WO2010111876A1 - 一种信号去噪的方法和装置及音频解码系统 - Google Patents

一种信号去噪的方法和装置及音频解码系统 Download PDF

Info

Publication number
WO2010111876A1
WO2010111876A1 PCT/CN2009/076155 CN2009076155W WO2010111876A1 WO 2010111876 A1 WO2010111876 A1 WO 2010111876A1 CN 2009076155 W CN2009076155 W CN 2009076155W WO 2010111876 A1 WO2010111876 A1 WO 2010111876A1
Authority
WO
WIPO (PCT)
Prior art keywords
adjusted
frame
coefficient
spectral
weighting
Prior art date
Application number
PCT/CN2009/076155
Other languages
English (en)
French (fr)
Inventor
陈龙吟
苗磊
胡晨
刘泽新
哈维·米希尔·塔迪
张清
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to JP2012502425A priority Critical patent/JP5459688B2/ja
Priority to KR1020117024686A priority patent/KR101390433B1/ko
Priority to EP09842532A priority patent/EP2407965B1/en
Priority to KR1020137015052A priority patent/KR101320963B1/ko
Publication of WO2010111876A1 publication Critical patent/WO2010111876A1/zh
Priority to US13/248,725 priority patent/US8965758B2/en

Links

Classifications

    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/02Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/04Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
    • G10L19/16Vocoder architecture
    • G10L19/18Vocoders using multiple modes
    • G10L19/24Variable rate codecs, e.g. for generating different qualities using a scalable representation such as hierarchical encoding or layered encoding
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00Speech 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
    • G10L21/02Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0316Speech enhancement, e.g. noise reduction or echo cancellation by changing the amplitude
    • G10L21/0364Speech enhancement, e.g. noise reduction or echo cancellation by changing the amplitude for improving intelligibility
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00Speech 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
    • G10L21/02Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/038Speech enhancement, e.g. noise reduction or echo cancellation using band spreading techniques

Definitions

  • the present invention relates to the field of audio codec technology, and in particular to a method and apparatus for signal denoising and an audio decoding system. Background technique
  • BWE Band Width Extension
  • the BWE parameter coding is characterized by less bit usage and bandwidth. Guaranteed, the quality is acceptable;
  • the code rate is high, the spectrum of the broadband or ultra-wideband part is quantized and encoded. The characteristics of the quantization code are more bits, higher precision and better quality.
  • FIG. 1 is a structural diagram of a prior art audio coding system supporting broadband or ultra-wideband.
  • the coding system adopts a hierarchical structure: a core encoder encodes low frequency information, and outputs a first layer code stream.
  • the BWE encoder encodes the high-band spectrum with less bits and outputs the second layer stream; the quantization encoder quantizes the high-band spectrum using the remaining bits and outputs the third-layer stream.
  • FIG. 2 is a structural diagram of a prior art audio decoding system supporting broadband or ultra-wideband.
  • the decoding system also adopts a hierarchical structure: a core decoder is used to decode a low-frequency first layer code stream. Information; BWE decoder is used to decode the bandwidth extended second layer stream The dequantization decoder is used to decode and dequantize the information of the third layer code stream of the remaining bits of the high frequency band; finally, the decoding system synthesizes the frequency bands of the three layers of code streams, and outputs the synthesized audio signal of the frequency band, due to the general core decoder.
  • the output signal is a time domain signal
  • the signal output by the BWE decoder and the dequantization decoder is a frequency domain signal, so in the frequency band synthesis, the frequency domain signals of the second and third layer code streams are converted into time domain signals.
  • the audio signal of the time domain after the frequency band is synthesized.
  • the decoding system can decode only the second layer code stream when the code rate is low, and obtain the BWE coded information to ensure the basic high-band quality; In the high case, the third layer code stream can be further decoded to obtain better high band quality.
  • the quantizer performs bit allocation, and allocates a large number of bits to some important frequency bands for high-accuracy quantization. And assigning fewer bits to some less important bands for less accurate quantization, even assigning bits to some less important bands, that is, not doing more important band quantizers for this part. Quantify.
  • the embodiment of the invention provides a signal denoising method and device and an audio decoding system, which can reduce the noise of the frequency band synthesis after decoding and improve the hearing effect.
