Patent Translate Powered by EPO and Google Notice This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate, complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or financial decisions, should not be based on machine-translation output. DESCRIPTION JPH09212196 [0001] BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is to suppress signals other than a target signal (in this case, noise) and efficiently pick up only a target signal when collecting speech signals etc. in various noise environments. The present invention relates to a noise suppressor capable of [0002] 2. Description of the Related Art In general, the main purpose of general audio equipment is to efficiently collect an acoustic signal and to amplify it. The basic components are (1) a microphone which picks up an acoustic signal and converts it into an electrical signal, (2) an amplification circuit which amplifies the electrical signal, and (3) the amplified electrical signal again as an acoustic signal. It is three elements of an acoustic transducer represented by a speaker, a receiver, etc. which convert. Here, in the component which picks up the acoustic signal of said (1), the objective can be classified into two. That is, one is to pick up all the acoustic signals as faithfully as possible, and the other is to pick up only the desired signal. [0003] The present invention relates to the latter "effectively collecting only a target signal". Among these, there are devices that use a plurality of microphones etc. to more efficiently pick up a 10-04-2019 1 target signal (hereinafter this will be called an audio signal for the convenience of explanation and the others will be called noise). The present invention relates to an apparatus for suppressing noise other than a target signal with respect to a collected input signal. In this so-called noise suppression device, many devices have been conventionally realized. Among them, the signal (S. Boll, "Suppression of acoustic noise in speech using spectral subtraction", IEEE Trans., ASSP, Vol. 27, No. 2, pp. 113-120, 1979) There is a noise suppressor focused on the power spectrum. [0004] An outline of the operation of the basic principle will be described with reference to FIG. In FIG. 4, 101 is an input signal end, 102 is a signal identification circuit that determines whether the input signal is an audio signal or noise, 103 is a frequency analysis circuit that obtains power spectrum and phase information of the input signal, 104 is a storage circuit, 105 is a storage circuit. The switch is controlled by the output of the signal identification circuit 102, and is closed only when the input signal is noise, and the output of the frequency analysis circuit 103 is stored in the storage circuit 104. Reference numeral 106 denotes subtraction means, and reference numeral 107 denotes an inverse frequency analysis circuit, which performs an operation reverse to that of the frequency analysis circuit 103. Reference numeral 108 denotes an output signal end. [0005] Next, the operation will be described. First, an input signal is taken from the input signal end 101 and sent to the signal identification circuit 102 and the frequency analysis circuit 103. In the signal identification circuit 102, frequency distribution characteristics of signal levels (RJ Mcaulay and MLMalpass, "Speech enhancement using a soft decision noise suppression filter, IEEE Trans., ASSP, Vol. 28, No. 2, pp. 137-145 , 1980) to identify the type of speech / noise. The frequency analysis circuit 103 obtains the power spectrum S (f) of the signal and the phase information P (f). The frequency analysis here is usually performed using FFT (Fast Fourier Transform) or the like. The storage circuit 104 is an average obtained by turning the switch 105 to the NS side only in the case of noise according to the identification result of the signal analyzed by the signal identification circuit 102 (a discrimination result of voice or noise). Power spectrum characteristic “Sns (f)” of the typical noise is stored. Here, when the identification result in the signal identification circuit 102 is identified as "voice", the switch 105 falls to the S side. Then, a spectral characteristic S ′ (f) is obtained by subtracting the average noise spectrum Sn (f) from the input signal spectrum S (f) by the subtracting means 106. Finally, using the S '(f) obtained here and the analyzed phase information P (f), the reverse frequency analysis circuit 10-04-2019 2 107 converts the signal into a time domain signal and outputs it from the output signal end 108. At this time, the phase information of the signal is used as it is as the analysis result without doing anything. The above process can be expressed by equation (1). [0006] Here, α is a subtraction coefficient, and ns (f) is a low level noise that is usually added so that the spectrum after subtraction is not negative. [0007] By such processing, the signal output terminal 108 outputs a signal from which the frequency spectrum component of the noise component has been removed. The noise suppression method as described above can ideally suppress noise if the power