JP4230414B2 - Sound signal processing method and sound signal processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and an apparatus to process the inputted sound signals which include degraded sound such as quantizing noise so that the degraded sound is subjectively unperceptible. <P>SOLUTION: The spectrum of a decoded sound after performing auditory weighting to the decoded sound as the input sound signals, is computed and transformation strength is calculated on the basis of the magnitude of the amplitude and the continuity of the spectrum. In a signal transformation section, the spectrum of the decoded sound is obtained, amplitude smoothing and phase disturbance adding are conducted on the basis of the transformation strength and the spectrum is returned to a signal region to provide transformed and decoded voice. In a signal evaluation section, the decoded sound is analyzed to obtain background noise likeness and the obtained value is made an addition control value. In a weighted value adding section, the weight for adding to the decoded sound is reduced, the weight for adding to the transformed sound is increased, and an output sound is obtained, when the addition control value indicates the background noise likeness. <P>COPYRIGHT: (C)2004,JPO&amp;NCIPI

Description

  The present invention makes it difficult to subjectively perceive subjectively undesirable components such as quantization noise generated by encoding / decoding processing such as voice and musical sound and distortion generated by various signal processing processing such as noise suppression processing. The present invention relates to a sound signal processing method and a sound signal processing apparatus.

  Increasing the compression rate of information source coding such as voice and musical sound will gradually increase the quantization noise that is distortion at the time of coding, and the quantization noise will be deformed and subjectively unbearable. Come. For example, in the case of a speech coding scheme that faithfully represents a signal itself such as PCM (Pulse Code Modulation) or ADPCM (Adaptive Differential Pulse Code Modulation), the quantization noise is a random number. Yes, but not too much subjectively, but as the compression rate increases and the coding scheme becomes more complex, the spectral characteristics unique to the coding scheme appear in the quantization noise, which can be subjectively degraded. Come. In particular, in a signal section in which background noise is dominant, the speech model used by the speech coding method with a high compression rate does not match, so the sound becomes very hard to hear.

  In addition, when noise suppression processing such as the spectral subtraction method is performed, the noise estimation error remains as distortion on the processed signal, and this has characteristics that are significantly different from the signal before processing, so the subjective evaluation is greatly degraded. Sometimes.

  As conventional methods for suppressing a decrease in subjective evaluation due to quantization noise and distortion as described above, JP-A-8-130513, JP-A-8-146998, JP-A-7-160296, JP-A-6-326670 are disclosed. JP-A-7-248793, and S.A. F. By Boll RactionSSP-27, No. 2, pp. 113-120, April 1979) (hereinafter referred to as Document 1).

  Japanese Patent Laid-Open No. 8-130513 is for the purpose of improving the quality of the background noise section. It is determined whether or not the section is a background noise only section, and a dedicated encoding process or decoding is performed for the background noise only section. In the case of performing processing and decoding the section of only background noise, the characteristic of the synthesis filter is suppressed to obtain an acoustically natural reproduced sound.

  Japanese Patent Laid-Open No. 8-146998 adds white noise or prestored background noise to decoded speech in order to suppress white noise from becoming a harsh tone by encoding and decoding. It is a thing.

  Japanese Patent Application Laid-Open No. 7-160296 aims to audibly reduce quantization noise and obtains an auditory masking threshold value based on decoded speech or an index related to a spectral parameter received by a speech decoding unit and reflects this. A filter coefficient is obtained, and this coefficient is used for the post filter.

  Japanese Patent Laid-Open No. 6-326670 generates and outputs pseudo background noise on the decoding side when there is no code transmission in a system that stops code transmission in a section that does not include speech for communication power control or the like. The aim is to reduce the sense of discomfort between the actual background noise included in the speech section and the pseudo background noise in the silence section. The pseudo background noise is superimposed not only on the section that does not contain speech but also on the speech section. It is what you do.

  Japanese Patent Application Laid-Open No. 7-248793 is intended to audibly reduce distortion sound generated by noise suppression processing. The encoding side first determines whether it is a noise interval or a speech interval, and in the noise interval a noise spectrum is determined. In the speech section, the spectrum after noise suppression processing is transmitted, and on the decoding side, a synthesized sound is generated and output using the received noise spectrum in the noise section, and after received noise suppression processing in the speech section The synthesized sound generated using the above spectrum is added to the synthesized sound generated using the noise spectrum received in the noise section, multiplied by the superposition magnification, and output.

Reference 1 aims to audibly reduce the distortion sound generated by the noise suppression processing, and smoothes the output speech after the noise suppression processing in terms of time intervals and amplitude spectra, Amplitude suppression processing is performed only in the background noise section.
JP-A-8-130513 JP-A-8-146998 JP 7-160296 A JP-A-6-326670 JP-A-7-248793 S. F. By Boll RactionSSP-27, No. 2, pp. 113-120, April 1979)

  The above conventional methods have the following problems.

  Japanese Patent Laid-Open No. 8-130513 has a problem that characteristics change suddenly at the boundary between the noise section and the voice section because the encoding process and the decoding process are largely switched according to the section determination result. In particular, when frequent misjudgment of a noise section as a speech section frequently occurs, the noise section, which is relatively stationary in nature, fluctuates in an unstable manner, which may cause deterioration of the noise section. When transmitting a noise section determination result, it is necessary to add information for transmission, and there is a problem that causes unnecessary degradation when the information is erroneous on the transmission path. Further, only suppressing the characteristics of the synthesis filter does not reduce the quantization noise generated at the time of sound source coding, and there is a problem that an improvement effect is hardly obtained depending on the noise type.

  In Japanese Patent Laid-Open No. 8-146998, there is a problem that the characteristic of the current encoded background noise is lost because the noise prepared in advance is added. In order to make it difficult to hear the deteriorated sound, it is necessary to add noise at a level exceeding the deteriorated sound, and there is a problem that the background noise to be reproduced becomes large.

  In Japanese Patent Laid-Open No. 7-160296, an auditory masking threshold value is obtained based on a spectrum parameter, and a spectrum post-filter is simply performed on the basis of the threshold value. There are almost no issues that cannot be improved at all. Moreover, since the main component which is not masked cannot give a big change, there exists a subject from which the improvement effect is not acquired about the distortion contained in the main component.

  In Japanese Patent Laid-Open No. 6-326670, pseudo background noise is generated regardless of actual background noise, and thus there is a problem that characteristics of actual background noise are lost.

  In Japanese Patent Laid-Open No. 7-248793, encoding and decoding processes are largely switched according to the section determination result, and therefore there is a problem that causes a large deterioration if a determination is made whether the section is a noise section or a speech section. If a part of the noise section is mistaken as a voice section, the sound quality in the noise section varies discontinuously and becomes difficult to hear. Conversely, if the speech segment is mistaken as a noise segment, speech components are mixed into the synthesized speech of the noise segment using the average noise spectrum and the synthesized speech using the noise spectrum superimposed in the speech segment. There is a problem that the sound quality deteriorates. Furthermore, it is necessary to superimpose noise that is not small so as not to hear the degraded sound in the voice section.

  Document 1 has a problem that processing delay of half a section (about 10 ms to 20 ms) occurs for smoothing. In addition, when a part of the noise section is erroneously determined as the voice section, there is a problem that the sound quality in the noise section varies discontinuously and becomes difficult to hear.

  The present invention has been made in order to solve the above-described problem. Actual background noise is less deteriorated due to a section determination error, less dependent on noise types and spectrum shapes, and does not require a large delay time. Sound signal processing that can retain the characteristics, does not excessively increase the background noise level, does not require the addition of new transmission information, and can provide a good suppression effect on degradation components due to sound source coding, etc. It is an object to provide a method and a sound signal processing apparatus.

  The input sound signal is processed to generate a first processed signal, the input sound signal is analyzed to calculate a predetermined evaluation value, and the input sound signal and the first processed signal are calculated based on the evaluation value. Weighted addition is performed to obtain a second processed signal, and this second processed signal is used as an output signal.

  Furthermore, in the first processed signal generation method, a spectral component for each frequency is calculated by performing a Fourier transform on the input sound signal, and a predetermined spectral component is calculated for the spectral component for each frequency calculated by the Fourier transform. It is characterized in that a deformation is given and a spectral component after the deformation is generated by inverse Fourier transform.

  Furthermore, the weighted addition is performed in a spectral region.

  Furthermore, the weighted addition is controlled independently for each frequency component.

  Further, the predetermined deformation of the spectral component for each frequency includes an amplitude spectral component smoothing process.

  Further, the predetermined deformation of the spectral component for each frequency includes a disturbance applying process for the phase spectral component.

  Furthermore, the smoothing intensity in the smoothing process is controlled by the magnitude of the amplitude spectrum component of the input sound signal.

  Furthermore, the disturbance applying intensity in the disturbance applying process is controlled by the magnitude of the amplitude spectrum component of the input sound signal.

  Further, the smoothing intensity in the smoothing process is controlled by the magnitude of continuity in the time direction of the spectral components of the input sound signal.

  Furthermore, the disturbance applying intensity in the disturbance applying process is controlled by the magnitude of continuity in the time direction of the spectral components of the input sound signal.

  Furthermore, an auditory weighted input sound signal is used as the input sound signal.

  Furthermore, the smoothing intensity in the smoothing process is controlled by the magnitude of temporal variability of the evaluation value.

  Furthermore, the disturbance applying intensity in the disturbance applying process is controlled by the magnitude of temporal variability of the evaluation value.

  Furthermore, the degree of background noise likelihood calculated by analyzing the input sound signal is used as the predetermined evaluation value.

  Furthermore, the degree of friction sound likelihood calculated by analyzing the input sound signal is used as the predetermined evaluation value.

  Furthermore, the input sound signal is characterized by using a decoded voice obtained by decoding a voice code generated by a voice encoding process.

  In the sound signal processing method of the present invention, the input sound signal is used as a first decoded sound obtained by decoding a sound code generated by a sound encoding process, and a post-filter process is performed on the first decoded sound to obtain a second decoded sound. The decoded speech is generated, the first decoded speech is processed to generate a first processed speech, one of the decoded speech is analyzed to calculate a predetermined evaluation value, and based on the evaluation value, the The second decoded sound and the first processed sound are weighted and added to form a second processed sound, and the second processed sound is output as an output sound.

  The sound signal processing apparatus according to the present invention includes a first processed signal generation unit that processes an input sound signal to generate a first processed signal, and an evaluation value that analyzes the input sound signal and calculates a predetermined evaluation value A calculation unit, and a second processed signal generation unit that weights and adds the input sound signal and the first processed signal based on the evaluation value of the evaluation value calculating unit, and outputs the second processed signal as a second processed signal. It is characterized by having.

  Further, the first processed signal generation unit calculates a spectral component for each frequency by performing a Fourier transform on the input sound signal, and smoothes an amplitude spectral component with respect to the calculated spectral component for each frequency. And a first processed signal is generated by performing inverse Fourier transform on the spectrum component after the amplitude spectrum component is smoothed.

  Further, the first processed signal generation unit calculates a spectral component for each frequency by performing a Fourier transform on the input sound signal, and a disturbance of a phase spectral component with respect to the calculated spectral component for each frequency. A first processing signal is generated by applying an applying process and performing an inverse Fourier transform on the spectral component after the phase spectral component is subjected to the disturbance applying process.

  As described above, the sound signal processing method and the sound signal processing apparatus according to the present invention perform predetermined signal processing on the input signal so that the deterioration component included in the input signal is not subjectively concerned. The processed signal is generated, and the addition weight of the input signal and the processed signal is controlled by the predetermined evaluation value, so the subjective quality is improved by increasing the ratio of the processed signal mainly in the section that contains many deterioration components There is an effect that can be done.