  • the method for denoising a signal includes: selecting at least two spectral coefficients having high correlation with a spectral coefficient to be adjusted according to an inter-frame correlation of a frame in which a spectral coefficient to be adjusted is located; The at least two spectral coefficients are weighted with the to-be-adjusted spectral coefficients to obtain a predicted value of the spectral coefficient to be adjusted; the decoded signal is spectrally adjusted by using the obtained predicted value, and the adjusted decoded signal is output.
  • the device for denoising a signal includes: a selecting unit, configured to select at least two spectral coefficients having high correlation with a spectral coefficient to be adjusted according to an inter-frame correlation of a frame in which a spectral coefficient to be adjusted is located; a weighting unit, configured to perform weighting by using at least two spectral coefficients selected by the selecting unit and the spectral coefficient to be adjusted, to obtain a predicted value of the spectral coefficient to be adjusted; and adjusting an output unit, configured to use the prediction obtained by using the weighting unit The value is spectrally adjusted to the decoded signal, and the adjusted decoded signal is output.
  • the audio decoding system includes a core decoder, a bandwidth extension decoder, a dequantization decoder, and the foregoing signal denoising device, wherein the core decoder is configured to decode information of a low frequency first layer code stream.
  • the bandwidth extension decoder is configured to decode information of the bandwidth-spread second layer code stream;
  • the dequantization decoder is configured to decode information of a third layer code stream that is used to dequantize the remaining bits of the high frequency band;
  • Means configured to receive the bandwidth extension decoder and the Decoding the decoded information output by the decoder, determining the spectral coefficients to be adjusted in the decoded information, and adjusting the spectral coefficients in the decoded information according to the obtained predicted values of the spectral coefficients to be adjusted.
  • the predicted value of the spectral coefficient to be adjusted is obtained by weighting at least two related spectral coefficients of the spectral coefficient to be adjusted, and the decoded signal is adjusted according to the predicted value of the spectral coefficient to be adjusted.
  • the spectrum is adapted to the predicted spectral coefficients (ie, the predicted values of the spectral coefficients to be adjusted) and other related spectral coefficients, so that the spectral coefficients obtained by different quantization precisions are mutually adapted, thereby increasing the smoothness of the decoded signal spectrum and reducing Understand the noise of the band synthesis after the code, so that the audio signal after the band synthesis can achieve better hearing effect.
  • FIG. 1 is a structural diagram of a conventional audio coding system
  • FIG. 2 is a structural diagram of an existing audio decoding system
  • FIG. 3 is a schematic flowchart of a signal denoising method according to Embodiment 1 of the present invention
  • FIG. 4 is a schematic flowchart of a signal denoising method according to Embodiment 2 of the present invention
  • FIG. 6 is a structural diagram of an audio decoding system according to Embodiment 5 of the present invention. detailed description
  • an embodiment of the present invention provides a method for signal denoising, including:
  • Step 31 Select at least two spectral coefficients having high correlation with the spectral coefficients to be adjusted according to the inter-frame correlation of the frame in which the spectral coefficients to be adjusted are located;
  • Step 32 Perform weighting by using at least two selected spectral coefficients and the spectral coefficients to be adjusted, and obtain a predicted value of the spectral coefficient to be adjusted;
  • Step 33 Perform spectrum adjustment on the decoded signal by using the obtained predicted value, and output the adjusted decoded signal.
  • the signal denoising method provided by the embodiment of the present invention obtains a predicted value of the spectral coefficient to be adjusted by weighting at least two related spectral coefficients of the adjusted spectral coefficient, and adjusts the decoded signal according to the predicted value of the spectral coefficient to be adjusted.
  • the spectrum is adapted to the predicted spectral coefficients (ie, the predicted values of the spectral coefficients to be adjusted) and other related spectral coefficients, so that the spectral coefficients obtained by different quantization precisions are mutually adapted, thereby increasing the smoothness of the decoded signal spectrum and reducing Understand the noise of the band synthesis after the code, so that the audio signal after the band synthesis can achieve better hearing effect.
  • a method for signal denoising includes: Step 41: Determine a to-be-adjusted spectral coefficient in a decoded signal according to a quantization precision of a spectral coefficient.
