WO2014129233A1 - 音声強調装置 - Google Patents
音声強調装置 Download PDFInfo
- Publication number
- WO2014129233A1 WO2014129233A1 PCT/JP2014/050573 JP2014050573W WO2014129233A1 WO 2014129233 A1 WO2014129233 A1 WO 2014129233A1 JP 2014050573 W JP2014050573 W JP 2014050573W WO 2014129233 A1 WO2014129233 A1 WO 2014129233A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- spectrum
- frequency
- power spectrum
- unit
- input signal
- Prior art date
Links
- 238000001228 spectrum Methods 0.000 claims abstract description 221
- 230000006835 compression Effects 0.000 claims abstract description 39
- 238000007906 compression Methods 0.000 claims abstract description 39
- 238000004458 analytical method Methods 0.000 claims abstract description 34
- 238000006243 chemical reaction Methods 0.000 claims abstract description 22
- 238000012937 correction Methods 0.000 claims abstract description 17
- 230000015572 biosynthetic process Effects 0.000 claims description 26
- 238000003786 synthesis reaction Methods 0.000 claims description 26
- 230000007423 decrease Effects 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 description 30
- 238000000034 method Methods 0.000 description 27
- 230000008569 process Effects 0.000 description 12
- 230000005236 sound signal Effects 0.000 description 10
- 230000002159 abnormal effect Effects 0.000 description 9
- 238000005311 autocorrelation function Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 8
- 238000000605 extraction Methods 0.000 description 7
- 230000003595 spectral effect Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 4
- 238000007796 conventional method Methods 0.000 description 4
- 238000009499 grossing Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000006866 deterioration Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002194 synthesizing effect Effects 0.000 description 2
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0316—Speech enhancement, e.g. noise reduction or echo cancellation by changing the amplitude
- G10L21/0364—Speech enhancement, e.g. noise reduction or echo cancellation by changing the amplitude for improving intelligibility
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0316—Speech enhancement, e.g. noise reduction or echo cancellation by changing the amplitude
- G10L21/0324—Details of processing therefor
- G10L21/0332—Details of processing therefor involving modification of waveforms
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/21—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being power information
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03G—CONTROL OF AMPLIFICATION
- H03G9/00—Combinations of two or more types of control, e.g. gain control and tone control
- H03G9/02—Combinations of two or more types of control, e.g. gain control and tone control in untuned amplifiers
- H03G9/025—Combinations of two or more types of control, e.g. gain control and tone control in untuned amplifiers frequency-dependent volume compression or expansion, e.g. multiple-band systems
Definitions
- the present invention relates to a voice communication system such as a car navigation system, a mobile phone, an interphone, a sound pickup device, a hands-free call system, a video conference system, a monitoring system, and a voice communication system, a voice storage system, a voice synthesis system, and a voice recognition system.
- the present invention relates to a speech enhancement device that is used for improving sound quality of a broadcast system / speech synthesis system or the like, and improving a recognition rate of a speech recognition system, and improving the quality and intelligibility of a speech signal.
- the frequency band of an audio signal sent through a telephone line is narrowly limited, for example, with an upper limit frequency of 3400 Hz. For this reason, the sound quality of conventional telephone lines is not very good. Also, in digital voice communication such as mobile phones and business radios, the bandwidth is limited in the same way as analog lines due to severe bit rate restrictions, so it cannot be said that the sound quality is good in this case. Consonant components of speech, “individuality (personality)” and “naturalness” of speech often exist in the band of 3400 Hz or more, but they are largely lost due to the above-mentioned bandwidth limitation.
- both the transmitting side terminal and the receiving side terminal need to support the corresponding wideband speech encoding / decoding method, and both base stations can complete a network for wideband encoding. Because it is necessary, it has only been put into practical use in some business communication systems, and it is not only an economic burden to implement on a public telephone communication network, but it also takes a lot of time to spread. .
- Patent Document 1 has an advantage that the individuality of the voice is maintained, but the compressed high frequency signal component is mapped to a predetermined fixed band in the low frequency range. Therefore, the above-mentioned bandwidth may not be optimal depending on the sound state of the input signal. In this case, there is a problem that deterioration of sound quality is unavoidable.
- the present invention has been made to solve the above-described problems, and an object thereof is to obtain a speech enhancement device capable of generating high-quality speech.
- a speech enhancement apparatus includes a time-frequency conversion unit that converts a time domain input signal into a power spectrum that is a frequency domain signal, an input signal analysis unit that analyzes a mode of the input signal from the power spectrum, and an input A band determining unit that determines a boundary frequency within a range that does not exceed a predetermined first frequency in terms of a signal, and a spectrum compressing unit that compresses a power spectrum of a frequency in a band above the first frequency in the frequency direction. And a spectrum synthesizer that reflects the compressed power spectrum in a band determined by the first frequency and the boundary frequency, and the combined power spectrum output from the spectrum synthesizer and the phase spectrum of the input signal are converted to the time domain. And a frequency-time conversion unit for obtaining an enhanced signal.
- the boundary frequency for determining the band reflecting the high frequency component is determined based on the mode of the input signal, it is possible to prevent abnormal noise due to spectrum synthesis, which is good and clear. Voice enhancement processing becomes possible.
- FIG. 1 is a block diagram showing a speech enhancement apparatus according to Embodiment 1 of the present invention.
- the speech enhancement apparatus according to the first embodiment includes an input terminal 1, a time-frequency conversion unit 2, an input signal analysis unit 3, a band determination unit 4, a high-frequency component extraction unit 5, a spectrum compression unit 6, a gain correction unit 7, It comprises a spectrum synthesizer 8, a frequency-time converter 9, and an output terminal 10.
- FIG. 2 shows an internal configuration of the input signal analysis unit 3.
- the input signal analysis unit 3 includes an autocorrelation analysis unit 11, a noise spectrum estimation unit 12, an SN ratio calculation unit 13, a power ratio analysis unit 14,
- the determination unit 15 is configured.
- the input terminal 1 is a terminal to which a signal such as an audio signal is input as an input signal.