  In addition, the conventional binary section judgment is eliminated, the evaluation value of the continuous quantity is calculated, and the weighted addition coefficient of the input signal and the processed signal can be controlled continuously based on this, thereby avoiding quality degradation due to the section judgment error. There is an effect that can be done.

  In addition, since the output signal can be generated by processing the input signal that contains a lot of background noise information, it has a stable quality improvement effect that does not depend much on the noise type and spectrum shape while retaining the actual background noise characteristics. In addition, there is an effect that an improvement effect can be obtained even for a deteriorated component due to excitation coding or the like.

  In addition, since processing can be performed using the input signals up to now, there is no need for a particularly large delay time, and there is an effect that delays other than the processing time can be eliminated depending on the method of adding the input signal and the processed signal. When increasing the level of the processed signal, if the input signal level is lowered, it is not necessary to superimpose a large pseudo noise to mask the degradation component as in the conventional case. Thus, the background noise level can be reduced or even increased. As a matter of course, even when the degraded sound due to speech coding / decoding is eliminated, it is not necessary to add new transmission information as in the conventional case.

  The sound signal processing method and sound signal processing apparatus according to the present invention perform predetermined processing in the spectral region on the input signal so that the deterioration component included in the input signal is not subjectively concerned. Since the processing signal is generated and the addition weight of the input signal and the processing signal is controlled by a predetermined evaluation value, in addition to the effect of the signal processing method, the fine degradation component suppression processing in the spectral region is performed. Can also improve the subjective quality.

  In the sound signal processing method of the present invention, since the input signal and the processed signal are weighted and added in the spectral region in the sound signal processing method of the above invention, in addition to the effects of the sound signal processing method, For example, when connecting to the subsequent stage of the noise suppression method that performs the above processing, part or all of the Fourier transform processing and inverse Fourier transform processing required by the sound signal processing method can be omitted, and the processing can be simplified. is there.

  In the sound signal processing method of the present invention, the weighted addition is controlled independently for each frequency component in the sound signal processing method of the above invention, so that in addition to the effects of the sound signal processing method, quantization noise and The dominant component of the deteriorated component is replaced with the processed signal mainly, and it is no longer replaced with a good component with little quantization noise and deteriorated component. The degradation component can be subjectively suppressed, and the subjective quality can be improved.

  In the sound signal processing method of the present invention, since the amplitude spectrum component is smoothed as the processing in the sound signal processing method of the above invention, in addition to the effects of the sound signal processing method, quantization noise is added. It is possible to satisfactorily suppress unstable fluctuations in the amplitude spectrum component caused by the above, and to improve the subjective quality.

  In the sound signal processing method according to the present invention, since the disturbance processing of the phase spectrum component is performed as the processing in the sound signal processing method of the above invention, in addition to the effects of the sound signal processing method, This has the effect of improving the subjective quality by disturbing the relationship between the phase components against the quantization noise and the degradation components that often have a characteristic correlation with each other. is there.

  In the sound signal processing method of the present invention, the smoothing strength or the disturbance imparting strength in the sound signal processing method of the above invention is controlled by the magnitude of the amplitude spectrum component of the input signal or the auditory weighted input signal. In addition to the effects of the sound signal processing method, quantization noise and deterioration components are mainly applied to the components where quantization noise and deterioration components are dominant because the amplitude spectrum component is small. Therefore, it is possible to subjectively suppress the quantization noise and the degradation component while leaving the characteristics of the input signal good, and to improve the subjective quality.

  In the sound signal processing method of the present invention, the smoothing strength or the disturbance imparting strength in the sound signal processing method of the above invention is controlled by the magnitude of continuity in the time direction of the spectral components of the input signal or the auditory weighted input signal. Therefore, in addition to the effects of the sound signal processing method described above, the continuity of the spectral components is low, so that processing that is prone to increase quantization noise and deterioration components is processed with a focus on quantization. There is no need to process a good component with less noise and deterioration components, and it is possible to subjectively suppress quantization noise and deterioration components while maintaining good characteristics of the input signal, thereby improving the subjective quality.

  In the sound signal processing method of the present invention, the smoothing strength or the disturbance imparting strength in the sound signal processing method of the above invention is controlled by the magnitude of temporal variability of the evaluation value. In addition to the effects, it is possible to suppress unnecessarily strong processing in a section where the characteristics of the input signal are fluctuating, and in particular, it is possible to prevent the occurrence of slack and echo due to amplitude smoothing.

  Since the sound signal processing method of the present invention uses the degree of background noise likelihood as the predetermined evaluation value in the sound signal processing method of the above invention, in addition to the effects of the sound signal processing method, quantization noise and Focused processing is added to the background noise section that tends to generate many degradation components, and appropriate processing (not processed, low-level processing, etc.) is selected for the section other than the background noise. Therefore, there is an effect that the subjective quality can be improved.

  Since the sound signal processing method of the present invention uses the degree of frictional sound likelihood as the predetermined evaluation value in the sound signal processing method of the above invention, in addition to the effects of the sound signal processing method, quantization noise and Focusing processing is applied to the frictional sound section that tends to generate many deterioration components, and appropriate processing (not processing, low-level processing, etc.) is selected for the other sections than the frictional sound. It has the effect of improving subjective quality.

  The sound signal processing method according to the present invention uses the sound code generated by the sound encoding process as an input, decodes the sound code to generate a decoded sound, and uses the decoded sound as an input to use the sound signal processing method. Since the processed speech is generated by performing the signal processing, and this processed speech is output as the output speech, the speech decoding with the subjective quality improvement effect etc. possessed by the sound signal processing method as it is can be realized. is there.

  In the sound signal processing method of the present invention, the speech code generated by the speech encoding process is input, the speech code is decoded to generate a decoded speech, and the decoded speech is subjected to a predetermined signal processing to produce the processed speech. Generate and perform post-filter processing on the decoded speech, further analyze the decoded speech before or after the post-filter, calculate a predetermined evaluation value, and weight the decoded speech after the post-filter and the processed speech based on this evaluation value Since it is added and output, in addition to the effect that speech decoding with the subjective quality improvement effect etc. of the sound signal processing method as it is is realized, processed speech that is not affected by the post filter can be generated, and the post filter Since the addition weight control with high accuracy can be performed based on the highly accurate evaluation value calculated without being influenced by the above, there is an effect of further improving the subjective quality.

  Embodiments of the present invention will be described below with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 shows the overall structure of a speech decoding method to which a sound signal processing method according to this embodiment is applied. In FIG. 1, 1 is a speech decoding apparatus, 2 is a signal processing unit for executing the signal processing method according to the present invention, A voice code, 4 is a voice decoding unit, 5 is a decoded voice, and 6 is an output voice. The signal processing unit 2 includes a signal transformation unit 7, a signal evaluation unit 12, and a weighted addition unit 18. The signal transformation unit 7 includes a Fourier transform unit 8, an amplitude smoothing unit 9, a phase disturbance unit 10, and an inverse Fourier unit 11. The signal evaluation unit 12 includes an inverse filter unit 13, a power calculation unit 14, a background noise likelihood calculation unit 15, an estimated background noise power update unit 16, and an estimated noise spectrum update unit 17.

  Hereinafter, the operation will be described with reference to the drawings.

  First, the speech code 3 is input to the speech decoding unit 4 in the speech decoding device 1. The speech code 3 is output as a result of separately encoding the speech signal by the speech encoding unit, and is input to the speech decoding unit 4 via a communication path or a storage device.

  The speech decoding unit 4 performs a decoding process that forms a pair with the speech encoding unit on the speech code 3 and outputs the obtained signal having a predetermined length (1 frame length) as the decoded speech 5. The decoded speech 5 is input to the signal transformation unit 7, the signal evaluation unit 12, and the weighted addition unit 18 in the signal processing unit 2.

  The Fourier transform unit 8 in the signal transformation unit 7 performs windowing on the combined signal of the input decoded speech 5 of the current frame and the latest part of the decoded speech 5 of the previous frame, if necessary. A spectral component for each frequency is calculated by performing a Fourier transform process on the signal, and this is output to the amplitude smoothing unit 9. Typical Fourier transform processing includes discrete Fourier transform (DFT), fast Fourier transform (FFT), and the like. Various methods such as trapezoidal windows, rectangular windows, and Hanning windows can be used for windowing, but here we use a modified trapezoidal window in which the slopes at both ends of the trapezoidal window are replaced with half of the Hanning window. To do. The actual shape example and the time relationship with the decoded speech 5 and the output speech 6 will be described later with reference to the drawings.

  The amplitude smoothing unit 9 performs a smoothing process on the amplitude component of the spectrum for each frequency input from the Fourier transform unit 8, and outputs the smoothed spectrum to the phase disturbance unit 10. As the smoothing process used here, the effect of suppressing degraded sound such as quantization noise can be obtained regardless of whether the frequency axis direction or the time axis direction is used. However, if the smoothing in the frequency axis direction is too strong, the spectrum will be neglected and the characteristics of the original background noise are often impaired. On the other hand, if the smoothing in the time axis direction is made too strong, the same sound remains for a long time, resulting in a feeling of echo. As a result of proceeding with adjustments for various background noises, the quality of the output sound 6 was good when smoothing in the frequency axis direction was not performed and amplitude was smoothed in the logarithmic region in the time axis direction. The smoothing method at that time is expressed by the following equation.

y _i = y _i−1 (1−α) + x _i α Equation 1
Here, x _i is a logarithmic amplitude spectrum value before smoothing of the current frame (i-th frame), y _i-1 is a logarithmic amplitude spectrum value after smoothing of the previous frame (i-th frame), and y _i. Is a logarithmic amplitude spectrum value of the current frame (i-th frame) after smoothing, α is a smoothing coefficient having a value of 0 to 1, smoothing coefficient α depends on the frame length, the level of degraded sound to be eliminated, etc. Although the optimum value is different, the value is about 0.5.

  The phase disturbance unit 10 gives a disturbance to the phase component of the smoothed spectrum input from the amplitude smoothing unit 9 and outputs the spectrum after the disturbance to the inverse Fourier transform unit 11. As a method for giving disturbance to each phase component, a phase angle in a predetermined range may be generated using a random number, and this may be added to the original phase angle. In the case where the range of phase angle generation is not limited, each phase component may be simply replaced with a phase angle generated with a random number. When the deterioration due to encoding or the like is large, the range of phase angle generation is not limited.

  The inverse Fourier transform unit 11 performs inverse Fourier transform processing on the spectrum after disturbance input from the phase disturbance unit 10 to return to the signal region, and opens a window for smooth connection with the preceding and following frames. The connection is performed while performing, and the obtained signal is output to the weighted addition unit 18 as the modified decoded speech 34.

  The inverse filter unit 13 in the signal evaluation unit 12 uses an estimated noise spectrum parameter stored in an estimated noise spectrum update unit 17 to be described later, and performs an inverse filter process on the decoded speech 5 input from the speech decoding unit 4. Then, the decoded speech that has been inversely filtered is output to the power calculation unit 14. By this inverse filter processing, the amplitude of the background noise is large, that is, the amplitude suppression of the component that is highly likely to antagonize the speech and the background noise, and compared with the case where the inverse filter processing is not performed, The signal power ratio in the background noise section can be increased.

  Note that the estimated noise spectrum parameter is selected from the viewpoint of compatibility with speech encoding processing and speech decoding processing, and software sharing. Currently, line spectral pairs (LSPs) are often used. In addition to the LSP, similar effects can be obtained by using spectral envelope parameters such as linear prediction coefficients (LPC), cepstrum, or the amplitude spectrum itself. As the update process in the estimated noise spectrum update unit 17 to be described later, a configuration using linear interpolation or average process is simple, and it is guaranteed that the filter is stable even if linear interpolation or average process is performed in the spectrum envelope parameter. Suitable LSP and cepstrum are suitable. The cepstrum is excellent as a power of expressing the noise component spectrum, but the LSP is superior in terms of the ease of configuration of the inverse filter unit. In the case of using an amplitude spectrum, LPC having this amplitude spectrum characteristic is calculated and used for an inverse filter, or an amplitude transformation process is performed on the result of Fourier transform of the decoded speech 5 (equal to the output of the Fourier transform unit 8). To achieve the same effect as the inverse filter.

