that obtains a gain-adjusted mixed signal by adding the gain-adjusted estimated first signal and the gain-adjusted estimated second signal.
(Supplementary Note 3)
There is provided the signal processing apparatus according to supplementary note 1, wherein the gain adjuster includes
a first gain calculator that obtains a first gain using the estimated first signal and a target value of the first signal, and
a second gain calculator that obtains a second gain using the estimated second signal and a target value of the second signal.
(Supplementary Note 4)
There is provided the signal processing apparatus according to supplementary note 1, wherein the gain adjuster includes
a third multiplier that obtains a gain-adjusted estimated second signal by applying a second gain to the estimated second signal,
a second adder that obtains a provisional gain-adjusted mixed signal by adding the estimated first signal and the gain-adjusted estimated second signal, and
a fourth multiplier that obtains a gain-adjusted mixed signal by applying a third gain to the provisional gain-adjusted mixed signal.
(Supplementary Note 5)
There is provided the signal processing apparatus according to supplementary note 4, wherein the gain adjuster further includes a first reciprocal converter that obtains a reciprocal of the third gain as the second gain.
(Supplementary Note 6)
There is provided the signal processing apparatus according to supplementary note 5, wherein the gain adjuster further includes a third gain calculator that obtains the third gain using the provisional gain-adjusted mixed signal, a target value of the first signal, and the gain-adjusted mixed signal.
(Supplementary Note 7)
There is provided the signal processing apparatus according to any one of supplementary notes 1 to 6, wherein the separator includes
a first enhancer that obtains the estimated first signal by enhancing the first signal, and
a first estimator that obtains the estimated second signal from the mixed signal and the estimated first signal.
(Supplementary Note 8)
There is provided the signal processing apparatus according to supplementary note 7, wherein the first enhancer includes
a second estimator that obtains a pseudo second signal by estimating the second signal, and
a third estimator that obtains the estimated first signal using the mixed signal and the pseudo second signal.
(Supplementary Note 9)
There is provided the signal processing apparatus according to supplementary note 7, wherein the first enhancer includes
a second estimator that obtains a pseudo second signal by estimating the second signal,
a fourth gain calculator that obtains a fourth gain using the mixed signal and the pseudo second signal, and
a fourth estimator that obtains the estimated first signal using the mixed signal and the fourth gain.
(Supplementary Note 10)
There is provided the signal processing apparatus according to any one of supplementary notes 1 to 6, wherein the separator includes a second enhancer that obtains the estimated first signal by enhancing the first signal, and obtains the estimated second signal by estimating the second signal.
(Supplementary Note 11)
There is provided the signal processing apparatus according to supplementary note 10, wherein the second enhancer includes
a second estimator that obtains a pseudo second signal by estimating the second signal, and
a third estimator that obtains the estimated first signal using the mixed signal and the pseudo second signal, and
outputs the pseudo second signal as the estimated second signal.
(Supplementary Note 12)
There is provided the signal processing apparatus according to supplementary note 10, wherein the second enhancer includes
a second estimator that obtains a pseudo second signal by estimating the second signal,
a fourth gain calculator that obtains a fourth gain using the mixed signal and the pseudo second signal, and
a fourth estimator that obtains the estimated first signal using the mixed signal and the fourth gain, and
outputs the pseudo second signal as the estimated second signal.
(Supplementary Note 13)
There is provided the signal processing apparatus according to any one of supplementary notes 1 to 6, wherein the separator includes
a third enhancer that obtains the estimated first signal by receiving a plurality of mixed signals and enhancing the first signal, and
a fifth estimator that obtains the estimated second signal from the plurality of mixed signals and the estimated first signal.
(Supplementary Note 14)
There is provided the signal processing apparatus according to supplementary note 13, wherein the third enhancer includes
a fourth enhancer that obtains an enhanced first signal by receiving the plurality of mixed signals and enhancing the first signal,
a sixth estimator that obtains a plurality of pseudo second signals uncorrelated with the enhanced first signal by receiving the plurality of mixed signals and the enhanced first signal, and
a seventh estimator that obtains the estimated first signal using the enhanced first signal and the plurality of pseudo second signals.
(Supplementary Note 15)
There is provided the signal processing apparatus according to any one of supplementary notes 1 to 6, wherein the separator includes a fifth enhancer that obtains the estimated first signal by receiving a plurality of mixed signals and enhancing the first signal, and obtains the estimated second signal by removing correlation with the first signal from the plurality of mixed signals.