  • a decoded signal is output, which is a low frequency signal output by the core decoder and output by the B WE decoder.
  • the bandwidth extension high frequency signal and the other high frequency signals of the dequantization decoder output, wherein the bandwidth extension high frequency signal output by the BWE decoder and the other high frequency signals output by the dequantization decoder are frequency domain signals.
  • the determined spectral coefficients to be adjusted may include: unquantized spectral coefficients, and/or spectral coefficients whose quantization accuracy is lower than a certain quantization precision threshold, where the quantization precision threshold may be set as needed.
  • Lbit can only represent the symbol information of the frequency point.
  • No bit ie, O bit
  • the frequency of the bit rate is 1 bit/frequency. There is no amplitude information.
  • the quantization accuracy of the frequency can be considered. If it is 0), the frequency point is not quantized, and it is determined that the frequency point of the bit rate is 1 bit/frequency point is the frequency point to be adjusted.
  • the average quantization precision of the vector where the frequency point is located may be first determined. If the quantization precision is less than a certain lower limit, such as 0.5 bit/frequency, it is determined that all frequency points in the vector need to be adjusted; if the average quantization precision is greater than a certain The upper threshold, such as 2 bit/frequency, determines that all frequency points in the vector do not need to be adjusted; if the average quantization precision is between the two, such as between 0.5 bit/frequency and 2 bit/frequency In between, it is further determined whether there are frequency points in the vector that are not vector quantized, and if so, these frequency points that are not vector quantized are determined to be adjusted, otherwise no adjustment is needed. Step 42: Select a weighting mode from three weighting modes of high inter-frame correlation, low inter-frame correlation, and medium inter-frame correlation according to the inter-frame correlation of the frame in which the spectral coefficient to be adjusted is located.
  • a certain lower limit such as 0.5 bit/frequency
  • the upper threshold such as 2 bit/frequency
  • the correlation between the frames can be judged according to the parameters related to the correlation, for example, a BWE algorithm, which uses the frame type to characterize the inter-frame correlation: the transient type frame represents low inter-frame correlation; Harmonic type frames represent high inter-frame correlation; normal type frames represent inter-frame correlation.
  • the frame type is a correlation-related parameter, and the correlation between the frames can be determined according to the frame type, thereby determining the weighting mode.
  • the correlation between the frames can also be determined by calculation. For example, first, a certain correlation calculation method calculates the correlation between the frame of the spectrum coefficient to be adjusted and the adjacent frame. If the correlation is greater than the upper threshold, the value is to be adjusted. The inter-frame correlation of the frame in which the spectral coefficient is located is high; if the correlation is less than the lower threshold, the inter-frame correlation of the frame in which the spectral coefficient is to be adjusted is low; in other cases, for example, the correlation is between the upper threshold and the lower threshold In between, the inter-frame correlation of the frame in which the spectral coefficients are to be adjusted is adjusted.
  • step 42 different weighting modes are selected according to the inter-frame correlation: when the inter-frame correlation is high, the high inter-frame correlation weighting mode is selected; when the inter-frame correlation is low, the low inter-frame correlation is selected. Weighting mode; When inter-frame correlation is selected, the inter-frame correlation weighting mode is selected. Different weighting modes correspond to different weighting weights for weighting the inter-frame spectral coefficients and the intra-frame spectral coefficients. In general, the higher the inter-frame correlation, the higher the weighting weight of the inter-frame spectral coefficients, and the lower the weighting weight of the intra-frame frequency coefficients; the lower the inter-frame correlation, the lower the weighting weight of the inter-frame spectral coefficients. The weighting weight of the intra-frame spectral coefficients is higher.
  • the weighted weight of the inter-frame spectral coefficients is proportional to the inter-frame correlation, and the weighted weight of the intra-frame spectral information is inversely proportional to the inter-frame correlation.
  • the interframe frequency The spectral coefficient weight is large, the intra-frame spectral coefficient weight is small or zero; for low inter-frame correlation, the intra-frame spectral coefficient weight is large, and the inter-frame spectral coefficient weight is small or zero;
  • the correlation frame, the size of the intra-frame and inter-frame frequency coefficient weights can be determined by the comparison of the inter-frame and intra-frame correlation.