  The power calculation unit 14 obtains the power of the decoded speech that has been subjected to inverse filtering input from the inverse filter unit 13, and outputs the calculated power value to the background noise likelihood calculation unit 15.

  The background noise likelihood calculation unit 15 uses the power input from the power calculation unit 14 and the estimated noise power stored in the estimated noise power update unit 16 described later to calculate the background noise likelihood of the current decoded speech 5. This is calculated and output to the weighted addition unit 18 as the addition control value 35. The calculated background noise likelihood is output to an estimated noise power updating unit 16 and an estimated noise spectrum updating unit 17 described later, and the power input from the power calculating unit 14 is output to an estimated noise power updating unit 16 described later. Output. Here, the likelihood of background noise can be calculated most simply by the following equation.

v = log (p _N ) −log (p) Equation 2
Here, p is the power input from the power calculation unit 14, p _N is the estimated noise power stored in the estimated noise power update unit 16, and v is the calculated background noise likelihood.

In this case, the greater the value of v (the smaller the negative value, the smaller the absolute value), the more likely it is background noise. In addition, various calculation methods such as calculating p _N / p and setting it to v are conceivable.

  The estimated noise power update unit 16 updates the estimated noise power stored therein using the background noise likelihood and power input from the background noise likelihood calculation unit 15. For example, when the input background noise likelihood is high (the value of v is large), updating is performed by reflecting the input power to the estimated noise power according to the following equation.

log (p _N ′) = (1−β) log (p _N ) + βlog (p) Equation 3
Here, β is an update rate constant that takes a value of 0 to 1, and may be set to a value relatively close to 0. The value on the right side of this equation is obtained, and updating is performed by using p _N 'on the left side as a new estimated noise power.

  As for the method of updating the estimated noise power, refer to the variability between frames in order to further improve the estimation accuracy, or store a plurality of input past powers and perform noise analysis by statistical analysis. Various modifications and improvements are possible, such as estimation of the above, or using the minimum value of p as the estimated noise power as it is.

  The estimated noise spectrum updater 17 first analyzes the input decoded speech 5 and calculates the spectrum parameter of the current frame. The spectrum parameter to be calculated is as described in the inverse filter unit 13, and in many cases LSP is used. Then, the estimated noise spectrum stored inside is updated by using the background noise likelihood input from the background noise likelihood calculation unit 15 and the spectrum parameter calculated here. For example, when the input background noise likelihood is high (the value of v is large), updating is performed by reflecting the calculated spectrum parameter in the estimated noise spectrum according to the following equation.

_{x N '= (1-γ} ) x N + γx ··· formula 4
Here, x is from the spectrum parameter of the current frame, x _N is the estimated noise spectrum (parameter). γ is an update rate constant that takes a value from 0 to 1, and may be set to a value relatively close to 0. The value on the right side of this equation is obtained, and updating is performed by setting x _N ′ on the left side as a new estimated noise spectrum (parameter).

  It should be noted that various improvements can be made to the method for updating the estimated noise spectrum as in the method for updating the estimated noise power.

  As a final process, the weighted addition unit 18 receives the decoded speech 5 input from the speech decoding unit 4 and the signal transformation unit 7 based on the addition control value 35 input from the signal evaluation unit 12. The modified decoded speech 34 is weighted and added, and the resulting output speech 6 is output. As the operation of the weighted addition control method, the weight for the decoded speech 5 is decreased and the weight for the modified decoded speech 34 is increased as the addition control value 35 becomes larger (likely background noise). Conversely, the weight for the decoded speech 5 is increased and the weight for the modified decoded speech 34 is decreased as the addition control value 35 becomes smaller (likely background noise is low).

  In addition, in order to suppress the quality deterioration of the output sound 6 due to a sudden change in weight between frames, it is desirable to perform smoothing so that the addition control value 35 or the weighting coefficient is gradually changed for each sample.

  FIG. 2 shows an example of weighted addition control based on the addition control value in the weighted addition unit 18.

FIG. 2A shows a case where linear control is performed using _two threshold values v ₁ and v ₂ for the addition control value 35. When the addition control value 35 is less than v ₁ , the weighting coefficient w _S for the decoded speech 5 is set to 1, and the weighting coefficient w _N for the modified decoded speech 34 is set to 0. When the addition control value 35 is v ₂ or more, the weighting coefficient w _S for the decoded speech 5 is set to 0, and the weighting coefficient w _N for the modified decoded speech 34 is set to A _N. And when the addition control value 35 is _v less than _{2 v 1} or more, linear weighting coefficients _{w S} for the decoded speech 5 1-0, the weighting coefficient _{w N} to deformation decoded speech 34 between 0 to A _N Is calculated and given.

By controlling in this way, when it can be reliably determined that it is the background noise section (v ₂ or more), only the modified decoded signal 34 is output, and when it can be determined that it is surely the speech section (less than v ₁ ). In the case where the decoded speech 5 itself is output and it is not possible to determine whether it is the speech interval or the background noise interval (v ₁ or more and less than v ₂ ), the decoded speech 5 and the modified decoding are proportionately depending on which tendency is strong. The result of mixing the sound 34 is output.

Note that if you give a value of 1 or less as the weighting coefficient value A _N for multiplying the modified decoded signal 34 when (v ₂ or more) which can be determined that where a strictly background noise period, resulting in the amplitude suppression of the background noise period An effect is obtained. Conversely, if a value of 1 or more is given, the effect of enhancing the amplitude of the background noise section can be obtained. In the background noise section, the amplitude is often lowered by the speech coding / decoding process. In this case, the background noise section can be improved in amplitude by enhancing the amplitude of the background noise section. Whether amplitude suppression or amplitude enhancement is performed depends on the application target, the user's request, and the like.

FIG. 2B shows a case where a new threshold value v ₃ is added and weighting coefficients are linearly calculated and given between v ₁ and v _{3 and} between v ₃ and v ₂ . By adjusting the value of the weighting coefficient at the position of the threshold value v ₃ , it is possible to set the mixing ratio more finely when it is not possible to determine whether it is a speech section or a background noise section (v ₁ or more and less than v ₂ ). In general, when two signals having a low phase correlation are added, the power of the obtained signal is smaller than the sum of the powers of the two signals before the addition. By making the sum of the two weighting coefficients in the range of v ₁ or more and less than v ₂ greater than 1 to w _N , this power reduction can be suppressed. The same effect can be obtained by taking the square root of the weighting coefficient obtained in FIG. 2A and further multiplying by a constant to obtain a new weighting coefficient.

In FIG. 2C, the weighting coefficient w _N given to the modified decoded speech 34 in the range of less than v _{1 in} FIG. 2A is given a value of B _N greater than 0 and v ₁ or more and less than v ₂ accordingly. This is a case where w _N in the range is also corrected. If the background noise level is high or the compression rate in encoding is very high, such as when the quantization noise or degraded sound in the speech section is large, even in the range that is surely known as the speech section, By adding the modified decoded speech, it is possible to make it difficult to hear the degraded sound.

FIG. 2D shows a control corresponding to the case where the background noise likelihood calculation unit 15 outputs the result (p _N / p) obtained by dividing the estimated noise power by the current power as the background noise likelihood (addition control value 35). It is an example. In this case, since the addition control value 35 indicates the ratio of the background noise included in the decoded speech 5, the weighting coefficient is calculated so as to be mixed at a ratio proportional to this value. Specifically, when the addition control value 35 is 1 or more, when w _N is 1 and w _S is 0 or less than 1, w _N is the addition control value 35 itself, and w _S is (1−w _N ).

    FIG. 3 is an explanatory diagram for explaining a temporal relationship with the cutout window in the Fourier transform unit 8, an actual shape example of the window for connection in the inverse Fourier transform unit 11, and the decoded speech 5.

    The decoded speech 5 is output from the speech decoding unit 4 every predetermined time length (one frame length). Here, this one frame length is assumed to be N samples. FIG. 3A shows an example of the decoded speech 5, which corresponds to the decoded speech 5 of the current frame in which x (0) to x (N-1) are input. The Fourier transform unit 8 cuts out a signal of length (N + NX) by multiplying the decoded speech 5 shown in FIG. 3A by a modified trapezoidal window shown in FIG. 3B. NX is the length of each section having a value less than 1 at both ends of the modified trapezoidal window. The section at both ends is equal to a Hanning window having a length (2NX) divided into two parts, the first half and the second half. The inverse Fourier transform unit 11 multiplies the signal generated by the inverse Fourier transform process by a modified trapezoidal window shown in FIG. 3C and obtains it in the previous and subsequent frames (as indicated by the broken line in FIG. 3C). The signals are added while maintaining the time relationship with the received signals, and a continuous modified decoded speech 34 (FIG. 3D) is generated.

  In the section (length NX) for connection with the signal of the next frame, the modified decoded speech 34 is not fixed at the time of the current frame. That is, the newly determined modified decoded speech 34 is x ′ (− NX) to x ′ (N−NX−1). For this reason, the output audio | voice 6 obtained with respect to the decoding audio | voice 5 of the present flame | frame becomes following Formula.

y (n) = x (n) + x ′ (n) (5)
(N = -NX, ..., N-NX-1)
Here, y (n) is the output sound 6. At this time, the processing delay as the signal processing unit 2 is required at least NX.

  In the case where the processing delay NX is not acceptable, the output sound 6 can be generated as shown in the following equation while allowing the time difference between the decoded sound 5 and the modified decoded sound 34.

y (n) = x (n) + x ′ (n−NX) Expression 6
(N = 0, ..., N-1)
In this case, since there is a difference in the time relationship between the decoded speech 5 and the modified decoded speech 34, the disturbance in the phase disturbance unit 10 is weak (that is, the phase characteristics of the decoded speech remain to some extent), or the spectrum and power in the frame Degradation may occur when the value changes suddenly. In particular, deterioration is likely to occur when the weighting coefficient in the weighted addition unit 18 changes greatly and when the two weighting coefficients are antagonizing. However, their deterioration is relatively small, and the effect of introducing the signal processing part is sufficiently large. Therefore, this method can also be used for an application target in which the processing delay NX is not allowed.

  In the case of FIG. 3, the modified trapezoidal window is multiplied before the Fourier transform and after the inverse Fourier transform, and the amplitude of the connected portion may be reduced. This decrease in amplitude is also likely to occur when the disturbance in the phase disturbance unit 10 is weak. In such a case, the amplitude reduction can be suppressed by changing the window before Fourier transform to a rectangular window. Normally, as a result of the phase being greatly deformed by the phase disturbance unit 10, the shape of the first deformed trapezoidal window does not appear in the signal after the inverse Fourier transform, so that smooth connection with the deformed decoded speech 34 of the previous and subsequent frames is achieved. A second window is required.

  Here, the processes of the signal transformation unit 7, the signal evaluation unit 12, and the weighting addition unit 18 are all performed for each frame, but the present invention is not limited to this. For example, one frame is divided into a plurality of subframes, the processing of the signal evaluation unit 12 is performed for each subframe to calculate the addition control value 35 for each subframe, and the weighting control in the weighting addition unit 18 is also performed for each subframe. You can go. Since the Fourier transform is used for the signal transformation process, if the frame length is too short, the analysis result of the spectral characteristics becomes unstable, and the modified decoded speech 34 is difficult to stabilize. On the other hand, since the likelihood of background noise can be calculated relatively stably even for a shorter section, the quality improvement effect at the rising portion of the voice or the like can be obtained by calculating for each subframe and finely controlling the weighting.