- the time-frequency conversion unit 2 is a processing unit that converts a time domain input signal into a power spectrum that is a frequency domain signal.
- the input signal analysis unit 3 determines whether the input signal is speech or noise from the power spectrum output from the time-frequency conversion unit 2, and whether the input signal is a vowel or consonant in the case of speech. It is a process part which analyzes an aspect.
- the band determination unit 4 is a processing unit that determines the boundary frequency within a range that does not exceed a predetermined first frequency from the aspect of the input signal analyzed by the input signal analysis unit 3.
- the high-frequency component extraction unit 5 is a processing unit that extracts a power spectrum of a frequency in a band above the first frequency.
- the spectrum compression unit 6 is a processing unit that compresses the power spectrum of the band extracted by the high frequency component extraction unit 5 in the frequency direction.
- the gain correction unit 7 is a processing unit that performs gain correction compressed by the spectrum compression unit 6.
- the spectrum synthesizing unit 8 is a processing unit that reflects the compressed power spectrum gain-corrected by the gain correcting unit 7 in a band determined by the first frequency and the boundary frequency.
- the frequency-time conversion unit 9 is a processing unit for converting the combined power spectrum output from the spectrum combining unit 8 and the phase spectrum of the input signal into the time domain to obtain enhanced speech (enhanced signal).
- the output terminal 10 is a terminal for outputting the emphasized sound output from the frequency-time conversion unit 9 to the outside.
- the autocorrelation analysis unit 11 is a processing unit that obtains a normalized autocorrelation function from the power spectrum of the input signal.
- the noise spectrum estimation unit 12 is a processing unit that obtains an estimated noise spectrum from the power spectrum of the input signal.
- the S / N ratio calculation unit 13 is a processing unit that calculates the average S / N ratio of the original frame from the power spectrum output from the time-frequency conversion unit 2 and the estimated noise spectrum output from the noise spectrum estimation unit 12.
- the power ratio analysis unit 14 is a processing unit that obtains a high frequency / low frequency power ratio from the power spectrum of the input signal.
- the determination unit 15 includes a power spectrum output from the time-frequency conversion unit 2, a maximum value of the normalized autocorrelation function obtained by the autocorrelation analysis unit 11, and a high frequency / low frequency obtained by the power ratio analysis unit 14. This is a processing unit that receives the region power ratio and the estimated noise spectrum output from the noise spectrum estimation unit 12 and determines whether the input signal of the current frame is speech (vowel / consonant) or noise.
- voice or music captured through a microphone or the like is A / D (analog / digital) converted and then sampled at a predetermined sampling frequency (for example, 16 kHz) and divided into frames (for example, 10 ms). , And input to the speech enhancement device through the input terminal 1.
- the band of the finally obtained processing signal is 0 Hz to 3400 Hz (first frequency), and 0 Hz to 3400 Hz is referred to as a pass band and 3400 Hz to 8000 Hz is referred to as a high frequency band.
- processing is performed on a signal including a pass band of 0 to 8000 Hz and a high frequency band.
- a high frequency band signal is also referred to as a high frequency component.
- the time-frequency conversion unit 2 performs, for example, Hanning windowing on the input signal x (t) divided into frame units, and then, for example, a fast Fourier transform (Fast Fourier) of 512 points as shown in the following equation (1): Transform (FFT) is performed to convert a time-domain signal into a spectral component X ( ⁇ , k) that is a frequency-domain signal.
- FFT fast Fourier transform
- t is a sampling time
- ⁇ is a frame number when the input signal is divided into frames
- k is a number designating a frequency component of a spectrum frequency band (hereinafter referred to as a spectrum number)
- FT [ ⁇ ] is a fast Fourier Represents the conversion process.
- Re ⁇ X ( ⁇ , k) ⁇ and Im ⁇ X ( ⁇ , k) ⁇ denote a real part and an imaginary part of the input signal spectrum after Fourier transform, respectively.
- the input signal analysis unit 3 determines whether the input signal of the current frame is speech or noise. In the case of voice, it is also determined whether the voice is a vowel or a consonant. First, the autocorrelation analysis unit 11 obtains a normalized autocorrelation function ⁇ N ( ⁇ , ⁇ ) from the power spectrum Y ( ⁇ , k) using the following equation (3).
- ⁇ is a delay time
- the expression (3) is a Wiener-Khintchin theorem and will not be described.
- the maximum value ⁇ max ( ⁇ ) of the normalized autocorrelation function is obtained using Expression (4).
- the expression (4) means that the maximum value of ⁇ ( ⁇ , ⁇ ) is searched in the range of 32 ⁇ ⁇ ⁇ 192.
- the SN ratio calculation unit 13 calculates the power spectrum Y ( ⁇ , k) output from the time-frequency conversion unit 2 and the estimated noise spectrum and N ( ⁇ , k) output from the noise spectrum estimation unit 12 described later.
- the average signal-to-noise ratio SNR AVE ( ⁇ ) of the current frame is calculated.
- N ( ⁇ , k) is an estimated noise spectrum
- S pow and N pow represent the sum of the power spectrum of the input signal and the sum of the estimated noise spectrum, respectively.
- subband SN ratio SNR ( ⁇ , k) is calculated as the SN ratio for each spectral component by the equation (6), and is output to the spectrum synthesis unit 8 described later.
- the power ratio analysis unit 14 obtains a ratio between the high frequency power and low frequency power of the input signal of the current frame (high frequency / low frequency power ratio). This ratio approximates the degree of spectral tilt from low to high, and there is a high possibility of consonants when the high frequency power is greater than the low frequency power. Thus, it can be determined whether the input signal is a vowel or a consonant. Specifically, for example, a power ratio between a band of 125 Hz to 3000 Hz and a band of 3000 Hz to 6000 Hz is obtained as in Expression (7).
- P s_Low ( ⁇ ) is a band power of 125 Hz to 3000 Hz of the input signal
- P s_High ( ⁇ ) is a band power of 3000 Hz to 6000 Hz, and is calculated as follows.
- the obtained high frequency / low frequency power ratio P Ratio ( ⁇ ) of the current frame is output to the determination unit 15.