spectrum characteristic of the noise signal is substantially stationary. However, in general, the characteristics of noise in the natural world, although "almost stationary", change from moment to moment. Therefore, although the output signal processed by the noise suppression apparatus as shown in the conventional method is suppressed in noise and difficult to hear, noise that can not be suppressed is newly heard, so this is actually an offensive noise. Then, this is called residual noise), which has been a major issue in realizing the noise suppression device in the conventional method. [0008] An object of the present invention is to provide a noise suppressor capable of picking up only a target signal efficiently. [0009] SUMMARY OF THE INVENTION In the present invention, the power spectrum characteristic of an input signal is provided to minimize the possibility of hearing residual noise, which is the biggest problem in noise suppressors implemented using the above-described conventional method. It is characterized in that auditory weighting is applied to the average power spectrum characteristic of noise which is subtracted from. 10-04-2019 3 That is, this is a method of newly using an auditory weighting coefficient W (f) instead of the subtraction coefficient α in the above-mentioned equation (1). By introducing such a weighting factor, it becomes possible to greatly reduce the residual noise audible to the ear. [0010] In other words, the value of α in the equation (1) in the conventional method uses a value of 1.0 or more in order to make the noise suppression amount as large as possible. However, while it is possible to greatly suppress noise by increasing this value, there are many cases where the target signal portion is also suppressed, and there is a possibility that it will be "excessive suppression". In the present invention, since the weighting coefficient W (f) is used so as to increase the noise suppression amount without giving a large distortion to the target signal, it is possible to suppress the quality degradation of the processed speech as a result. . [0011] Furthermore, in the present invention, although the residual noise can be reduced as much as possible by the above-mentioned method, there are still cases where residual noise can not be completely suppressed due to the type and size of noise (signal to noise ratio). Noise is often heard in sections where there is no audio signal. Therefore, the device according to the present invention is characterized in that residual noise can be suppressed almost completely in a signal section where there is no audio signal by performing loss control on this noise in order to further suppress the residual noise. [0012] In the present invention, since voice and noise are identified, noise is multiplied by an auditory weighting factor to obtain a noise spectrum characteristic, and this is subtracted from the power spectrum of the input signal. As much as possible, and significantly reduce the residual noise that can be heard aurally. [0013] Further, by controlling the loss to the residual noise, the residual noise disappears almost completely. 10-04-2019 4 [0014] FIG. 1 is a block diagram showing an embodiment of the present invention. [0015] In FIG. 1, 1000 indicates a noise suppressor, 1001 indicates an auditory weighting side, and 1002 indicates a loss control side. 201 is an input signal end, 202 is a frequency analysis circuit, 203 is a linear prediction analysis circuit, 204 is an autocorrelation analysis circuit, 205 is a maximum value selection circuit, and 206 is a voice / non-voice discrimination circuit. 207A and 207B are on / off controlled. [0016] Reference numeral 208 denotes a noise spectrum characteristic calculation and storage circuit where auditory weighting is performed. Reference numeral 209 denotes subtraction means, and reference numeral 210 denotes an inverse frequency analysis circuit which operates in the reverse order of the frequency analysis circuit 202. The above corresponds to the auditory weighting side 1001. [0017] Reference numeral 211 denotes an average noise level storage circuit, 212 denotes a loss control coefficient circuit, 213 denotes an output signal calculation circuit, 214 denotes an arithmetic means, 215 denotes an output signal end, and the above corresponds to the loss control side 1002. [0018] 10-04-2019 5 Next, the operation of the embodiment of the present invention will be described. The input signal is taken from the signal input terminal 201, and the power spectrum S (f) and the phase information P (f) are obtained in the frequency analysis circuit 202 as in the conventional method. At the same time, the input signal is extracted by the linear prediction analysis circuit 203 as a linear prediction residual signal (herein referred to as a residual signal). The residual signal is sent to the autocorrelation analysis circuit 204 where it obtains the autocorrelation function (Cor [i]) of the residual signal. Then, the peak value of the autocorrelation coefficient (the maximum value, which is called here Rmax) is determined in the maximum value selection circuit 205, and