  It is also possible to calculate a small number of addition control values 35 by performing the processing of the signal evaluation unit 12 for each subframe and combining all the addition control values in the frame. When it is not desired to make an error when the speech section seems to be background noise, a minimum value (minimum value of background noise likelihood) of all the addition control values may be selected and output as an addition control value 35 representing a frame.

  Furthermore, the frame length of the decoded speech 5 and the processing frame length of the signal transformation unit 7 do not have to be the same. For example, if the decoded speech 5 has a short frame length and is too short for the spectrum analysis in the signal transformation unit 7, a plurality of frames of the decoded speech 5 can be accumulated so as to perform signal transformation processing collectively. good. However, in this case, a processing delay occurs because the decoded audio 5 of a plurality of frames is accumulated. In addition, the processing frame length of the entire signal transformation unit 7 and the signal processing unit 2 may be set completely independently of the frame length of the decoded speech 5. In this case, the buffering of the signal becomes complicated, but the optimum processing frame length for the signal processing can be selected without depending on the frame lengths of the various decoded speech 5, and the quality of the signal processing unit 2 becomes the best. effective.

    Here, the inverse filter unit 13, the power calculation unit 14, the background noise likelihood calculation unit 15, the estimated background noise level update unit 16, and the estimated noise spectrum update unit 17 are used to calculate the background noise likelihood. The structure is not limited to this structure as long as it evaluates the uniqueness.

  According to the first embodiment, by performing predetermined signal processing on the input signal (decoded speech), the processed signal (deformed decoding) that does not subjectively care about the degradation component included in the input signal. Voice), and the addition weight of the input signal and the processed signal is controlled by a predetermined evaluation value (likeness of background noise), so the ratio of the processed signal is increased mainly in a section containing a lot of deterioration components, It has the effect of improving subjective quality.

  In addition, since signal processing is performed in the spectral region, it is possible to perform processing for suppressing fine degradation components in the spectral region, and to further improve the subjective quality.

  In addition, since the amplitude spectrum component smoothing process and the phase spectrum component disturbance applying process are performed as the processing process, it is possible to satisfactorily suppress unstable fluctuations in the amplitude spectrum component caused by quantization noise, Furthermore, for quantization noise that often has a unique correlation between phase components and is often felt as a characteristic deterioration, the relationship between phase components can be disturbed, and the subjective quality can be improved. There is.

  Also, the conventional binary section determination of either the speech section or the background noise section is abolished, and a continuous amount of background noise likelihood is calculated, and based on this, the weighted addition coefficient of the decoded speech and the modified decoded speech is continuously calculated. Therefore, there is an effect that quality deterioration due to a section determination error can be avoided.

  In addition, when quantization noise or deteriorated sound is large in the speech section, it is possible to make it difficult to hear the deteriorated sound by adding the modified decoded speech even in a section that is surely known as the speech section. .

  In addition, since the output speech is generated by processing the decoded speech that contains a lot of background noise information, the quality of the output remains stable, while maintaining the actual background noise characteristics, and stable quality improvement that does not depend much on the noise type or spectrum shape. The effect is obtained, and there is an effect that an improvement effect can be obtained even for a deteriorated component due to excitation coding or the like.

  Further, since processing is performed using the decoded speech up to now, there is no need for a particularly large delay time, and there is an effect that delays other than the processing time can be eliminated depending on the method of adding the decoded speech and the modified decoded speech. When raising the level of the modified decoded speech, the level of the decoded speech is lowered, so there is no need to superimpose a large pseudo noise to make the quantization noise inaudible as in the past, and conversely depending on the application target Thus, the background noise level can be reduced or even increased. As a matter of course, since the process is closed in the speech decoding apparatus or the signal processing unit, it is not necessary to add new transmission information as in the prior art.

  Furthermore, in this Embodiment 1, since the speech decoding unit and the signal processing unit are clearly separated and there is little exchange of information between them, it is introduced into various speech decoding apparatuses including existing ones. Is easy.

Embodiment 2. FIG.
FIG. 4 shows a part of the configuration of a sound signal processing apparatus to which the sound signal processing method according to the present embodiment is applied in combination with a noise suppression method. In the figure, 36 is an input signal, 8 is a Fourier transform unit, 19 is a noise suppression unit, 39 is a spectrum transformation unit, 12 is a signal evaluation unit, 18 is a weighted addition unit, 11 is an inverse Fourier transform unit, and 40 is an output signal. is there. The spectrum modification unit 39 includes an amplitude smoothing unit 9 and a phase disturbance unit 10.
Hereinafter, the operation will be described with reference to the drawings.

  First, the input signal 36 is input to the Fourier transform unit 8 and the signal evaluation unit 12.

  The Fourier transform unit 8 performs windowing on the signal obtained by combining the input signal 36 of the input current frame and the latest part of the input signal 36 of the previous frame as necessary, and performs Fourier processing on the signal after windowing. By performing the conversion process, a spectrum component for each frequency is calculated and output to the noise suppression unit 19. The Fourier transform process and the windowing process are the same as those in the first embodiment.

  The noise suppression unit 19 subtracts the estimated noise spectrum stored in the noise suppression unit 19 from the spectrum component for each frequency input from the Fourier transform unit 8, and weights and adds the obtained result as the noise suppression spectrum 37. To the amplitude smoothing unit 9 in the unit 18 and the spectrum transformation unit 39. This is a process corresponding to the main part of the so-called spectral subtraction process. Then, the noise suppression unit 19 determines whether or not it is a background noise interval, and if it is a background noise interval, an internal estimated noise spectrum is calculated using a spectrum component for each frequency input from the Fourier transform unit 8. Update. Note that it is possible to simplify the process by determining whether or not the background noise section is in use by using the output result of the signal evaluation unit 12 described later.

  The amplitude smoothing unit 9 in the spectrum modification unit 39 performs a smoothing process on the amplitude component of the noise suppression spectrum 37 input from the noise suppression unit 19, and the smoothed noise suppression spectrum is supplied to the phase disturbance unit 10. Output. As the smoothing process used here, the effect of suppressing the deteriorated sound generated by the noise suppression unit can be obtained regardless of whether the frequency axis direction or the time axis direction is used. As a specific smoothing method, the same one as in the first embodiment can be used.

  The phase disturbance unit 10 in the spectrum modification unit 39 gives disturbance to the phase component of the smoothed noise suppression spectrum input from the amplitude smoothing unit 9, and weighted addition is performed with the spectrum after the disturbance as the modified noise suppression spectrum 38. To the unit 18. As a method for giving disturbance to each phase component, the same method as in the first embodiment can be used.

  The signal evaluation unit 12 analyzes the input signal 36 to calculate the likelihood of background noise, and outputs this to the weighted addition unit 18 as the addition control value 35. In addition, about the structure and each process in this signal evaluation part 12, the thing similar to Embodiment 1 can be used.

    Based on the addition control value 35 input from the signal evaluation unit 12, the weighted addition unit 18 uses the noise suppression spectrum 37 input from the noise suppression unit 19 and the modified noise suppression spectrum 38 input from the spectrum modification unit 39. The weighted addition is performed, and the obtained spectrum is output to the inverse Fourier transform unit 11. As the operation of the weighted addition control method, as in the first embodiment, the weight for the noise suppression spectrum 37 is decreased and the weight for the modified noise suppression spectrum 38 is increased as the addition control value 35 is increased (likely to be background noise). Greatly control. Conversely, as the addition control value 35 becomes smaller (the likelihood of background noise is lower), the weight for the noise suppression spectrum 37 is increased, and the weight for the modified noise suppression spectrum 38 is decreased.

  Then, as the last process, the inverse Fourier transform unit 11 performs the inverse Fourier transform process on the spectrum input from the weighted addition unit 18 to return to the signal region, and smooth connection with the previous and subsequent frames. For this purpose, the connection is performed while the window is opened, and the obtained signal is output as the output signal 40. The windowing and connection processing for connection are the same as in the first embodiment.

  According to the second embodiment, a predetermined processing is performed on a spectrum that has deteriorated due to noise suppression processing or the like, so that a processing spectrum (deformed noise suppression spectrum) that does not bother subjectively with the deterioration component is obtained. Generated and controlled the addition weight of the spectrum before processing and the processing spectrum by a predetermined evaluation value (likeness of background noise), so that there are many deterioration components and it leads to a decrease in subjective quality (background noise) There is an effect that the subjective quality can be improved by increasing the ratio of the processing spectrum centering on the section).

  In addition, since weighted addition in the spectral region is performed, the Fourier transform and inverse Fourier transform for processing are not required compared to the first embodiment, and the processing is simplified. Note that the Fourier transform unit 8 and the inverse Fourier transform 11 in the second embodiment are originally required for the noise suppression unit 19.

  In addition, since the amplitude spectrum component smoothing process and the phase spectrum component disturbance applying process are performed as the processing process, it is possible to satisfactorily suppress unstable fluctuations in the amplitude spectrum component caused by quantization noise, In addition, quantization noise and deterioration components, which often have a characteristic correlation between phase components and are often perceived as characteristic deterioration, can disturb the relationship between phase components, reducing subjective quality. There is an effect that can be improved.

  In addition, instead of the binary interval determination of whether or not it is a background noise interval, a continuous amount of background noise likelihood is calculated, and the weighted addition coefficient is continuously controlled based on this, so an interval determination error There is an effect that quality deterioration due to can be avoided.

  In addition, when the deteriorated sound outside the background noise section is large, weighted addition as shown in FIG. 2 (c) is performed, so that the modified noise suppression spectrum is added even in the section known to be outside the background noise section. There is an effect of making it difficult to hear the deteriorated sound.

  In addition, since a modified noise suppression spectrum is generated by directly performing a simple process on the noise suppression spectrum, there is an effect that a stable quality improvement effect that does not depend much on the noise type and spectrum shape can be obtained.

  Further, since processing is performed using the noise suppression spectrum up to now, there is a feature that a large delay time is not necessary in addition to the delay time of the noise suppression unit 19. When increasing the addition level of the modified noise suppression spectrum, the addition level of the original noise suppression spectrum is lowered, so it is not necessary to superimpose a relatively large noise in order to make the quantization noise inaudible. There is an effect that can be reduced. Of course, even when this process is used as a pre-process for the speech encoding process, it is a process closed in the encoding unit, so that it is not necessary to add new transmission information as in the prior art. .

Embodiment 3 FIG.
FIG. 5, in which the same reference numerals are assigned to the parts corresponding to those in FIG. 1, shows the overall configuration of the speech decoding apparatus to which the sound signal processing method according to this embodiment is applied. In FIG. It is a deformation intensity control part which outputs the information to do. The deformation intensity control unit 20 includes an auditory weighting unit 21, a Fourier transform unit 22, a level determination unit 23, a continuity determination unit 24, and a deformation intensity calculation unit 25.

  Hereinafter, the operation will be described with reference to the drawings.

  The decoded speech 5 output from the speech decoding unit 4 is input to the signal transformation unit 7, the transformation strength control unit 20, the signal evaluation unit 12, and the weighted addition unit 18 in the signal processing unit 2.

    The auditory weighting unit 21 in the deformation intensity control unit 20 performs auditory weighting processing on the decoded speech 5 input from the speech decoding unit 4, and outputs the obtained auditory weighted speech to the Fourier transform unit 22. Here, as auditory weighting processing, processing similar to that used in speech coding processing (which is paired with speech decoding processing performed in speech decoding unit 4) is performed.