- the determination unit 15 includes the power spectrum Y ( ⁇ , k) output from the time-frequency conversion unit 2, the maximum value ⁇ max ( ⁇ ) of the normalized autocorrelation function obtained by the autocorrelation analysis unit 11, and the power ratio.
- the high frequency / low frequency power ratio P Ratio ( ⁇ ) obtained by the analysis unit 14 and the estimated noise spectrum N ( ⁇ , k) output from the noise spectrum estimation unit 12 described later are input, and the input signal of the current frame is It is determined whether the sound is a voice (vowel / consonant) or noise, and the result is output as a determination flag.
- the determination flag Vflag is set to “1 (voice)” as being voice, and otherwise, noise is determined. As a result, the determination flag Vflag is set to “0 (noise)” and output.
- the high frequency / low frequency power ratio P Ratio ( ⁇ ) output by the power ratio analysis unit 14 has a predetermined threshold value using the following equation (10). If it exceeds, it is determined as a consonant, and the determination flag Vflag is set to “2 (consonant)” and output.
- the input speech analysis method uses the autocorrelation function method, the average SN ratio of the input signal, and the high frequency / low frequency power ratio.
- the present invention is not limited to this method, and a known method such as cepstrum analysis is used. Other methods such as the above method may be used. Further, it is possible to improve the determination accuracy by combining various known methods such as the number of zero crossings at the discretion of those skilled in the art.
- the noise spectrum estimation unit 12 receives the power spectrum Y ( ⁇ , k) output from the time-frequency conversion unit 2 and the determination flag Vflag output from the determination unit 15, and the following equation (11) and the determination flag Vflag are input.
- the noise spectrum is estimated and updated according to, and the estimated noise spectrum N ( ⁇ , k) is output.
- N ( ⁇ -1, k) is an estimated noise spectrum in the previous frame, and is held in a storage means such as a RAM (Random Access Memory) in the noise spectrum estimation unit 12.
- a storage means such as a RAM (Random Access Memory) in the noise spectrum estimation unit 12.
- the band determination unit 4 receives the determination flag Vflag output from the input signal analysis unit 3, and generates a boundary frequency B ( ⁇ ) representing a bandwidth for synthesizing (or convolving) a high-frequency component described later with respect to the pass band. decide.
- the boundary frequency B ( ⁇ ) can be determined using, for example, Expression (12) and Expression (13).
- Expression (12) when the current frame is determined to be a noise section, the bandwidth is minimized, that is, the boundary frequency is set to a value close to 3400 Hz (3300 Hz) which is the upper limit frequency of the pass band.
- the voice formants are often concentrated at a frequency of 3000 Hz or less, so that the high frequency component is reflected while maintaining the formants of 3000 Hz or less.
- a high frequency component is convolved with a bandwidth of 3000 Hz to 3400 Hz with a boundary frequency of 3000 Hz.
- consonants there are many consonant components even in the high region above 3400 Hz. Therefore, in order to reflect more of these components in the passband, the band is widened, that is, the boundary frequency is lower than the boundary frequency of the vowel.
- the value (2500 Hz) is set, and the high frequency component is convoluted in the band of 2500 Hz to 3400 Hz.
- the boundary frequency is moved up and down to adjust to the optimum frequency.
- B F 3000 Hz
- the power spectrum of the temporary boundary frequency is compared with the power spectrum of 3400 Hz in the upper and lower 100 Hz range (2900 Hz to 3100 Hz).
- the boundary frequency B ( ⁇ 1) of the previous frame is used as shown in Equation (13) in order to suppress the occurrence of abnormal noise due to a sudden change in the bandwidth between frames.
- the boundary frequency is smoothed in the time direction.
- the high frequency component extraction unit 5 extracts high frequency components based on the determination flag output from the input signal analysis unit 3.
- the frequency range of the high frequency component is 3400 Hz to 8000 Hz
- the power spectrum Y ( ⁇ , k) in this range is cut out, and the power spectrum Y h ( ⁇ , k) of the high frequency component is extracted. Is output.
- the determination flag is a voice section
- clipping is performed, but when the determination flag is a noise section (or silent section), it is possible not to perform clipping. In this case, the amount of processing can be reduced because spectrum compression processing and spectrum synthesis processing described later are not performed.
- the extracted high frequency component is sent to the subsequent spectrum compression unit 6.
- the spectrum compression unit 6 performs band compression in the frequency direction of the spectrum so that the power spectrum Y h ( ⁇ , k) of the high frequency component extracted according to the high frequency component extraction unit 5 can be combined with the spectrum on the pass band. Match the bandwidth with the passband.
- the subband signal-to-noise ratio SNR ( ⁇ , k) of the spectrum component of the passband is frequency-compressed using the subband signal-to-noise ratio output from the input signal analysis unit 3 as shown in equation (15).
- the power spectrum of the emphasized speech is obtained by comparing the subband SN ratio SNR m ( ⁇ , k) of the spectral component in the high band and selecting the one with the larger subband SN ratio.
- smoothing between frequencies is performed in a band near the boundary frequency.
- a known method such as a moving average filter is used to smooth the boundary frequency so that it is smooth at 100 Hz above and below the boundary frequency (width of 200 Hz).
- FIG. 3 schematically shows a series of operation principles of the first embodiment for easier understanding.
- A represents the power spectrum of the input audio signal
- (b) represents the power spectrum of the high-frequency component, which is the output of the high-frequency component extraction unit 5
- (c) represents the output of the spectrum compression unit 6.
- a frequency-compressed high-frequency component power spectrum is represented
- (d) represents a frequency-compressed and gain-corrected high-frequency component power spectrum that is an output of the gain correction unit 7.
- the arrow represents the order of processing.
- FIG. 4 is an example of the input signal spectrum in the voice section.
- (A) is a speech signal in a consonant section
- (b) is a signal when the power of a high frequency component of 3400 Hz or higher is small and the consonant characteristics are poor, although it is a consonant section.
- FIG. 5 is an example when the input signal shown in FIG. 4 is processed by the conventional method
- FIG. 6 is an example when the input signal shown in FIG. 4 is also processed by the first embodiment. is there.