  • Step 43 Determine at least two spectral coefficients having high correlation with the spectral coefficients to be adjusted according to the selected weighting mode.
  • determining at least two spectral coefficients having high correlation with the spectral coefficients to be adjusted according to the weighting mode may be: when a high inter-frame correlation weighting mode is selected, indicating that the inter-frame correlation is high, At least two spectral coefficients may be determined from a frame adjacent to a frame in which the spectral coefficient to be adjusted is located; when a low inter-frame correlation weighting mode is selected, it indicates that the inter-frame correlation is low, and the frame in which the spectral coefficient to be adjusted is located may be Determining at least two spectral coefficients in the middle; when the inter-frame correlation weighting mode is selected, indicating that the inter-frame correlation can be simultaneously from the frame in which the spectral coefficient to be adjusted is located and adjacent to the frame in which the spectral coefficient to be adjusted is located At least two spectral coefficients are determined in the frame.
  • Step 44 Perform weighting by using the determined at least two spectral coefficients and the spectral coefficients to be adjusted, and obtain a predicted value of the spectral coefficient to be adjusted.
  • the method for weighting the determined at least two spectral coefficients and the spectral coefficients to be adjusted may be predicted by using weighting values of at least one of the following information: 1. Dequantizing the quantized spectral coefficients output by the decoder; 2. Outputting by the BWE decoder Bandwidth spread spectrum coefficient; 3. The predicted value of the frequency coefficient obtained by the prediction. The product of the frequency coefficient and its corresponding weighting weight is the weighting value of the frequency coefficient; since the spectral coefficient to be adjusted may be the spectral coefficient corresponding to the unquantized frequency point, at least two spectral coefficients are used in step 44.
  • the weighting value of the spectral coefficient to be adjusted may be 0, that is, only the determined weighting values of at least two spectral coefficients are used to obtain the predicted value of the spectral coefficient to be adjusted.
  • the frequency coefficient is predicted according to the weighting value of at least one of the following information: (1) the predicted value of the previous frame; (2) the quantized frequency coefficient of the previous frame; (3) The bandwidth spread spectrum factor of the previous frame.
  • the spectral coefficients are predicted according to the weighting values of at least one of the following: (1) the quantized spectral coefficients of the current frame; (2) the bandwidth spread spectrum coefficients of the current frame; (3) the current frame has Some predictions.
  • the intra-frame correlation weighting mode predicts the spectral coefficients according to the weighting values of at least one of the following information: (1) the predicted values existing in the previous frame or the current frame; (2) the quantized spectral coefficients of the previous frame or the current frame; (3) The bandwidth spread spectrum factor of the previous frame or the current frame.
  • the weighting weight of each of the above spectrum information may also be adjusted according to the quantization precision of the frequency to be adjusted.
  • the quantized result can still be weighted and the weighted weight is proportional to the quantization precision of the spectral coefficient.
  • Step 45 Control the energy of the obtained predicted value, and perform spectrum adjustment on the decoded signal.
  • the upper limit of the energy of the spectral coefficient to be adjusted is first determined, and then the energy of the adjusted spectral coefficient is controlled to be less than or equal to the upper limit.
  • the upper threshold can be determined according to the quantization error or the minimum non-zero quantization value in the range of the spectral coefficient to be adjusted, wherein the quantization error or the minimum non-zero quantization value can be obtained by the prior art, and details are not described herein again.
  • the energy of the obtained predicted value is controlled, and the spectrum adjustment of the decoded signal may be: according to the upper limit, correcting the predicted value of the spectral coefficient to be adjusted, and obtaining the corrected value of the spectral coefficient to be adjusted,
  • the energy of the correction value is within a range less than or equal to the upper limit threshold, and the modified signal is used for spectrum adjustment, wherein when the predicted value is less than or equal to the upper limit, the correction value is equal to the predicted value, and when the predicted value is greater than The correction value is equal to the upper threshold when the upper limit is exceeded.
  • the threshold value a can be determined using an empirical value based on experimental statistics, or the quantization precision can be used to control the size of a.
  • the value of the control threshold coefficient a is from 1 to a value less than one.
  • the spectral coefficients to be adjusted are determined by the quantization precision of the spectral coefficients, and different weighting modes are selected according to the inter-frame correlation of the frame in which the spectral coefficients to be adjusted are located, and determined according to the selected weighting mode.