  An auditory weighting process often used in an encoding process such as CELP is to analyze a speech to be encoded to calculate a linear prediction coefficient (LPC), and perform constant multiplication on this to obtain two modified LPCs. An ARMA filter having two modified LPCs as filter coefficients is configured, and auditory weighting is performed by a filtering process using this filter. In order to perform auditory weighting on the decoded speech 5 in the same manner as in the encoding process, an LPC obtained by decoding the received speech code 3 or an LPC calculated by reanalyzing the decoded speech 5 is used as a starting point. What is necessary is just to obtain | require two deformation | transformation LPC and comprise an auditory weighting filter using this.

  In an encoding process such as CELP, encoding is performed so as to minimize distortion on speech after auditory weighting. Therefore, in a speech component after auditory weighting, a spectral component having a large amplitude has little superposition of quantization noise. ,It turns out that. Therefore, if the speech close to the auditory weighted speech at the time of encoding can be generated in the decoding unit 1, it is useful as control information for the deformation intensity in the signal deformation unit 7.

  If the speech decoding process in the speech decoding unit 4 includes a processing process such as a spectrum post filter (mostly included in the case of CELP), the spectrum is first converted from the decoded speech 5 first. Encoding is performed by generating a sound from which the influence of processing such as a post filter has been removed or by extracting the sound immediately before the processing from the sound decoding unit 4 and performing auditory weighting on the sound. Sound close to the auditory weighted sound of the time can be obtained. However, when the main purpose is to improve the quality of the background noise section, the influence of processing such as a spectrum post filter in this section is small, and even if the influence is not removed, the effect does not vary greatly. In the third embodiment, the influence of processing such as a spectrum post filter is not removed.

  As a matter of course, the auditory weighting unit 21 is not necessary when auditory weighting is not performed in the encoding process or when the effect is small and can be ignored. In that case, since the output of the Fourier transform unit 8 in the signal transformation unit 7 may be given to a level determination unit 23 and a continuity determination unit 24 described later, the Fourier transform unit 22 can also be omitted.

  In addition, there is a method that brings an effect close to auditory weighting, such as non-linear amplitude conversion processing, in the spectral domain, so if you can ignore errors with the auditory weighting method used in the encoding process, The output of the Fourier transform unit 8 in the unit 7 is used as an input to the auditory weighting unit 21, and the auditory weighting unit 21 performs auditory weighting in the spectral region on this input, omitting the Fourier transform unit 22, and will be described later. It is also possible to configure so that an auditory weighted spectrum is output to the level determination unit 23 and the continuity determination unit 24.

  The Fourier transform unit 22 in the deformation intensity control unit 20 performs windowing on a signal obtained by combining the auditory weighted speech input from the auditory weighting unit 21 and the latest part of the auditory weighted speech of the previous frame as necessary. A spectrum component for each frequency is calculated by performing Fourier transform processing on the signal after windowing, and this is output to the level determination unit 23 and the continuity determination unit 24 as an auditory weighted spectrum. The Fourier transform process and the windowing process are the same as those of the Fourier transform unit 8 of the first embodiment.

  The level determination unit 23 calculates the first deformation intensity for each frequency based on the magnitude of each amplitude component value of the auditory weighting spectrum input from the Fourier transform unit 22, and uses this as the deformation intensity calculation unit 25. Output to. The smaller the value of each amplitude component of the auditory weighting spectrum, the larger the ratio of quantization noise, so the first deformation strength may be increased. In the simplest case, an average value of all amplitude components is obtained, and a predetermined threshold Th is added to the average value. The first deformation strength is 0 for components exceeding this value, and for components below this value. For example, the first deformation strength may be 1. FIG. 6 shows the relationship between the auditory weighting spectrum and the first deformation intensity when this threshold value Th is used. Note that the first deformation strength calculation method is not limited to this.

  The continuity determination unit 24 evaluates the continuity in the time direction of each amplitude component or each phase component of the auditory weighting spectrum input from the Fourier transform unit 22, and based on this evaluation result, the second continuity for each frequency. The deformation strength is calculated and output to the deformation strength calculation unit 25. It is said that good coding was performed for frequency components with low continuity in the time direction of the amplitude component of the auditory weighting spectrum and low continuity of the phase component (after compensating for phase rotation due to temporal transition between frames). Since it is difficult to think, increase the second deformation strength. For the calculation of the second deformation strength, the method of giving 0 or 1 by the determination using a predetermined threshold can be used most simply.

  Based on the first deformation strength input from the level determination unit 23 and the second deformation strength input from the continuity determination unit 24, the deformation strength calculation unit 25 calculates the final deformation strength for each frequency. This is calculated and output to the amplitude smoothing unit 9 and the phase disturbance unit 10 in the signal transformation unit 7. For the final deformation strength, the minimum value, the weighted average value, the maximum value, etc. of the first deformation strength and the second deformation strength can be used. Above, description of operation | movement of the deformation | transformation intensity | strength control part 20 newly added in this Embodiment 3 is complete | finished.

  Next, a description will be given of constituent elements whose operation is changed with the addition of the deformation strength control unit 20.

  The amplitude smoothing unit 9 performs a smoothing process on the amplitude component of the spectrum for each frequency input from the Fourier transform unit 8 according to the deformation intensity input from the deformation intensity control unit 20, and converts the smoothed spectrum to Output to the phase disturbance unit 10. It should be noted that control is performed so as to increase the smoothing of the frequency component having a higher deformation strength. The simplest method for controlling the strength of the smoothing strength is to perform smoothing only when the input deformation strength is large. As another method of enhancing smoothing, the spectrum after the smoothing coefficient α in the smoothing equation described in Embodiment 1 is reduced or fixed smoothed and the spectrum before smoothing are used. It is possible to use various methods such as generating a final spectrum by weighting and adding a weight to the spectrum before smoothing.

  The phase perturbation unit 10 perturbs the phase component of the smoothed spectrum input from the amplitude smoothing unit 9 according to the deformation intensity input from the deformation intensity control unit 20, and converts the spectrum after the disturbance into an inverse Fourier transform unit. 11 is output. It should be noted that control is performed so that a frequency component having a higher deformation strength gives a larger phase disturbance. The simplest method for controlling the magnitude of the disturbance is to give the disturbance only when the input deformation strength is large. As other methods for controlling the disturbance, various methods such as increasing or decreasing the range of the phase angle generated by random numbers can be used.

  Since other components are the same as those in the first embodiment, description thereof is omitted.

  Although the output results of both the level determination unit 23 and the continuity determination unit 24 are used here, a configuration in which only one is used and the other is omitted is also possible. In addition, a configuration in which only one of the amplitude smoothing unit 9 and the phase disturbance unit 10 is controlled by the deformation strength may be used.

  According to the third embodiment, the magnitude of each frequency component of the input signal (decoded speech) or the auditory weighted input signal (decoded speech), the magnitude of the continuity of the amplitude and phase for each frequency. In addition to the effect of the first embodiment, since the amplitude spectrum component is small, quantization is performed because the deformation intensity at the time of generating the processed signal (modified decoded speech) is controlled for each frequency. Components that are dominated by noise and degradation components, and the continuity of spectral components is low, so that quantization noise and degradation components tend to increase. The effect of improving the subjective quality by suppressing the quantization noise and the degradation component subjectively while keeping the characteristics of the input signal and the actual background noise relatively good, without processing the good component with few components. There .

Embodiment 4 FIG.
FIG. 7 in which the same reference numerals are assigned to the corresponding parts in FIG. 5 shows the overall configuration of the speech decoding apparatus to which the sound signal processing method according to the present embodiment is applied. In the figure, 41 is an addition control value dividing unit. 5 is changed to a configuration of a Fourier transform unit 8, a spectrum transform unit 39, and an inverse Fourier transform unit 11.

  Hereinafter, the operation will be described with reference to the drawings.

  The decoded speech 5 output from the speech decoding unit 4 is input to the Fourier transform unit 8, the deformation strength control unit 20, and the signal evaluation unit 12 in the signal processing unit 2.

  In the same manner as in the second embodiment, the Fourier transform unit 8 performs windowing on a signal obtained by combining the input decoded speech 5 of the current frame and the latest part of the decoded speech 5 of the previous frame as necessary. A spectrum component for each frequency is calculated by performing Fourier transform processing on the signal after windowing, and this is output as a decoded speech spectrum 43 to the weighted addition unit 18 and the amplitude smoothing unit 9 in the spectrum transformation unit 39. To do.

  In the same manner as in the second embodiment, the spectrum modification unit 39 performs the processes of the amplitude smoothing unit 9 and the phase disturbance unit 10 on the input decoded speech spectrum 43 in order, and converts the obtained spectrum into the modified decoded speech. The spectrum 44 is output to the weighted addition unit 18.

    In the deformation intensity control unit 20, the auditory weighting unit 21, the Fourier transform unit 22, the level determination unit 23, the continuity determination unit 24, and the deformation intensity calculation are performed on the input decoded speech 5 as in the third embodiment. The processing of the unit 25 is sequentially performed, and the obtained deformation strength for each frequency is output to the addition control value dividing unit 41.

  As in the third embodiment, the perceptual weighting unit 21 and the Fourier transform unit 22 are not required when perceptual weighting is not performed in the encoding process or when the effect is small. In that case, what is necessary is just to give the output of the Fourier-transform part 8 to the level determination part 23 and the continuity determination part 24. FIG.

  The output of the Fourier transform unit 8 is used as an input to the perceptual weighting unit 21. The perceptual weighting unit 21 performs perceptual weighting on the input in the spectral region, omits the Fourier transform unit 22, and is described later. It is also possible to configure the judgment unit 23 and the continuity judgment unit 24 to output auditory weighted spectra. By configuring in this way, a process simplification effect can be obtained.

  Similarly to the first embodiment, the signal evaluation unit 12 obtains the likelihood of background noise for the input decoded speech 5 and outputs this to the addition control value division unit 41 as the addition control value 35.

  The newly added addition control value dividing unit 41 uses the deformation strength for each frequency input from the deformation strength control unit 20 and the addition control value 35 input from the signal evaluation unit 12 to perform addition control for each frequency. A value 42 is generated and output to the weighted addition unit 18. For a frequency having a high deformation strength, the value of the addition control value 42 of the frequency is controlled so that the weight of the decoded speech spectrum 43 in the weighted addition unit 18 is weakened and the weight of the modified decoded speech spectrum 44 is increased. On the other hand, for a frequency with a weak deformation strength, the value of the addition control value 42 for that frequency is controlled to increase the weight of the decoded speech spectrum 43 in the weighted addition unit 18 and weaken the weight of the modified decoded speech spectrum 44. In other words, since the frequency of strong deformation is high in background noise, the addition control value 42 for the frequency is increased, and in the opposite case, it is decreased.

  The weighted addition unit 18 is based on the addition control value 42 for each frequency input from the addition control value dividing unit 41, and the decoded speech spectrum 43 input from the Fourier transform unit 8 and the modification input from the spectrum modification unit 39. The decoded speech spectrum 44 is weighted and added, and the obtained spectrum is output to the inverse Fourier transform unit 11. As the operation of the weighted addition control method, as described with reference to FIG. 2, the weight for the decoded speech spectrum 43 is applied to a frequency component having a large addition control value 42 for each frequency (high background noise likelihood). A small weight and a large weight for the modified decoded speech spectrum 44 are controlled. On the other hand, for a frequency component having a small addition control value 42 for each frequency (low background noise likelihood), the weight for the decoded speech spectrum 43 is increased, and the weight for the modified decoded speech spectrum 44 is decreased.

  As a final process, the inverse Fourier transform unit 11 returns to the signal region by performing an inverse Fourier transform process on the spectrum input from the weighted addition unit 18 in the same manner as in the second embodiment. The connection is performed while performing windowing for smooth connection with the front and rear frames, and the obtained signal is output as the output sound 6.