- the bandwidth is fixed as indicated by the arrow 500a in (a) and the arrow 500b in (b).
- a peak component is generated in the spectrum and abnormal noise is generated as indicated by the broken line portion 501).
- the method is indicated by the arrow 600a in (a) and the arrow 600b in (b).
- the audio signal in (a) reflects more consonant features in the passband (as shown in the range 601a, the high-frequency component has a wider range)
- the time-frequency conversion unit that converts the time domain input signal into the power spectrum that is the frequency domain signal, and the input that analyzes the mode of the input signal from the power spectrum.
- a signal analysis unit a band determination unit that determines a boundary frequency within a range that does not exceed a predetermined first frequency based on an aspect of an input signal, and a power spectrum of a frequency in a band above the first frequency in a frequency direction
- a spectrum compression unit that compresses the compressed power spectrum into a band determined by the first frequency and the boundary frequency, a combined power spectrum output from the spectrum synthesis unit, and a phase of the input signal
- a frequency-time conversion unit that converts the spectrum into the time domain and obtains an enhanced signal is provided. Since the input signal can be enhanced by reflecting the high frequency band signal, the high frequency characteristics can be appropriately reflected in the passband, and there is no abnormal noise due to spectrum synthesis, enabling good and clear voice enhancement processing. Become.
- the power compressed by the spectrum compression unit so that the power of the power spectrum before compression and the power spectrum after compression in the band compressed by the spectrum compression unit become equal.
- a gain correction unit that performs power correction of a power spectrum that is compressed by correcting the spectrum or multiplying by a predetermined correction coefficient determined based on auditory factors, and the spectrum synthesis unit is a gain correction unit Since the power spectrum corrected in step S5 is reflected, the generation of abnormal noise during spectrum synthesis is suppressed, and satisfactory speech enhancement processing is possible.
- the band determining unit when reflecting the compressed power spectrum, sets the boundary frequency to the frequency with the smallest power difference from the power spectrum belonging to the first frequency. Therefore, the generation of abnormal noise during spectrum synthesis is suppressed, and good speech enhancement processing is possible.
- the band determination unit smoothes the boundary frequency of the current frame in the time direction using the boundary frequency of the previous frame. Occurrence of abnormal noise is suppressed, and good speech enhancement processing is possible.
- the spectrum synthesis unit compares the SN ratio of the power spectrum of the input signal with the SN ratio of the compressed power spectrum, and the power spectrum having the higher SN ratio. Is selected to generate a combined power spectrum, so when the subband SN ratio of the high-frequency band signal is low at the time of spectrum synthesis with the high-frequency component, the spectral component is not reflected in the passband, Since it is possible to prevent the deteriorated high-frequency component from being reflected in the pass band, it is possible to perform a good speech enhancement process while suppressing an increase in the feeling of speech degradation.
- Embodiment 2 the determination according to the state of the input signal is performed by using the determination flag that is the input signal analysis result for the determination of the boundary frequency in the band determination unit 4, but the present invention is not limited to this.
- the boundary frequency is not expressed by three discrete values as in the equation (11), but may be a continuous value between 2500 Hz and 3400 Hz according to the values of both S / N ratios, for example. Good.
- the boundary frequency B ( ⁇ ) is lowered to widen the bandwidth reflecting the high frequency component.
- the high frequency / low frequency power ratio P Ratio ( ⁇ ) increases, the possibility of consonants increases, so the boundary frequency B ( ⁇ ) is decreased.
- the boundary frequency B ( ⁇ ) is increased to narrow the bandwidth reflecting the high frequency component.
- the boundary frequency can be continuously controlled according to the S / N ratio of the input signal, it is possible to set the optimum bandwidth according to the S / N ratio of the input signal, and the average S / N ratio of the input signal.
- the frequency is low, the band is narrowed to suppress the reflection of extra high frequency components in the pass band, so that it is possible to prevent an increase in the feeling of deterioration of the voice and to perform a better voice enhancement process.
- the band determining unit sets the boundary frequency low when the SN ratio of the input signal is high, and sets the boundary frequency higher as the SN ratio decreases. As a result, it is possible to prevent an increase in the sense of deterioration of the voice and to perform a better voice enhancement process.
- Embodiment 3 FIG.
- the spectrum synthesizer 8 compares the subband SN ratio of the high-frequency component and the subband SN ratio of the passband, and performs spectrum synthesis according to the subband SN ratio. Without selecting a spectral component using the subband signal-to-noise ratio, the power spectrum Y ( ⁇ , k) of the input signal for each spectral component and the power of the high frequency component subjected to frequency compression and gain correction as shown in Equation (16) It is also possible to obtain a power spectrum of the enhancement signal by taking a weighted average of the spectra Y m ( ⁇ , k) and combining them.
- W S (k) is a predetermined constant for weighting.
- the weight of the power spectrum of the high frequency component is increased as the frequency is increased as in Expression (17).
- the spectrum synthesis unit generates the synthesized power spectrum by weighted averaging the power spectrum of the input signal and the compressed power spectrum.
- the power discontinuity of the spectrum in the frequency direction can be alleviated.
- by increasing the weight of the power spectrum of the high frequency component as the frequency increases more high frequency components can be reflected in the high frequency, enabling more natural and clear voice enhancement processing. Become.
- Embodiment 4 FIG.
- the peak of the spectrum (the “mountain” portion of the peak-and-valley structure of the spectrum) as represented by the formant frequency approaches.
- the emphasized speech may have a feeling of echo.
- this feeling of reverberation for example, by comparing the power spectrum corresponding to the peak and selecting the larger component, or by not compressing the spectrum at a frequency near the power spectrum corresponding to the peak. Place the peaks apart.
- the speech enhancement apparatus of the fourth embodiment when the spectrum compression unit compresses a plurality of power spectrum peaks within a predetermined frequency, the power corresponding to the plurality of peaks is generated. A large component of the spectrum is selected, or spectrum compression is not performed at frequencies close to the power spectrum corresponding to multiple peaks, so the spectrum peaks that may cause reverberation during spectrum compression are placed apart. As a result, it is possible to suppress the sense of unusual noise in the emphasized speech, and a good speech enhancement process is possible.