  • the coefficient (ie, the predicted value of the spectral coefficient to be adjusted) is adapted to other related spectral coefficients, so that the spectral coefficients obtained by different quantization precisions are adapted to each other, and the spectrum of the decoded signal is increased.
  • the smoothness reduces the noise of the synthesized frequency band after decoding, and enables the audio signal after the band synthesis to achieve a better hearing effect.
  • the embodiment provides a method for performing weighted prediction on the adjusted spectral coefficients, and describes frequency information that can be used in different weighting modes, including:
  • the intra-frame spectral information is f_ inner[n]
  • the weighting weight in the frame is w_ inner [n]
  • the inter-frame spectral information is f-inter[n]
  • the weighted weight between frames is w-inter[n].
  • N is the maximum number of frequency points that a frame has; if the spectral coefficient of the frequency point n is the spectral coefficient to be adjusted, the predicted value f[n] of the frequency coefficient of the frequency point n is expressed by a formula such as Formula 1 :
  • f[n] w_inner[0] xf_inner[0] + w inner [ 1 ] x f inner [ 1 ] + ... +
  • the weight weight w- inner [n] in the frame is proportional to the intra-frame correlation; the weighted weight w-inter[n] between frames is proportional to the inter-frame correlation; and the sum of all weighted weights is 1 .
  • the following is a specific example of how to perform weighted prediction on the adjusted frequency coefficient.
  • the quantized spectral coefficient fQ [n] of the current frame intermediate frequency point n is determined as the spectral coefficient to be adjusted, and the bandwidth spread spectrum coefficient of the current frame intermediate frequency point n is ffi[n]; the quantized spectrum of the intermediate frame of the current frame of the current frame n
  • the coefficient is expressed as fS[l][n]
  • the quantized spectral coefficient of the intermediate frequency point n of the previous frame is expressed as fS[0][n]
  • the prediction of the quantized frequency coefficient of the intermediate frequency point n is f[n].
  • the above frequency coefficient or predicted value may be 0 or a non-zero number.
  • step 41 If the frequency point 17 needs to be adjusted according to the step 41 in the second embodiment, and the frame in which the frequency point is located, different weighting modes are selected according to step 42, for different weighting modes.
  • the following processing is performed, wherein the frequency point 16 and the frequency point 18 are adjacent frequency points of the frequency point 17:
  • fQ[17] (fB[17]+fQ[16]+fQ[18] )/3 , at this time, ffi[17], fQ[16], fQ[18] are The determined spectral coefficients with high correlation with the spectral coefficients to be adjusted, B[17], fQ[16], and fQ[18] have weighted weights of 1/3, 1/3, and 1/3, respectively, and other weighted prediction formulas below. The meaning in it is similar to here, no longer praised;
  • f[17] (0.4 X fB[17]+fQ[17]+0.8 ⁇ fQ[16]+0.8 x fQ[18] )/3;
  • f[17] ( 0.3 ⁇ fS[0][17]+0.7 ⁇ fS[l][17]+fQ[17] )12;
  • f[17] ( fB[17]+fQ[16]+fQ[18]+ fS[l][16]+ fS[l][17]+ fS[l] [18] )/6;
  • weighted weights and the frequency range of the values are derived from the experimental results, that is, the empirical values, and in practical applications of different scenarios, the choice of weighting weights and frequency points will vary from scene to scene, for example, different core encoders will have Different bandwidth extension ranges. Therefore, the above-mentioned inter-frame spectral information, the value range of the intra-frame spectral information, and the specific value of the weighting weight may be rooted. According to experiments in different scenarios.
  • the method for performing weighted prediction on the adjusted spectral coefficients provided in the third embodiment is described by using specific weighting weights, frequency coefficients and calculation formulas. These specific weighting weights, frequency coefficients and calculation formulas are only based on experience. The preferred implementation manners do not constitute a limitation on the scope of the present invention. In practice, these specific weighting weights, spectral coefficients, and calculation formulas can be flexibly adjusted according to specific situations, which are extensions and modifications without departing from the invention. All belong to the scope of protection of the present invention.