  The addition control value dividing unit 41 is abolished, the output of the signal evaluation unit 12 is given to the weighted addition unit 18, and the deformation strength that is the output of the deformation strength control unit 20 is sent to the amplitude smoothing unit 9 and the phase disturbance unit 10. A configuration is also possible. This is equivalent to the weighted addition process in the configuration of Embodiment 3 performed in the spectral domain.

Furthermore, as in the case of the third embodiment, it is possible to use only one of the level determination unit 23 and the continuity determination unit 24 and omit the remaining one.
According to the fourth embodiment, the magnitude of the amplitude for each frequency component of the input signal (decoded speech) or the auditory weighted input signal (decoded speech), the magnitude of the continuity of the amplitude and phase for each frequency. Since the weighted addition of the spectrum of the input signal (decoded speech spectrum) and the processed spectrum (modified decoded speech spectrum) is controlled independently for each frequency component, in addition to the effects of the first embodiment, Emphasis is placed on components where quantization noise and degradation components are dominant because the amplitude spectral components are small, and components where quantization noise and degradation components tend to increase due to low continuity of spectrum components. The weight of the processing spectrum is increased, and the weight of the processing spectrum is not increased to a good component with little quantization noise and deterioration component. While leaving sex relatively good can subjectively suppressed quantization noise or the degraded component, there is an effect of improving the subjective quality.

  Compared to the third embodiment, the transformation processing for each frequency, that is, smoothing and disturbance, is changed to the transformation processing for each frequency, which has the effect of simplifying the processing.

Embodiment 5 FIG.
FIG. 8, in which parts corresponding to those in FIG. 5 are assigned the same reference numerals, shows the overall configuration of the speech decoding apparatus to which the sound signal processing method according to this embodiment is applied. In the figure, reference numeral 26 denotes background noise likelihood (addition control value 35). It is a variability determination part which determines the variability of the time direction.

  Hereinafter, the operation will be described with reference to the drawings.

  The decoded speech 5 output from the speech decoding unit 4 is input to the signal transformation unit 7, the transformation strength control unit 20, the signal evaluation unit 12, and the weighted addition unit 18 in the signal processing unit 2. The signal evaluation unit 12 evaluates the likelihood of background noise for the input decoded speech 5 and outputs the evaluation result as the addition control value 35 to the variability determination unit 26 and the weighted addition unit 18.

The variability determination unit 26 compares the addition control value 35 input from the signal evaluation unit 12 with the past addition control value 35 stored therein, and whether or not the variability of the value in the time direction is high. And the third deformation strength is calculated based on the determination result, and is output to the deformation strength calculation unit 25 in the deformation strength control unit 20. Then, the past addition control value 35 stored therein is updated using the input addition control value 35.
When the variability in the time direction of the parameter representing the characteristics of the frame (or subframe) such as the addition control value 35 is high, the spectrum of the decoded speech 5 often changes greatly in the time direction, which is more than necessary. If strong amplitude smoothing or phase disturbance is applied, an unnatural echo is generated. Therefore, when the variability in the time direction of the addition control value 35 is high, the third deformation strength is set so that the smoothing in the amplitude smoothing unit 9 and the disturbance application in the phase disturbance unit 10 are weakened. Note that the same effect can be obtained by using parameters other than the addition control value 35, such as decoded speech power and spectrum envelope parameters, as long as they represent parameters of frames (or subframes).

  As a variability determination method, the simplest method is to compare the absolute value of the difference from the addition control value 35 of the previous frame with a predetermined threshold value and to determine that the variability is high if the threshold value is exceeded. . In addition, the absolute value of the difference between the previous frame and the previous frame's addition control value 35 may be calculated, and determination may be made based on whether one of the absolute values exceeds a predetermined threshold. In addition, when the signal evaluation unit 12 calculates the addition control value 35 for each subframe, the absolute value of the difference of the addition control value 35 between all the subframes in the current frame or in the previous frame is calculated as necessary. It can be determined by determining whether any of them exceeds a predetermined threshold. As a specific processing example, the third deformation strength is set to 0 if the threshold is exceeded, and the third deformation strength is set to 1 if it is below the threshold.

  In the deformation intensity control unit 20, the auditory weighting unit 21, the Fourier transform unit 22, the level determination unit 23, and the continuity determination unit 24 perform the same processing as in the third embodiment on the input decoded speech 5. Do.

  In the deformation strength calculation unit 25, the first deformation strength input from the level determination unit 23, the second deformation strength input from the continuity determination unit 24, and the third deformation strength input from the variability determination unit 26. Based on the deformation strength, the final deformation strength for each frequency is calculated and output to the amplitude smoothing section 9 and the phase disturbance section 10 in the signal deformation section 7. As a method for calculating the final deformation strength, the third deformation strength is given as a constant value with respect to all frequencies, and the third deformation strength, the first deformation strength, It is possible to use a method of obtaining a final deformation strength by obtaining a minimum value, a weighted average value, a maximum value, etc. of the second deformation strength.

  Subsequent operations of the signal transformation unit 7 and the weighted addition unit 18 are the same as those in the third embodiment, and a description thereof will be omitted.

  Although the output results of both the level determination unit 23 and the continuity determination unit 24 are used here, it is possible to use only one or neither. In addition, a configuration in which only one of the amplitude smoothing unit 9 and the phase disturbance unit 10 is controlled by the deformation strength, or only one of the third deformation strengths is controlled.

  According to the fifth embodiment, in addition to the configuration of the third embodiment, the smoothing strength or the disturbance imparting strength is set to the time variability (variability between frames or subframes) of a predetermined evaluation value (likeness of background noise) In addition to the effects of the third embodiment, it is possible to suppress processing that is stronger than necessary in a section where the characteristics of the input signal (decoded speech) fluctuate, This has the effect of preventing the occurrence of echoes.

Embodiment 6 FIG.
FIG. 9, in which the same reference numerals are assigned to the parts corresponding to FIG. 5, shows the overall configuration of the speech decoding apparatus to which the sound signal processing method according to this embodiment is applied. In the figure, 27 is a friction sound likelihood evaluation unit, 31 is a background noise likelihood evaluation unit, and 45 is an addition control value calculation unit. The frictional sound likelihood evaluation unit 27 includes a low-frequency cut filter 28, a zero-crossing number counting unit 29, and a frictional sound likelihood calculation unit 30. The background noise likelihood evaluation unit 31 has the same configuration as the signal evaluation unit 12 in FIG. 5, and includes an inverse filter unit 13, a power calculation unit 14, a background noise likelihood calculation unit 15, an estimated noise power update unit 16, and an estimated noise spectrum update unit. 17. Unlike the case of FIG. 5, the signal evaluation unit 12 includes a frictional sound likelihood evaluation unit 27, a background noise likelihood evaluation unit 31, and an addition control value calculation unit 45.

  Hereinafter, the operation will be described with reference to the drawings.

  The decoded speech 5 output from the speech decoding unit 4 includes a signal deformation unit 7 in the signal processing unit 2, a deformation strength control unit 20, a frictional sound likelihood evaluation unit 27 and a background noise likelihood evaluation unit 31 in the signal evaluation unit 12, and Input to the weighted addition unit 18.

  Similar to the signal evaluation unit 12 in the third embodiment, the background noise likelihood evaluation unit 31 in the signal evaluation unit 12 applies an inverse filter unit 13, a power calculation unit 14, and background noise likelihood to the input decoded speech 5. The processing of the calculation unit 15 is performed, and the obtained background noise likelihood 46 is output to the addition control value calculation unit 45. Further, the estimated noise power update unit 16 and the estimated noise spectrum update unit 17 are processed to update the estimated noise power and the estimated noise spectrum stored therein.

  The low frequency cut filter 28 in the frictional sound likelihood evaluation unit 27 performs low frequency cut filtering processing for suppressing low frequency components on the input decoded speech 5, and outputs the decoded speech after filtering to the zero-crossing number counting unit 29. Output. The purpose of this low-frequency cut filtering process is to prevent a direct current component or a low-frequency component included in the decoded speech from becoming an offset and reducing the count result of the zero-crossing number counting unit 29 described later. Therefore, simply, an average value of the decoded speech 5 in the frame may be calculated and subtracted from each sample of the decoded speech 5.

  The zero-crossing number counting unit 29 analyzes the voice input from the low-frequency cut filter 28, counts the number of zero-crossings included, and outputs the obtained number of zero-crossings to the friction sound likelihood calculation unit 30. To count the number of zero crossings, compare the positive and negative values of adjacent samples and count them as crossing zero if they are not the same. Take the product of the values of adjacent samples, and the result is negative or zero. If there is, there is a method of counting as crossing zero.

  The frictional sound likelihood calculating unit 30 compares the number of zero crossings input from the zero crossing number counting unit 29 with a predetermined threshold value, obtains the frictional sound likelihood 47 based on the comparison result, and adds this to the addition control value calculating unit 45. Output to. For example, when the number of zero crossings is larger than the threshold value, it is determined that the noise is likely to be a friction noise, and the friction noise likelihood is set to 1. Conversely, when the number of zero crossings is smaller than the threshold value, it is determined that the frictional sound is not likely and the frictional noise likelihood is set to zero. In addition, two or more threshold values may be provided to set the frictional sound likelihood in a stepwise manner, or a predetermined function may be prepared to calculate a continuous value of the frictional sound likelihood from the number of zero crossings. .

  Note that the configuration within the frictional sound likelihood evaluation unit 27 is merely an example, and the evaluation is based on the analysis result of the spectrum inclination, the evaluation is based on the continuity of power and spectrum, or the zero crossing. A plurality of parameters including the number may be combined and evaluated.

  The addition control value calculation unit 45 calculates an addition control value 35 based on the background noise likelihood 46 input from the background noise likelihood evaluation unit 31 and the friction sound likelihood 47 input from the friction sound likelihood evaluation unit 27. The result is output to the weighted addition unit 18. In both cases of background noise and friction noise, quantization noise is often difficult to hear. Therefore, the addition control value 35 is calculated by appropriately weighted addition of background noise likelihood 46 and friction noise likelihood 47. do it.

    The subsequent operations of the signal transformation unit 7, the transformation strength control unit 20, and the weighted addition unit 18 are the same as those in the third embodiment, and a description thereof will be omitted.

  According to the sixth embodiment, when the background noise and the frictional noise are high in the input signal (decoded voice), the processed signal (modified decoded voice) is output more greatly instead of the input signal (decoded voice). Therefore, in addition to the effects of the third embodiment, a focused processing is applied to a frictional sound section that tends to generate a lot of quantization noise and deterioration components, and sections other than the frictional sound are also appropriate for the section. Since processing (not processing, performing low level processing, etc.) is selected, there is an effect that subjective quality can be improved. In addition to the frictional sound, when a part that tends to generate a lot of quantization noise and degradation components can be specified to some extent, it is possible to evaluate the likelihood of the part and reflect it in the addition control value. With such a configuration, it is possible to suppress large quantization noises and degradation components one by one, so that there is an effect that the subjective quality can be further improved.

  As a matter of course, a configuration in which the background noise likelihood evaluation unit is deleted is also possible.

Embodiment 7 FIG.
FIG. 10, in which the same reference numerals are assigned to the parts corresponding to those in FIG. 1, shows the overall configuration of the speech decoding apparatus to which the signal processing method according to this embodiment is applied. In FIG.

  Hereinafter, the operation will be described with reference to the drawings.

  First, the speech code 3 is input to the speech decoding unit 4 in the speech decoding device 1.

  The speech decoding unit 4 performs a decoding process on the input speech code 3 and outputs the obtained decoded speech 5 to the post filter unit 32, the signal transformation unit 7, and the signal evaluation unit 12.