- Embodiment 5 As a modification of the fourth embodiment, a process for a power spectrum peak generated at the time of spectrum synthesis in the spectrum synthesis unit 8 will be described as a fifth embodiment.
- the spectrum synthesis unit 8 synthesizes the power spectrum of the input signal and the power spectrum of the high frequency component subjected to the spectrum compression, when the peak of the spectrum, as represented by the formant frequency, approaches, the emphasis voice is reverberated. May occur.
- the components corresponding to the peaks of the power spectrum of the input signal and the power spectrum of the high frequency component are compared, and the larger component is selected or equivalent to that peak.
- the spectrum is not compressed and the peaks are separated and synthesized. Thereby, the peak of the spectrum which may become a factor of an echo at the time of spectrum compression can be arranged apart.
- the speech enhancement device of the fifth embodiment when the spectrum synthesizer synthesizes a plurality of power spectrum peaks within a predetermined frequency, the power spectrum of the input signal and the high frequency band Compare the component corresponding to each peak of the power spectrum of the component and select the larger component, or the spectrum compression unit does not compress the spectrum at frequencies near the power spectrum corresponding to multiple peaks Since it did in this way, since the spectrum peak which may become a factor of a reverberation at the time of spectrum compression can be arrange
- the passband is described as 3400 Hz.
- the present invention is not limited to this, and for example, broadband transmission at 7000 Hz is possible.
- an input signal in the 11 kHz band is input and analyzed.
- the present invention realizes a sense of broadness exceeding the passband and clarity, and does not require re-development processing of high-frequency components on the receiving side, that is, no additional processing on the receiving terminal side is required. There is an effect that the quality is improved without increasing the amount and regardless of the type of the receiving terminal.
- the present invention is not only intended for voice, but can also be applied to non-voice signals such as music.
- the input signal analysis unit 3 is replaced with one adapted to the musical sound, but it may be determined using a known analysis means corresponding to the musical sound corresponding to the consonant / vowel of the voice.
- the present invention is not only effective as a sound quality improvement measure for bandwidth limitation during wireless communication transmission, but also effective when the high frequency reproduction capability of a speaker is poor, or when high frequency components are attenuated in loud sound broadcasting or the like. Since the characteristics of the high frequency component can be reflected in the band that can be reproduced by the speaker and the low frequency band that is difficult to attenuate, there is an effect that clear sound can be reproduced.
- the memory capacity of storage devices and D / A (digital / analog) conversion for consumer electronics and toys that are demanding cost such as voice reading of TV program guides and synthetic voice output of toys, etc. Even when the sampling frequency is limited by this limitation, it is possible to reproduce a clear audio signal that exceeds the sampling frequency audibly.
- the enhanced output signal is sent in a digital data format to various audio / acoustic processing apparatuses such as a voice encoding device, a voice recognition device, a voice storage device, and a hands-free call device.
- the voice emphasis device of each embodiment can be realized by a DSP (digital signal processor) alone or together with the other devices described above, or executed as a software program.
- the program may be stored in a storage device of a computer device that executes the software program, or may be distributed in a storage medium such as a CD-ROM. It is also possible to provide a program through a network.
- D / A conversion it can be amplified by an amplifying apparatus and directly output as an audio signal from a speaker or the like.