the type of the input signal is determined in the speech / non-speech discrimination circuit 206 using this Rmax. Identify That is, when Rmax is larger than a certain value (for example, Th), it is determined to be a voice signal and less than that as noise. This Rmax is often used as a feature that can well express the strength of the periodicity of the signal waveform. That is, of the input signals, most of the noise signals often have random characteristics in the time or frequency domain, and on the other hand, most of the voice signals are occupied by voiced sound, and the signals have periodicity. Therefore, it is effective to distinguish a non-periodic signal section as noise. Of course, the voice signal contains unvoiced consonants, and such a feature quantity relating to periodicity alone can not accurately discriminate voice / non-voice. However, it is very difficult to accurately detect unvoiced consonants (for example, p, t, k, s, h, f, etc.) having a very small signal level among various environmental noises. Therefore, in the device of the present invention, speech / non-speech discrimination is performed based on the idea that "a signal section that is considered not to be a speech signal is surely identified and its long-time average spectral feature is determined". . [0019] In other words, it is only necessary to obtain "the average spectral characteristic of a signal which is surely considered as a noise signal", and by setting Rmax to a small value, a typical noise spectral characteristic can be obtained. For example, FIG. 2 shows an example of the average spectral characteristic (Ssel [f]) of a signal section that is considered to be noise by identifying a noise signal collected by "cafeteria" using Rmax. The same figure also shows the average spectrum characteristics (Sall [f] and the difference characteristics between the two (| Sall [f]-Ssel [f] |) when the noise section is specified in the inspection. Here, the value of Rmax was 0.14, the measurement time length was 12 seconds, and the noise discrimination rate at this time was 77.8%. From the same figure, the difference between Sall [f] and Ssel [f] is very small, and it is possible to obtain an average noise spectrum characteristic by Rmax even for environmental noise mixed with various noises such as "cafeteria". It is well understood that 10-04-2019 6 [0020] Now, only when the frequency-analyzed signal spectrum S (f) is identified as noise, the switch 207A closes and is stored in the noise spectrum characteristic calculation and storage circuit 208 as a noise spectrum Sns (f). The update of the noise spectrum characteristic when the input signal is determined to be noise at time t is obtained by equation (2). [0021] Here, Snew (t, f) represents the updated noise spectrum, Sold (f) represents the noise spectrum before update, and St (f) represents the noise spectrum when the input signal is identified as noise. Also, β is an average weighting coefficient. [0022] Although the noise suppression process is performed by the same method as that of equation (1), in the present invention, W (f) as shown by equation (3) is used instead of α in equation (1). [0023] W (f) works to reduce the "hearing" of the residual noise as much as possible, and the effect is enhanced by the equation (4). That is, when f of W (f) is replaced with i as a point of frequency, [0024] It is represented by. Here, fc is a value corresponding to the frequency band of the input signal, B and K are weighting factors, and the larger the value, the larger the amount of suppression. This auditory weighting coefficient is not limited to the characteristic as shown in the equation (4), but even the one simulating the average characteristic of noise naturally has the same effect, and 10-04-2019 7 is not limited to the equation (4). Furthermore, the weighting factors B and K may be fixed to values that are devices, but the efficiency of noise suppression can be further increased by adaptively changing them according to the type and size of noise. [0025] By the above processing, the average spectrum of the noise superimposed on the input signal is removed, and a new spectrum S ′ (f) is obtained from the calculation means 209. This and the previously analyzed phase information P (f) are processed by the inverse frequency analysis circuit 210 to return from the frequency domain to the time domain to obtain a signal waveform. Although this signal waveform has the noise frequency component suppressed, only the speech signal will remain, but in fact, spectral characteristics such as various environmental noises changing from moment to moment slightly differ from the average spectral characteristics. . Therefore, even if the residual noise can be greatly reduced, it is necessary to further suppress the noise level depending on the type and size of the noise. Therefore, in the present invention, the following process is performed to solve this problem. [0026] That is, the residual noise level when the input signal is identified as noise is stored in the average noise level storage circuit 211. This average noise level Lns will be updated