  The post filter unit 32 performs spectrum enhancement processing, pitch periodicity enhancement processing, and the like on the input decoded speech 5 and outputs the obtained result to the weighted addition unit 18 as the post filter decoded speech 48. This post-filter process is generally used as a post-process of CELP decoding process, and is introduced for the purpose of suppressing quantization noise generated by encoding / decoding. Since the portion where the spectrum intensity is weak contains a lot of quantization noise, the amplitude of this component is suppressed. In some cases, the pitch periodicity enhancement process is not performed and only the spectrum enhancement process is performed.

  In the first embodiment and the third to sixth embodiments, the post filter processing has been described as being applicable to any of those included in the speech decoding unit 4 or those that do not exist. 7, all or part of the post-filter processing is made independent as post-filter unit 32 from what the post-filter processing is included in the speech decoding unit 4.

  As in the first embodiment, the signal transformation unit 7 performs the processes of the Fourier transform unit 8, the amplitude smoothing unit 9, the phase disturbance unit 10, and the inverse Fourier transform unit 11 on the input decoded speech 5, The obtained modified decoded speech 34 is output to the weighted addition unit 18.

  As in the first embodiment, the signal evaluation unit 12 evaluates the likelihood of background noise for the input decoded speech 5 and outputs the evaluation result as the addition control value 35 to the weighted addition unit 18.

  Then, as a final process, the weighted addition unit 18 performs post-filter decoded speech input from the post-filter unit 32 based on the addition control value 35 input from the signal evaluation unit 12 as in the first embodiment. 48 and the modified decoded speech 34 input from the signal transformation unit 7 are weighted and added, and the resulting output speech 6 is output.

  According to the seventh embodiment, the modified decoded speech is generated based on the decoded speech before processing by the post filter, and the decoded speech before processing by the post filter is analyzed to obtain the likelihood of background noise, based on this Since the weight at the time of adding the post-filter decoded speech and the modified decoded speech is controlled, in addition to the effect of the first embodiment, the modified decoded speech that does not include the deformation of the decoded speech by the post-filter can be generated. Since the addition weight control with high accuracy can be performed based on the high accuracy of background noise calculated without being affected by the deformation of the decoded speech by the filter, the subjective quality is further improved.

  In the background noise section, the degraded sound is often emphasized by the post-filter, and it is often difficult to hear, and it is better to generate the modified decoded speech using the decoded speech before processing by the post-filter as the starting point. Becomes smaller. In addition, when the post-filter processing has multiple modes, and switching the processing often, there is a high risk that the switching will affect the evaluation of the likelihood of background noise. A more stable evaluation result can be obtained by evaluating the likelihood of background noise.

  In the configuration of the third embodiment, when the post filter unit is separated as in the seventh embodiment, the output result of the auditory weighting unit 21 in FIG. 5 is more perceptually weighted in the encoding process. As the voice approaches, the accuracy of specifying a component with a lot of quantization noise is improved, better deformation intensity control can be obtained, and the subjective quality can be further improved.

  Further, in the configuration of the sixth embodiment, when the post filter unit is separated as in the seventh embodiment, the evaluation accuracy in the frictional sound likelihood evaluation unit 27 of FIG. 9 is increased, and the subjective quality is further improved. An effect is obtained.

  Note that the configuration in which the post filter unit is not separated is less connected to the speech decoding unit (including the post filter) than the separated configuration of the seventh embodiment, and is independent of the decoded speech. There is an advantage that it can be easily realized by a device and a program. In the seventh embodiment, there is a disadvantage that it is not easy to be realized by an apparatus and program independent of a speech decoding unit having a post filter. However, the seventh embodiment has various effects described above.

Embodiment 8 FIG.
FIG. 11, in which the same reference numerals are assigned to the parts corresponding to those in FIG. 10, shows the overall configuration of the speech decoding apparatus to which the sound signal processing method according to this embodiment is applied. In FIG. It is a spectral parameter. The difference from FIG. 10 is that a deformation strength control unit 20 similar to that of the third embodiment is added, and the spectrum parameter 33 is input from the speech decoding unit 4 to the signal evaluation unit 12 and the deformation strength control unit 20. It is.

  Hereinafter, the operation will be described with reference to the drawings.

  First, the speech code 3 is input to the speech decoding unit 4 in the speech decoding device 1.

  The speech decoding unit 4 performs a decoding process on the input speech code 3 and outputs the obtained decoded speech 5 to the post filter unit 32, the signal transformation unit 7, the transformation strength control unit 20, and the signal evaluation unit 12. . Further, the spectrum parameter 33 generated in the course of the decoding process is output to the estimated noise spectrum updating unit 17 in the signal evaluation unit 12 and the auditory weighting unit 21 in the deformation intensity control unit 20. As the spectrum parameter 33, a linear prediction coefficient (LPC), a line spectrum pair (LSP), etc. are generally used in many cases.

  The perceptual weighting unit 21 in the deformation intensity control unit 20 performs perceptual weighting processing on the decoded speech 5 input from the speech decoding unit 4 using the spectral parameter 33 also input from the speech decoding unit 4. The perceived weighted speech is output to the Fourier transform unit 22. Specifically, when the spectral parameter 33 is a linear prediction coefficient (LPC), this is used as it is. When the spectral parameter 33 is a parameter other than LPC, the spectral parameter 33 is converted to LPC. Then, the LPC is subjected to constant multiplication to obtain two modified LPCs, an ARMA filter having the two modified LPCs as filter coefficients is configured, and auditory weighting is performed by a filtering process using the filter. The auditory weighting process is desirably performed in the same manner as that used in the speech encoding process (which is paired with the speech decoding process performed by the speech decoding unit 4).

  In the deformation intensity control unit 20, following the process of the auditory weighting unit 21, the processes of the Fourier transform unit 22, the level determination unit 23, the continuity determination unit 24, and the deformation intensity calculation unit 25, as in the third embodiment. And the obtained deformation strength is output to the signal deformation unit 7.

  Similarly to the third embodiment, the signal transformation unit 7 processes the input decoded speech 5 and the transformation intensity by the Fourier transform unit 8, the amplitude smoothing unit 9, the phase disturbance unit 10, and the inverse Fourier transform unit 11. The modified decoded speech 34 obtained is output to the weighted addition unit 18.

  In the signal evaluation unit 12, as in the first embodiment, the input decoded speech 5 is first subjected to the processing of the inverse filter unit 13, the power calculation unit 14, and the background noise likelihood calculation unit 15 to be the background noise likelihood. And the evaluation result is output to the weighted addition unit 18 as the addition control value 35. Moreover, the process of the estimated noise power update part 16 is performed, and internal estimated noise power is updated.

Then, the estimated noise spectrum updating unit 17 updates the estimated noise spectrum stored therein using the spectrum parameter 33 input from the speech decoding unit 4 and the background noise input from the background noise likelihood calculating unit 15. . For example, when the likelihood of input background noise is high, updating is performed by reflecting the spectrum parameter 33 in the estimated noise spectrum according to the equation shown in the first embodiment.
Since the subsequent operations of the post filter unit 32 and the weighted addition unit 18 are the same as those in the seventh embodiment, description thereof will be omitted.

  According to the eighth embodiment, the spectral parameters generated in the course of the speech decoding process are used to perform the auditory weighting process and the update of the estimated noise spectrum, so the third and seventh embodiments. In addition to the effects of the process, the process is simplified.

  Furthermore, exactly the same auditory weighting process as that of the encoding process is realized, the accuracy of specifying components with a lot of quantization noise is improved, better deformation intensity control is obtained, and the effect of improving the subjective quality is obtained.

  In addition, the estimation accuracy of the estimated noise spectrum used to calculate the background noise likelihood (in the sense that it is close to the speech spectrum input to the speech encoding process) is increased, and the resulting high-precision background noise likelihood is obtained. Based on this, it becomes possible to perform addition weight control with high accuracy, and there is an effect of improving the subjective quality.

  In the eighth embodiment, the post-filter unit 32 is separated from the speech decoding unit 4. However, the spectrum output by the speech decoding unit 4 as in the eighth embodiment also in a configuration in which the post-filter unit 32 is not separated. The signal processing unit 2 can be processed using the parameter 33. Even in this case, the same effect as in the eighth embodiment can be obtained.

Embodiment 9 FIG.
In the configuration of the fourth embodiment shown in FIG. 7, the outline of the spectrum after the addition control value dividing unit 41 has multiplied the weight for each frequency of the modified decoded speech spectrum 44 added by the weighted addition unit 18 is as follows. It is also possible to control the output deformation intensity so as to match the estimated spectral shape of the quantization noise.

  FIG. 12 is a schematic diagram illustrating an example of a spectrum obtained by multiplying the decoded speech spectrum 43 and the modified decoded speech spectrum 44 in this case by a weight for each frequency.

  The decoded speech spectrum 43 is superposed with quantization noise having a spectrum shape depending on the encoding method. In the CELP speech coding method, a code search is performed so as to minimize distortion in speech after auditory weighting processing. For this reason, the quantization noise has a flat spectral shape in the sound after perceptual weighting processing, and the final spectral shape of the quantization noise has a spectral shape opposite to that of perceptual weighting processing. become. Therefore, it is possible to obtain the spectrum characteristic of the auditory weighting process, obtain the spectrum shape of the inverse characteristic, and control the output of the addition control value dividing unit 41 so that the spectrum shape of the modified decoded speech spectrum matches this. is there.

  According to the ninth embodiment, the spectral shape of the modified decoded speech component included in the final output speech 6 is made to coincide with the approximate shape of the estimated spectrum of the quantization noise. In addition, it is possible to make it difficult to hear unpleasant quantization noise in the speech section by adding the modified decoded speech with the minimum power.

Embodiment 10 FIG.
In the configuration of the first embodiment and the third to eighth embodiments, the amplitude spectrum after smoothing may be processed so as to match the amplitude spectrum shape of the estimated quantization noise within the processing of the amplitude smoothing unit 9. Is possible. The calculation of the amplitude spectrum shape of the estimated quantization noise may be performed in the same manner as in the ninth embodiment.

  According to the tenth embodiment, since the spectrum shape of the modified decoded speech matches the estimated spectrum shape of the quantization noise, in addition to the effects of the first embodiment and the third to eighth embodiments, it is necessary. There is an effect that it is possible to make it difficult to hear unpleasant quantization noise in the speech section by adding the deformed decoded speech with the minimum power.

Embodiment 11 FIG.
In the first embodiment and the third to tenth embodiments, the signal processing unit 2 is used for processing the decoded speech 5. However, only the signal processing unit 2 is taken out and an acoustic signal decoding unit (acoustic signal encoding) is used. The decoder can be used for other signal processing processing, such as connecting to a subsequent stage of noise suppression processing. However, it is necessary to change or adjust the deformation process in the signal deformation unit and the evaluation method in the signal evaluation unit according to the characteristics of the degradation component to be eliminated.

  According to the eleventh embodiment, it is possible to process a component including a deteriorated component other than the decoded speech so that a subjectively undesirable component is difficult to feel.

Embodiment 12 FIG.
In the first to eleventh embodiments, the signal up to the current frame is used to process the signal. However, a configuration in which processing delay is allowed and signals from the next frame are used is also possible.

  According to the twelfth embodiment, since the signal after the next frame can be referred to, it is possible to improve the smoothing characteristic of the amplitude spectrum, improve the accuracy of continuity determination, and improve the evaluation accuracy such as noise.

Embodiment 13 FIG.
In the first embodiment, the third embodiment, and the fifth to twelfth embodiments, the spectral component is calculated by Fourier transform, transformed, and returned to the signal domain by inverse Fourier transform. In addition, it is possible to perform a transformation process on each output of the bandpass filter group and reconstruct the signal by adding the band-specific signals.

  According to the thirteenth embodiment, the same effect can be obtained even in a configuration that does not use Fourier transform.