- the speech enhancement apparatus includes a band determination unit that determines a band that reflects a high-frequency component within a range that does not exceed a predetermined first frequency based on an aspect of an input signal. Since sound can be prevented and good and clear voice enhancement processing is possible, voice communication, voice storage, voice synthesis, and voice recognition systems have been introduced, such as car navigation systems, mobile phones, intercoms, and sound collection devices. Suitable for use in.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
- Telephone Function (AREA)
Abstract
Description
実施の形態1.
図1は、この発明の実施の形態1による音声強調装置を示す構成図である。
本実施の形態1の音声強調装置は、入力端子1、時間-周波数変換部2、入力信号分析部3、帯域決定部4、高域成分切出し部5、スペクトル圧縮部6、ゲイン補正部7、スペクトル合成部8、周波数-時間変換部9、出力端子10で構成されている。また、図2は入力信号分析部3の内部構成を示すものであり、入力信号分析部3は、自己相関分析部11、雑音スペクトル推定部12、SN比計算部13、パワー比分析部14、判定部15で構成されている。
自己相関分析部11は、入力信号のパワースペクトルから正規化自己相関関数を求める処理部である。雑音スペクトル推定部12は、入力信号のパワースペクトルから推定雑音スペクトルを求める処理部である。SN比計算部13は、時間-周波数変換部2が出力するパワースペクトルと雑音スペクトル推定部12が出力する推定雑音スペクトルから、原フレームの平均SN比を計算する処理部である。パワー比分析部14は、入力信号のパワースペクトルから高域・低域パワー比を求める処理部である。判定部15は、時間-周波数変換部2が出力するパワースペクトルと、自己相関分析部11で得られた正規化自己相関関数の最大値と、パワー比分析部14で得られた高域・低域パワー比と、雑音スペクトル推定部12が出力する推定雑音スペクトルを入力し、現フレームの入力信号が音声(母音・子音)であるか雑音であるかどうかの判定を行う処理部である。
まず、マイクロホンなどを通じて取り込まれた音声や音楽などが、A/D(アナログ・デジタル)変換された後、所定のサンプリング周波数(例えば、16kHz)でサンプリングされる共にフレーム単位に分割(例えば10ms)され、入力端子1を通じて音声強調装置へ入力される。なお、本実施の形態では、最終的に得る処理信号の帯域は0Hz~3400Hz(第1の周波数)とし、0Hz~3400Hzを通過帯域、3400Hz~8000Hzを高周波帯域と称して説明する。また、特に指定がなければ、0~8000Hzの通過帯域と高周波帯域を含めた信号に対して処理を行うものとする。また、高周波帯域の信号を高域成分とも称して説明する。
ここで、Ps_Low(λ)は入力信号の125Hz~3000Hzの帯域パワー、Ps_High(λ)は3000Hz~6000Hzの帯域パワーであり、以下のように計算される。
以上、得られた現フレームの高域・低域パワー比PRatio(λ)を判定部15へ出力する。
ここで、THFR_SN、THACFおよびTHPOW_Ratioは、判定用の所定の定数閾値であり、好適な例としてTHPR_SN=3.0、THACF=0.3およびTHPOW_Ratio=1.25であるが、入力信号の状態や雑音レベルに応じて適宜変更することもできる。
一方、判定フラグVflag=1あるいはVflag=2の場合には、現フレームの入力信号が音声であり、前フレームの推定雑音スペクトルN(λ-1,k)を、そのまま現フレームの推定雑音スペクトルN(λ,k)として出力する。
ここでWBは、時間方向平滑化のための所定の定数であり、好適な例としてWB=0.8であるが、入力信号の種類等に応じて良好に動作するように、適宜変更することができる。
なお、判定フラグが音声区間の場合は切出しを行うが、判定フラグが雑音区間(あるいは無音区間)の場合には、切出しを行わないことも可能である。この場合、後述のスペクトル圧縮処理や、スペクトル合成処理を行わないので処理量の削減ができる。切出された高域成分は後段のスペクトル圧縮部6へ送出される。
境界周波数B(λ)と通過帯域の上限周波数BNL=3400Hzが現す帯域幅、すなわち、通過帯域へ高域成分を反映する帯域幅をBW_C(λ)とし、高周波帯域の上限周波数BNH=8000Hzと、通過帯域の上限周波数BNL=3400Hzの間の帯域、すなわち、高域成分の帯域幅をBW_Hとすると、例えばスペクトル帯域の線形圧縮は次の式(14)のように表すことができる。
実施の形態1では、帯域決定部4における境界周波数の決定に、入力信号分析結果である判定フラグを用いることで入力信号の様態に応じた制御を行っているが、これに限定されることは無く、例えば、SN比計算部13が出力する平均SN比と、パワー比計算部14が出力する高域・低域パワー比を用いて制御することも可能である。その際、式(11)のように境界周波数を3つの離散値で表現するのではなく、例えば、両SN比の値に応じて2500Hz~3400Hzの間の連続的な値を取るようにしてもよい。
実施の形態1では、スペクトル合成部8において、高域成分のサブバンドSN比と、通過帯域のサブバンドSN比とを比較して、サブバンドSN比に応じてスペクトル合成を行っていたが、サブバンドSN比を用いてスペクトル成分を選択せずに、式(16)のようにスペクトル成分毎に入力信号のパワースペクトルY(λ,k)と、周波数圧縮ならびにゲイン補正した高域成分のパワースペクトルYm(λ,k)の重み付き平均を取って合成し、強調信号のパワースペクトルを求めることも可能である。
ここでBW_c(λ)は、スペクトル圧縮部6で述べたのと同様に、境界周波数B(λ)と通過帯域の上限周波数BNL=3400Hzが表す帯域幅であり、また、WS(k)を決定している定数値は、入力信号の様態に応じて予め好適な値に調整が可能である。
上述の実施の形態1において、スペクトル圧縮部6でスペクトルを周波数方向に圧縮した際に、フォルマント周波数に代表されるような、スペクトルのピーク(スペクトルの山谷構造の“山”の部分)が接近した場合、強調音声に反響感を生じることがある。この反響感を抑制するために、例えば、ピークに相当するパワースペクトルを比較して大きい成分の方を選択したり、あるいは、そのピークに相当するパワースペクトルの近傍周波数ではスペクトル圧縮をしないといったことにより、ピークを離して配置する。
実施の形態4の変形として、スペクトル合成部8におけるスペクトル合成の際に生成されるパワースペクトルのピークに対する処理を実施の形態5として以下説明する。
Claims (9)
- 時間領域の入力信号を周波数領域の信号であるパワースペクトルに変換する時間-周波数変換部と、
前記パワースペクトルから前記入力信号の態様を分析する入力信号分析部と、
前記入力信号の態様から、予め定めた第1の周波数を越えない範囲で、境界周波数を決定する帯域決定部と、
前記第1の周波数より上の帯域の周波数のパワースペクトルを周波数方向に圧縮するスペクトル圧縮部と、
前記圧縮されたパワースペクトルを、前記第1の周波数と前記境界周波数で決定される帯域に反映するスペクトル合成部と、
前記スペクトル合成部から出力された合成パワースペクトルと前記入力信号の位相スペクトルを時間領域へ変換して強調信号を得る周波数-時間変換部とを備えたことを特徴とする音声強調装置。 - 前記スペクトル圧縮部が圧縮する帯域における圧縮前のパワースペクトルのパワーと圧縮後のパワースペクトルのパワーが同等になるよう当該スペクトル圧縮部が圧縮するパワースペクトルを補正するか、または、聴感上の要因に基づいて決定される所定の補正係数を乗じるかにより前記圧縮されるパワースペクトルのパワー補正を行うゲイン補正部を備え、
前記スペクトル合成部は、前記ゲイン補正部で補正されたパワースペクトルを反映させることを特徴とする請求項1記載の音声強調装置。 - 前記帯域決定部は、前記圧縮されたパワースペクトルを反映した場合に、前記第1の周波数に属するパワースペクトルとのパワーの差が最も小さい周波数に前記境界周波数を決定することを特徴とする請求項1記載の音声強調装置。
- 前記帯域決定部は、前フレームの境界周波数を用いて、現フレームの境界周波数の時間方向の平滑化を行うことを特徴とする請求項1記載の音声強調装置。
- 前記スペクトル合成部は、前記入力信号のパワースペクトルのSN比と、前記圧縮されたパワースペクトルのSN比とを比較し、SN比が高い方のパワースペクトルを選択して合成パワースペクトルを生成することを特徴とする請求項1記載の音声強調装置。
- 前記帯域決定部は、前記入力信号のSN比が高い場合、前記境界周波数を低く設定し、前記SN比が低くなるに従って前記境界周波数を高く設定することを特徴とする請求項1記載の音声強調装置。