only when the input signal is identified as noise, similar to the average spectral characteristic described above. For example, the average noise level Lnew [t] updated at time t can be obtained by equation (5). [0027] Here, Lold represents the average noise level before updating, and Ln [t] represents the residual noise level at time t. As in the equation (2), β is a coefficient of averaging and is set under the condition of β <1.0. Then, using both of them, a control coefficient for loss control of the output signal level is calculated. Specifically, the loss control coefficient at time t is determined by the loss control coefficient calculation circuit 212 as A [t] by equation (6). [0028] 10-04-2019 8 Here, Ls [t] is an output signal of the inverse frequency analysis circuit 210 and is calculated by the output signal calculation circuit 213, and μ indicates a desired loss amount. However, this loss control coefficient A [t] is A [t] <= 1.0. In the final output signal of the present apparatus, the output signal waveform of the inverse frequency analysis circuit 210 is multiplied by the loss control coefficient A [t] by the computing means 214, and an output signal is obtained from the output signal end 215. [0029] The processing described above is based on frame processing, and it is preferable that signal output is performed by multiplying each of the cutout window lengths and multiplying them after multiplication by a cosine window or the like. [0030] FIG. 3 is a block diagram illustrating an example of the use of the present invention. In this figure, 2000 is a multi-microphone system, which comprises, for example, 10 multimicrophones 2001 and a processing circuit 2002, which is provided at the front stage of the noise suppression apparatus 1000 of the present invention. [0031] With this configuration, there is a remarkable effect that various residual noises generated in the multi-microphone system 2000 are almost completely suppressed. [0032] As described above, according to the present invention, a noise suppression apparatus that uses an acoustic signal in which noise and a target signal are mixed as an input signal and suppresses only noise from the input signal analyzes the frequency of the input signal. Frequency analysis circuit for extracting the power spectrum component and the phase component of the signal, the target signal / noise discrimination circuit for discriminating whether the input signal is the target signal or the other noise, and the average noise from the signal discrimination result A power spectrum is calculated, and a noise spectrum characteristic calculation and storage circuit for acoustically weighting the average power spectrum characteristic, and the acoustically 10-04-2019 9 weighted average power spectrum is subtracted from the power spectrum of the input signal Since it consists of subtraction means and an inverse frequency analysis circuit which converts the result back to the time domain, Suppressing signal distortion to a minimum, yet it is possible to greatly reduce the residual noise to his aurally ear. [0033] Also, an average noise level storage circuit for storing the residual noise level to further reduce residual noise remaining as a result of subtracting the power spectrum of noise from the power spectrum of the input signal, and residual noise, Since it has a loss control coefficient calculation circuit for calculating a loss control coefficient of noise and an operation means for controlling the loss of the output signal of the inverse frequency analysis circuit from the loss control coefficient, the weight coefficient alone can not suppress Residual noise can also be suppressed almost completely. [0034] As described above, according to the present invention, noise can be efficiently removed by processing residual noise that could not be completely removed so as to make it difficult for the user to hear aurally in the noise suppression device according to the conventional method. A noise suppressor can be realized that can be used in a comfortable state. [0035] Brief description of the drawings [0036] 1 is a block diagram showing the configuration of an embodiment of the present invention. [0037] It is a figure which shows an example of the average spectrum characteristic of the noise signal identified using FIG. 2Rmax. [0038] 3 is a block diagram illustrating an example of the use of the present invention. [0039] 10-04-2019 10 4 is a block diagram showing the configuration of the noise suppression device in the conventional method. [0040] Explanation of sign [0041] 201 signal input terminal 202 frequency analysis circuit 203 linear prediction analysis circuit 204 autocorrelation analysis circuit 205 maximum value selection circuit 206 voice / non-speech discrimination circuit 207A switch 207B switch 208 noise spectrum characteristic calculation and storage circuit 209 subtraction means 210 inverse frequency analysis circuit 211 Average noise level storage circuit 212 Loss control coefficient calculation circuit 213 Output signal calculation circuit 214 Operation means 215 Signal output terminal 10-04-2019 11

1/--страниц