(Supplementary Note 16)
There is provided the signal processing apparatus according to supplementary note 15, wherein the fifth enhancer includes
a fourth enhancer that obtains an enhanced first signal by receiving the plurality of mixed signals and enhancing the first signal,
a sixth estimator that obtains a plurality of pseudo second signals uncorrelated with the enhanced first signal by receiving the plurality of mixed signals and the enhanced first signal,
a seventh estimator that obtains the estimated first signal using the enhanced first signal and the plurality of pseudo second signals, and
an integrator that obtains the estimated second signal by integrating the plurality of pseudo second signals.
(Supplementary Note 17)
There is provided the signal processing apparatus according to any one of supplementary notes 1 to 6, wherein the separator includes
a sixth enhancer that further receives a reference signal correlated with the second signal and obtains the estimated first signal using the mixed signal and the reference signal, and
a first estimator that obtains the estimated second signal from the mixed signal and the estimated first signal.
(Supplementary Note 18)
There is provided the signal processing apparatus according to supplementary note 17, wherein the sixth enhancer includes
an eighth estimator that obtains a pseudo second signal correlated with the second signal using the reference signal, and
a ninth estimator that obtains the estimated first signal by removing the pseudo second signal from the mixed signal.
(Supplementary Note 19)
There is provided the signal processing apparatus according to any one of supplementary notes 1 to 6, wherein the separator includes a seventh enhancer that further receives a reference signal correlated with the second signal, obtains the estimated second signal by estimating the second signal based on the reference signal, and obtains the estimated first signal by removing the estimated second signal from the mixed signal.
(Supplementary Note 20)
There is provided the signal processing apparatus according to supplementary note 19, wherein the seventh enhancer includes
an eighth estimator that obtains a pseudo second signal correlated with the second signal using the reference signal, and
a ninth estimator that obtains the estimated first signal by removing the pseudo second signal from the mixed signal, and
outputs the pseudo second signal as the estimated second signal.
(Supplementary Note 21)
There is provided the signal processing apparatus according to any one of supplementary notes 3 to 20, wherein the first gain calculator includes
a second reciprocal converter that obtains a reciprocal of the estimated first signal,
a fourth multiplier that obtains a normalized signal by multiplying the reciprocal of the estimated first signal by a step size,
a subtracter that obtains, as an error, a difference between the gain-adjusted mixed signal and the target value of the first signal,
a fifth multiplier that obtains a gain-adjusted signal by multiplying the normalized signal by the error,
a third adder that obtains an updated value of the first gain using the gain-adjusted signal and a past value of the first gain,
a storage unit that stores the updated value of the first gain, and
a delay unit that delays the updated value of the first gain stored in the storage unit and supplies the delayed updated value to the adder.
(Supplementary Note 22)
There is provided the signal processing apparatus according to supplementary note 21, wherein the first gain calculator further includes a limiter that receives the estimated second signal, limits the updated value of the first gain, and supplies the limited updated value to the storage unit.
(Supplementary Note 23)
There is provided the signal processing apparatus according to supplementary note 21 or 22, wherein the first gain calculator further includes
a first average unit that averages the estimated first signal and supplies the averaged estimated first signal to the reciprocal unit, and
a second average unit that averages the gain-adjusted mixed signal and supplies the averaged gain-adjusted mixed signal to the subtracter.
(Supplementary Note 24)
There is provided the signal processing apparatus according to any one of supplementary notes 3 to 23, wherein at least one of the estimated first signal, the estimated second signal, and the provisional gain-adjusted mixed signal is supplied as a frame signal whose unit is a frame formed from a plurality of signal samples, and one of the first gain calculator and the second gain calculator that is supplied with the frame signal calculates one gain for each frame.
(Supplementary Note 25)
There is provided a gain adjustment method comprising:
obtaining an estimated first signal and an estimated second signal by receiving a mixed signal in which a first signal and a second signal are mixed and separating the first signal and the second signal; and
obtaining a gain-adjusted mixed signal by receiving the estimated first signal and the estimated second signal.
(Supplementary Note 26)
There is provided a gain adjustment method comprising:
obtaining an estimated first signal and an estimated second signal by receiving a mixed signal in which a first signal and a second signal are mixed and separating the first signal and the second signal; and
obtaining a gain-adjusted mixed signal by applying different gains to the estimated first signal and the estimated second signal.
(Supplementary Note 27)
There is provided the gain adjustment method according to supplementary note 26, wherein one of the different gains is a reciprocal of the other gain.
(Supplementary Note 28)
There is provided a gain adjustment program for causing a computer to execute a method, comprising:
obtaining an estimated first signal and an estimated second signal by receiving a mixed signal in which a first signal and a second signal are mixed and separating the first signal and the second signal; and
obtaining a gain-adjusted mixed signal by receiving the estimated first signal and the estimated second signal.