- 前記スペクトル合成部は、入力信号のパワースペクトルと、圧縮されたパワースペクトルとを重み付け平均して合成パワースペクトルを生成することを特徴とする請求項1記載の音声強調装置。
- 前記スペクトル圧縮部は、圧縮する際に、パワースペクトルのピークが所定周波数以内に複数生成される場合、当該複数のピークに相当するパワースペクトルのうち大きい成分を選択するか、または、前記複数のピークに相当するパワースペクトルの近傍周波数ではスペクトル圧縮を行わないことを特徴とする請求項1記載の音声強調装置。
- 前記スペクトル合成部が合成する際に、パワースペクトルのピークが所定周波数以内に複数生成される場合、入力信号のパワースペクトルと高域成分のパワースペクトルのそれぞれのピークに相当する成分を比較して、大きい成分の方を選択するか、または、前記スペクトル圧縮部が、前記複数のピークに相当するパワースペクトルの近傍周波数ではスペクトル圧縮を行わないことを特徴とする請求項1記載の音声強調装置。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/769,639 US9530430B2 (en) | 2013-02-22 | 2014-01-15 | Voice emphasis device |
CN201480008333.7A CN104981870B (zh) | 2013-02-22 | 2014-01-15 | 声音增强装置 |
JP2015501357A JP6073456B2 (ja) | 2013-02-22 | 2014-01-15 | 音声強調装置 |
DE112014000945.8T DE112014000945B4 (de) | 2013-02-22 | 2014-01-15 | Sprachbetonungsgerät |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2013033347 | 2013-02-22 | ||
JP2013-033347 | 2013-02-22 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014129233A1 true WO2014129233A1 (ja) | 2014-08-28 |
Family
ID=51391029
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2014/050573 WO2014129233A1 (ja) | 2013-02-22 | 2014-01-15 | 音声強調装置 |
Country Status (5)
Country | Link |
---|---|
US (1) | US9530430B2 (ja) |
JP (1) | JP6073456B2 (ja) |
CN (1) | CN104981870B (ja) |
DE (1) | DE112014000945B4 (ja) |
WO (1) | WO2014129233A1 (ja) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6140685B2 (ja) * | 2012-03-23 | 2017-05-31 | パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカPanasonic Intellectual Property Corporation of America | 帯域パワー算出装置及び帯域パワー算出方法 |
JP6216553B2 (ja) * | 2013-06-27 | 2017-10-18 | クラリオン株式会社 | 伝搬遅延補正装置及び伝搬遅延補正方法 |
DE102014101307A1 (de) * | 2014-02-03 | 2015-08-06 | Osram Opto Semiconductors Gmbh | Kodierverfahren zur Datenkompression von Leistungsspektren eines optoelektronischen Bauteils und Dekodierverfahren |
US10121488B1 (en) * | 2015-02-23 | 2018-11-06 | Sprint Communications Company L.P. | Optimizing call quality using vocal frequency fingerprints to filter voice calls |
JP2018159759A (ja) * | 2017-03-22 | 2018-10-11 | 株式会社東芝 | 音声処理装置、音声処理方法およびプログラム |
CN111337213A (zh) * | 2020-02-21 | 2020-06-26 | 中铁大桥(南京)桥隧诊治有限公司 | 一种基于合成功率谱桥梁模态频率识别方法及系统 |
CN113936694B (zh) * | 2021-12-17 | 2022-03-18 | 珠海普林芯驰科技有限公司 | 人声实时检测方法、计算机装置及计算机可读存储介质 |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS6254300A (ja) * | 1985-05-28 | 1987-03-09 | 日本電気株式会社 | フオルマント抽出器 |
JPH08321792A (ja) * | 1995-05-26 | 1996-12-03 | Tohoku Electric Power Co Inc | 音声信号帯域圧縮伝送方法 |
JP2002244686A (ja) * | 2001-02-13 | 2002-08-30 | Hitachi Ltd | 音声加工方法、これを用いた電話機及び中継局 |
JP2008537174A (ja) * | 2005-04-20 | 2008-09-11 | キューエヌエックス ソフトウェア システムズ (ウェイブメイカーズ), インコーポレイテッド | 音声の品質および了解度を改善するためのシステム |
JP2011141551A (ja) * | 2005-12-09 | 2011-07-21 | Qnx Software Systems (Wavemakers) Inc | 高周波数圧縮を通じてスピーチ了解度を改良するためのシステム |
JP2012083790A (ja) * | 2006-11-09 | 2012-04-26 | Sony Corp | 信号処理装置及び信号処理方法、プログラム及び記録媒体、並びに再生装置 |
JP2012145659A (ja) * | 2011-01-07 | 2012-08-02 | Jvc Kenwood Corp | 音声信号圧縮装置、音声信号圧縮方法及びプログラム |
Family Cites Families (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA1250368A (en) | 1985-05-28 | 1989-02-21 | Tetsu Taguchi | Formant extractor |
US5774837A (en) * | 1995-09-13 | 1998-06-30 | Voxware, Inc. | Speech coding system and method using voicing probability determination |
DE19941311C1 (de) * | 1999-08-31 | 2001-06-07 | Cryoelectra Ges Fuer Kryoelek | Bandfilter |
AU2002230151B2 (en) * | 2001-08-06 | 2006-08-03 | Index Corporation | Apparatus for determining dog's emotions by vocal analysis of barking sounds and method for the same |
WO2003038812A1 (en) * | 2001-11-02 | 2003-05-08 | Matsushita Electric Industrial Co., Ltd. | Audio encoding and decoding device |
SE0400997D0 (sv) * | 2004-04-16 | 2004-04-16 | Cooding Technologies Sweden Ab | Efficient coding of multi-channel audio |
AU2005201813B2 (en) | 2005-04-29 | 2011-03-24 | Phonak Ag | Sound processing with frequency transposition |
JPWO2007007476A1 (ja) * | 2005-07-13 | 2009-01-29 | 株式会社村田製作所 | 弾性境界波フィルタ装置 |
US8295507B2 (en) | 2006-11-09 | 2012-10-23 | Sony Corporation | Frequency band extending apparatus, frequency band extending method, player apparatus, playing method, program and recording medium |
JP4984983B2 (ja) * | 2007-03-09 | 2012-07-25 | 富士通株式会社 | 符号化装置および符号化方法 |