Embodiment 14 FIG.
In the first to thirteenth embodiments, the amplitude smoothing unit 9 and the phase disturbance unit 10 are both provided. However, a configuration in which one of the amplitude smoothing unit 9 and the phase disturbance unit 10 is omitted is also possible. However, a configuration in which another deformed portion is introduced is also possible.

  According to the fourteenth embodiment, depending on the characteristics of quantization noise and deteriorated sound to be eliminated, there is an effect that the processing can be simplified by omitting a deforming portion having no introduction effect. In addition, by introducing an appropriate deformation unit, it is possible to expect an effect of eliminating quantization noise and degraded sound that cannot be eliminated by the amplitude smoothing unit 9 and the phase disturbance unit 10.

It is a figure which shows the whole structure of the audio | voice decoding apparatus to which the audio | voice decoding method by Embodiment 1 of this invention is applied. It is a figure which shows the control example of the weighting addition based on the addition control value in the weighting addition part 18 of Embodiment 1 of this invention. It is explanatory drawing explaining the time relationship with the cut-out window in the Fourier-transform part 8 of Embodiment 1 of this invention, the actual shape example of the window for the connection in the inverse Fourier-transform part 11, and the decoding audio | voice 5. FIG. It is a figure which shows a part of structure of the audio | voice decoding apparatus which applied the sound signal processing method of Embodiment 2 of this invention in combination with the noise suppression method. It is a figure which shows the whole structure of the audio | voice decoding apparatus to which the audio | voice decoding method by Embodiment 3 of this invention is applied. It is a figure which shows the relationship between the auditory weighting spectrum of Embodiment 3 of this invention, and 1st deformation intensity. It is a figure which shows the whole structure of the audio | voice decoding apparatus to which the audio | voice decoding method by Embodiment 4 of this invention is applied. It is a figure which shows the whole structure of the speech decoding apparatus to which the speech decoding method by Embodiment 5 of this invention is applied. It is a figure which shows the whole structure of the speech decoding apparatus to which the speech decoding method by Embodiment 6 of this invention is applied. It is a figure which shows the whole structure of the audio | voice decoding apparatus to which the audio | voice decoding method by Embodiment 7 of this invention is applied. It is a figure which shows the whole structure of the audio | voice decoding apparatus to which the audio | voice decoding method by Embodiment 8 of this invention is applied. It is a schematic diagram which shows an example of the spectrum after multiplying the weight for every frequency to the decoding audio | voice spectrum 43 to which Embodiment 9 of this invention is applied, and the deformation | transformation decoding audio | voice spectrum 44. FIG.

Claims (14)

A decoded speech generation step of generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters ;
A first processed sound generation step of generating a first processed sound by smoothing the amplitude of the decoded sound generated in the decoded sound generation step in the frequency axis direction ;
A mixing ratio used for generating a second processed voice generated by mixing the decoded voice and the first processed voice, and as the noise likelihood of the decoded voice increases, the second processing A mixing ratio for mixing the decoded speech and the first processed speech so that a ratio of the first processed speech in the speech is increased is any one of a plurality of parameters generated in the decoded speech generation step. A mixing ratio acquisition step to acquire based on one;
And a second processed sound generation step of generating a second processed sound by mixing the decoded sound and the first processed sound at the mixing ratio acquired in the mixing ratio acquisition step. Sound signal processing method.   A decoded speech generation step of generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  A first processed sound generation step of generating a first processed sound by smoothing the amplitude of the decoded sound generated in the decoded sound generation step in the frequency axis direction;
  A weighting coefficient acquisition step of acquiring a weighting coefficient that increases as the noise likelihood of the decoded voice increases based on any one of the plurality of parameters generated in the decoded voice generation step;
  A second processed voice generating step of adding the decoded voice and the first processed voice weighted with the weighting coefficient acquired in the weighting coefficient acquiring step to generate a second processed voice; A sound signal processing method characterized by the above.   A decoded speech generation step of generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  A first processed speech generation step of generating a first processed speech by smoothing the amplitude of the decoded speech generated in the decoded speech generation step in a time axis direction;
  A mixing ratio used for generating a second processed voice generated by mixing the decoded voice and the first processed voice, and as the noise likelihood of the decoded voice increases, the second processing A mixing ratio for mixing the decoded speech and the first processed speech so that a ratio of the first processed speech in the speech is increased is any one of a plurality of parameters generated in the decoded speech generation step. A mixing ratio acquisition step to acquire based on one;
  And a second processed sound generation step of generating a second processed sound by mixing the decoded sound and the first processed sound at the mixing ratio acquired in the mixing ratio acquisition step. Sound signal processing method.   A decoded speech generation step of generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  A first processed speech generation step of generating a first processed speech by smoothing the amplitude of the decoded speech generated in the decoded speech generation step in a time axis direction;
  A weighting coefficient acquisition step of acquiring a weighting coefficient that increases as the noise likelihood of the decoded voice increases based on any one of the plurality of parameters generated in the decoded voice generation step;
  A second processed voice generating step of adding the decoded voice and the first processed voice weighted with the weighting coefficient acquired in the weighting coefficient acquiring step to generate a second processed voice; A sound signal processing method characterized by the above.   A decoded speech generation step of generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  The first processed speech in which fluctuation in the time axis direction of the amplitude of the decoded speech generated in the decoded speech generation step is reduced to the amplitude at a predetermined time of the decoded speech by a predetermined time before the predetermined time. A first processed speech generation step for performing weighted addition processing for weighted addition of the amplitude of the decoded speech at a time point and the amplitude of the decoded speech at a time point after the predetermined time from the predetermined time point;
  A mixing ratio used for generating a second processed voice generated by mixing the decoded voice and the first processed voice, and as the noise likelihood of the decoded voice increases, the second processing A mixing ratio for mixing the decoded speech and the first processed speech so that a ratio of the first processed speech in the speech is increased is any one of a plurality of parameters generated in the decoded speech generation step. A mixing ratio acquisition step to acquire based on one;
  And a second processed sound generation step of generating a second processed sound by mixing the decoded sound and the first processed sound at the mixing ratio acquired in the mixing ratio acquisition step. Sound signal processing method.   A decoded speech generation step of generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  The first processed speech in which fluctuation in the time axis direction of the amplitude of the decoded speech generated in the decoded speech generation step is reduced to the amplitude at a predetermined time of the decoded speech by a predetermined time before the predetermined time. A first processed speech generation step for performing weighted addition processing for weighted addition of the amplitude of the decoded speech at a time point and the amplitude of the decoded speech at a time point after the predetermined time from the predetermined time point;
  A weighting coefficient acquisition step of acquiring a weighting coefficient that increases as the noise likelihood of the decoded voice increases based on any one of the plurality of parameters generated in the decoded voice generation step;
  A second processed voice generating step of adding the decoded voice and the first processed voice weighted with the weighting coefficient acquired in the weighting coefficient acquiring step to generate a second processed voice; A sound signal processing method characterized by the above.   7. The sound signal according to claim 5, wherein, in the first processed sound generation step, the weighted addition process is repeatedly performed for a plurality of time points at equal time intervals to generate a first processed sound. 8. Processing method.   Decoded speech generating means for generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  First processed sound generating means for smoothing the amplitude of the decoded sound generated by the decoded sound generating means in the frequency axis direction to generate a first processed sound;
  A mixing ratio used for generating a second processed voice generated by mixing the decoded voice and the first processed voice, and as the noise likelihood of the decoded voice increases, the second processing Any one of a plurality of parameters generated by the decoded sound generation means is a mixing ratio for mixing the decoded sound and the first processed sound so that a ratio of the first processed sound in the sound is increased. A mixing ratio acquisition means to acquire based on one;
  And a second processed sound generating means for generating a second processed sound by mixing the decoded sound and the first processed sound at the mixing ratio acquired by the mixing ratio acquiring means. Sound signal processing device.   Decoded speech generating means for generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  First processed sound generating means for smoothing the amplitude of the decoded sound generated by the decoded sound generating means in the frequency axis direction to generate a first processed sound;
  Based on any one of a plurality of parameters generated by the decoded speech generation means, weighting coefficient acquisition means for acquiring a weighting coefficient that increases as the noise likelihood of the decoded speech increases;
  A second processed sound generating means for adding the decoded sound and the first processed sound weighted by the weighting coefficient acquired by the weighting coefficient acquiring means to generate a second processed sound; A sound signal processing apparatus characterized by.   Decoded speech generating means for generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  First processed sound generating means for smoothing the amplitude of the decoded sound generated by the decoded sound generating means in the time axis direction to generate a first processed sound;
  A mixing ratio used for generating a second processed voice generated by mixing the decoded voice and the first processed voice, and as the noise likelihood of the decoded voice increases, the second processing Any one of a plurality of parameters generated by the decoded sound generation means is a mixing ratio for mixing the decoded sound and the first processed sound so that a ratio of the first processed sound in the sound is increased. A mixing ratio acquisition means to acquire based on one;
  And a second processed sound generating means for generating a second processed sound by mixing the decoded sound and the first processed sound at the mixing ratio acquired by the mixing ratio acquiring means. Sound signal processing device.   Decoded speech generating means for generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  First processed sound generating means for smoothing the amplitude of the decoded sound generated by the decoded sound generating means in the time axis direction to generate a first processed sound;
  Based on any one of a plurality of parameters generated by the decoded speech generation means, weighting coefficient acquisition means for acquiring a weighting coefficient that increases as the noise likelihood of the decoded speech increases;
  A second processed sound generating means for adding the decoded sound and the first processed sound weighted by the weighting coefficient acquired by the weighting coefficient acquiring means to generate a second processed sound; A sound signal processing apparatus characterized by.   Decoded speech generating means for generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  The first processed speech in which the fluctuation in the time axis direction of the amplitude of the decoded speech generated by the decoded speech generation means is reduced to the amplitude at a predetermined time of the decoded speech by a predetermined time before the predetermined time. First processed speech generation means for performing weighted addition processing for weighted addition of the amplitude of the decoded speech at a time point and the amplitude of the decoded speech at a time point after the predetermined time from the predetermined time point;
  A mixing ratio used for generating a second processed voice generated by mixing the decoded voice and the first processed voice, and as the noise likelihood of the decoded voice increases, the second processing Any one of a plurality of parameters generated by the decoded sound generation means is a mixing ratio for mixing the decoded sound and the first processed sound so that a ratio of the first processed sound in the sound is increased. A mixing ratio acquisition means to acquire based on one;
  And a second processed sound generating means for generating a second processed sound by mixing the decoded sound and the first processed sound at the mixing ratio acquired by the mixing ratio acquiring means. Sound signal processing device.   Decoded speech generating means for generating a plurality of parameters from the speech code and generating a decoded speech corresponding to the speech code using the plurality of parameters;
  The first processed speech in which the fluctuation in the time axis direction of the amplitude of the decoded speech generated by the decoded speech generation means is reduced to the amplitude at a predetermined time of the decoded speech by a predetermined time before the predetermined time. First processed speech generation means for performing weighted addition processing for weighted addition of the amplitude of the decoded speech at a time point and the amplitude of the decoded speech at a time point after the predetermined time from the predetermined time point;
  Based on any one of a plurality of parameters generated by the decoded speech generation means, weighting coefficient acquisition means for acquiring a weighting coefficient that increases as the noise likelihood of the decoded speech increases;
  A second processed sound generating means for adding the decoded sound and the first processed sound weighted by the weighting coefficient acquired by the weighting coefficient acquiring means to generate a second processed sound; A sound signal processing apparatus characterized by.   The sound signal according to claim 12 or 13, wherein the first processed sound generation means generates the first processed sound by repeatedly performing the weighted addition processing at a plurality of time points at equal time intervals. Processing equipment.

Images