Number | Date | Country | Kind |
---|---|---|---|
2016-003076 | Jan 2016 | JP | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/JP2016/087968 | 12/20/2016 | WO | 00 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2017/119284 | 7/13/2017 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
5727073 | Ikeda | Mar 1998 | A |
5822728 | Applebaum | Oct 1998 | A |
5930336 | Junqua | Jul 1999 | A |
7184937 | Su | Feb 2007 | B1 |
7590528 | Kato et al. | Sep 2009 | B2 |
7720233 | Sato et al. | May 2010 | B2 |
8050717 | Sugiyama | Nov 2011 | B2 |
8223989 | Sugiyama | Jul 2012 | B2 |
8736359 | Sugiyama | May 2014 | B2 |
9318119 | Sugiyama et al. | Apr 2016 | B2 |
9543926 | Sato et al. | Jan 2017 | B2 |
9905247 | Sugiyama et al. | Feb 2018 | B2 |
20040049383 | Kato et al. | Mar 2004 | A1 |
20070021958 | Visser | Jan 2007 | A1 |
20070053455 | Sugiyama | Mar 2007 | A1 |
20070071253 | Sato et al. | Mar 2007 | A1 |
20080201138 | Visser | Aug 2008 | A1 |
20080262835 | Oshikiri | Oct 2008 | A1 |
20090299742 | Toman | Dec 2009 | A1 |
20100010808 | Sugiyama et al. | Jan 2010 | A1 |
20100217587 | Sato et al. | Aug 2010 | A1 |
20110286604 | Matsuo | Nov 2011 | A1 |
20120033725 | Sugiyama | Feb 2012 | A1 |
20120158404 | Shin | Jun 2012 | A1 |
20120268198 | Sugiyama | Oct 2012 | A1 |
20150269947 | Liu | Sep 2015 | A1 |
20160254008 | Sugiyama et al. | Sep 2016 | A1 |
20160261951 | Matheja | Sep 2016 | A1 |
20160329949 | Cloutier | Nov 2016 | A1 |
Number | Date | Country |
---|---|---|
8-033091 | Feb 1996 | JP |
9-018291 | Jan 1997 | JP |
2001-095082 | Apr 2001 | JP |
2002-204175 | Jul 2002 | JP |
2007-068125 | Mar 2007 | JP |
2009-193031 | Aug 2009 | JP |
2011-100030 | May 2011 | JP |
2011-244232 | Dec 2011 | JP |
2012-120133 | Jun 2012 | JP |
2013-168878 | Aug 2013 | JP |
2014-160156 | Sep 2014 | JP |
9-311696 | Dec 2017 | JP |
2005024787 | Mar 2005 | WO |
2007026691 | Mar 2007 | WO |
2015049921 | Apr 2015 | WO |
Entry |
---|
Steven F. Boll, “Suppression of Acoustic Noise in Speech Using Spectral Subtraction”, IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP-27, No. 2, Apr. 1979, pp. 113-120 (8 pages total). |
Yariv Ephraim et al., “Speech Enhancement Using a Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator”, IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP-32, No. 6, Dec. 1984, pp. 1109-1121 (13 pages total). |
Jalal Taghia et al., “An Evaluation of Noise Power Spectral Density Estimation Algorithms in Adverse Acoustic Environments”, IEEE Proceedings of International Conference on Acoustics, Speech, and Signal Processing, pp. 4640-4643, May 2011 (4 pages total). |
Lloyd J. Griffiths et al., “An Alternative Approach to Linearly Constrained Adaptive Beamforming”, IEEE Transactions on Antennas and Propagation, vol. AP-30, No. 1, Jan. 1982, pp. 27-34 (8 pages total). |
S. Gannot et al., “Adaptive Beamforming and Postfiltering”, Handbook of Speech Processing, Springer, Berlin Heidelberg New York, 2008, pp. 945-978 (34 pages total). |
Akihiko Sugiyama et al., “A Directional Noise Suppressor With a Specified Beamwidth”, IEEE Proceedings of International Conference on Acoustics, Speech, and Signal Processing, pp. 524-528, Apr. 2015 (5 pages total). |
Bernard Widrow et al., “Adaptive Noise Cancelling: Principles and Applications”, Proceedings of the IEEE, Vol. 63, No. 12, pp. 1692-1716, Dec. 1975 (25 pages total). |
International Search Report dated Mar. 7, 2017 issued by the International Searching Authority in International Application No. PCT/JP2016/087968. |
Number | Date | Country | |
---|---|---|---|
20190027159 A1 | Jan 2019 | US |