EP2077550B8 (en) * | 2008-01-04 | 2012-03-14 | Dolby International AB | Audio encoder and decoder |
JP5453740B2 (ja) * | 2008-07-02 | 2014-03-26 | 富士通株式会社 | 音声強調装置 |
JP5293817B2 (ja) * | 2009-06-19 | 2013-09-18 | 富士通株式会社 | 音声信号処理装置及び音声信号処理方法 |
PL3570278T3 (pl) * | 2010-03-09 | 2023-03-20 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Rekonstrukcja wysokiej częstotliwości wejściowego sygnału audio przy użyciu kaskadowych banków filtrów |
DK2375782T3 (en) | 2010-04-09 | 2019-03-18 | Oticon As | Improvements in sound perception by using frequency transposing by moving the envelope |
CN102880767A (zh) * | 2012-10-19 | 2013-01-16 | 西南交通大学 | 轨道交通桥梁结构噪声仿真预测方法 |
US9420368B2 (en) * | 2013-09-24 | 2016-08-16 | Analog Devices, Inc. | Time-frequency directional processing of audio signals |
CN103632676B (zh) * | 2013-11-12 | 2016-08-24 | 广州海格通信集团股份有限公司 | 一种低信噪比语音降噪方法 |
JP6260504B2 (ja) * | 2014-02-27 | 2018-01-17 | 株式会社Jvcケンウッド | オーディオ信号処理装置、オーディオ信号処理方法及びオーディオ信号処理プログラム |
-
2014
- 2014-01-15 DE DE112014000945.8T patent/DE112014000945B4/de active Active
- 2014-01-15 CN CN201480008333.7A patent/CN104981870B/zh active Active
- 2014-01-15 US US14/769,639 patent/US9530430B2/en active Active
- 2014-01-15 JP JP2015501357A patent/JP6073456B2/ja active Active
- 2014-01-15 WO PCT/JP2014/050573 patent/WO2014129233A1/ja active Application Filing
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS6254300A (ja) * | 1985-05-28 | 1987-03-09 | 日本電気株式会社 | フオルマント抽出器 |
JPH08321792A (ja) * | 1995-05-26 | 1996-12-03 | Tohoku Electric Power Co Inc | 音声信号帯域圧縮伝送方法 |
JP2002244686A (ja) * | 2001-02-13 | 2002-08-30 | Hitachi Ltd | 音声加工方法、これを用いた電話機及び中継局 |
JP2008537174A (ja) * | 2005-04-20 | 2008-09-11 | キューエヌエックス ソフトウェア システムズ (ウェイブメイカーズ), インコーポレイテッド | 音声の品質および了解度を改善するためのシステム |
JP2011141551A (ja) * | 2005-12-09 | 2011-07-21 | Qnx Software Systems (Wavemakers) Inc | 高周波数圧縮を通じてスピーチ了解度を改良するためのシステム |
JP2012083790A (ja) * | 2006-11-09 | 2012-04-26 | Sony Corp | 信号処理装置及び信号処理方法、プログラム及び記録媒体、並びに再生装置 |
JP2012145659A (ja) * | 2011-01-07 | 2012-08-02 | Jvc Kenwood Corp | 音声信号圧縮装置、音声信号圧縮方法及びプログラム |
Also Published As
Publication number | Publication date |
---|---|
DE112014000945B4 (de) | 2021-10-28 |
DE112014000945T5 (de) | 2015-11-12 |
JPWO2014129233A1 (ja) | 2017-02-02 |
JP6073456B2 (ja) | 2017-02-01 |
US9530430B2 (en) | 2016-12-27 |
CN104981870A (zh) | 2015-10-14 |
US20160005420A1 (en) | 2016-01-07 |
CN104981870B (zh) | 2018-03-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6073456B2 (ja) | 音声強調装置 | |
JP4707739B2 (ja) | 音声の品質および了解度を改善するためのシステム | |
RU2585987C2 (ru) | Устройство и способ обработки речевого/аудио сигнала | |
US8521530B1 (en) | System and method for enhancing a monaural audio signal | |
JP5127754B2 (ja) | 信号処理装置 | |
JP6147744B2 (ja) | 適応音声了解度処理システムおよび方法 | |
US8271292B2 (en) | Signal bandwidth expanding apparatus | |
CN1971711B (zh) | 语音信号自适应增强系统 | |
JP4836720B2 (ja) | ノイズサプレス装置 | |
JP5301471B2 (ja) | 音声符号化システム及び方法 | |
US20040138876A1 (en) | Method and apparatus for artificial bandwidth expansion in speech processing | |
CN110265046B (zh) | 一种编码参数调控方法、装置、设备及存储介质 | |
US20110286605A1 (en) | Noise suppressor | |
EP2483888A2 (en) | Suppressing noise in an audio signal | |
JP2005165021A (ja) | 雑音低減装置、および低減方法 | |
JPWO2006046293A1 (ja) | 雑音抑圧装置 | |
US7606702B2 (en) | Speech decoder, speech decoding method, program and storage media to improve voice clarity by emphasizing voice tract characteristics using estimated formants | |
JPH0946233A (ja) | 音声符号化方法とその装置、音声復号方法とその装置 | |
JP5589631B2 (ja) | 音声処理装置、音声処理方法および電話装置 | |
JP2004341339A (ja) | 雑音抑圧装置 | |
US10147434B2 (en) | Signal processing device and signal processing method | |
JP2008309955A (ja) | ノイズサプレス装置 | |
JP2012181561A (ja) | 信号処理装置 | |
JP4922427B2 (ja) | 信号補正装置 | |
CN112908350B (zh) | 一种音频处理方法、通信装置、芯片及其模组设备 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 14753462 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2015501357 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14769639 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 112014000945 Country of ref document: DE Ref document number: 1120140009458 Country of ref document: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 14753462 Country of ref document: EP Kind code of ref document: A1 |