  • the method for performing weighted prediction on the adjusted spectral coefficients provided in the third embodiment can be applied to the embodiments of the present invention, and the weighted prediction is performed on the adjusted frequency coefficient, and the predicted value of the spectral coefficient to be adjusted is obtained.
  • a signal denoising method is provided, where the BWE algorithm and the 8-dimensional trellis vector quantization are used as an example for description, but the present invention is not limited thereto.
  • the method can also be applied to other vector quantization, such as 4-dimensional quantization.
  • the two weights here can be the empirical values obtained from the experiment.
  • the frame in which the spectrum coefficient to be adjusted is located is referred to as the current frame.
  • the current frame and the previous frame are both harmonic frames, that is, there is a high interframe correlation. Then, the current frame vector has spectral coefficients decoded, and the vector of the corresponding frequency band of the current frame has no spectral coefficients decoded.
  • the method for recovering the spectral coefficients to be adjusted may be: if the amplitude of the quantized spectral coefficients of the previous frame of the previous frame is larger than the amplitude of the quantized spectral coefficients corresponding to the previous frame by a given multiple (eg, twice),
  • the amplitude of the frequency coefficient is the weighted sum of the amplitude of the BWE spectral coefficient of the current frame and the amplitude of the quantized spectral coefficient corresponding to the previous frame, and the symbol is the symbol of the BWE spectral coefficient of the current frame; otherwise, if the previous frame of the previous frame
  • the amplitude of the corresponding quantized frequency coefficient is not greater than the amplitude of the quantized spectral coefficient corresponding to the previous frame
  • the amplitude of the spectral coefficient to be adjusted is the amplitude of the quantized frequency coefficient corresponding to the previous frame of the previous frame, before The weight of the corresponding quantized frequency coefficient of one frame, and the weighted sum of the amplitude
  • the method for recovering the spectral coefficient to be adjusted of the frequency point may be: obtaining a weighted average of the amplitude of the BWE spectral coefficient of the current frequency point and the amplitude of the quantized spectral coefficient of the adjacent frequency point.
  • the value En as the magnitude of the spectral coefficient to be adjusted.
  • the current frequency point is the frequency point at which the spectrum coefficient to be adjusted is located, and may be referred to as the frequency point to be adjusted.
  • the adjacent frequency point may be a frequency point in the same frame that is higher than the frequency to be adjusted or lower in frequency, and may be one or several One. If En is greater than the threshold thr[i], then En is set to thr[i], and the amplitude of the spectral coefficient to be adjusted is set to thr[i].
  • the sign of the spectral coefficient to be adjusted is the sign of the BWE frequency coefficient of the frequency.
  • the amplitude of the spectral coefficient to be adjusted is multiplied by the sign of the spectral coefficient to be adjusted as the adjustment result of the frequency point.
  • the recovery method of the spectral coefficient to be adjusted of the frequency point may be: the amplitude of the BWE spectral coefficient of the current frequency point, and the BWE of the current frequency point adjacent to the current frequency point.
  • the current frequency point is the frequency point at which the spectrum coefficient to be adjusted is located, and may be referred to as the frequency point to be adjusted.
  • the adjacent frequency point may be a frequency point in the same frame that is higher than the frequency to be adjusted or lower in frequency, and may be one or several One. If En is greater than the threshold thr[i], then En is set to thr[i], and the amplitude of the spectral coefficient to be adjusted is set to thr[i].
  • the symbol of the spectral coefficient to be adjusted is the sign of the BWE frequency coefficient of the frequency point.
  • the amplitude of the spectral coefficient to be adjusted is multiplied by the sign of the spectral coefficient to be adjusted as the result of the frequency adjustment.
  • the weighting coefficients in the weighting operation may be different to control the degree of spectral coefficient adjustment so that it does not affect the auditory resolution of the quantized spectral coefficients, and does not introduce additional noise.
  • the present invention further provides an apparatus for signal denoising.
  • the method includes:
  • the selecting unit 51 is configured to select at least two spectral coefficients having high correlation with the spectral coefficient to be adjusted according to the inter-frame correlation of the frame in which the spectral coefficient to be adjusted is located;
  • a weighting unit 52 configured to perform weighting by using at least two spectral coefficients selected by the selecting unit 51 and the to-be-adjusted spectral coefficients, to obtain a predicted value of the spectral coefficient to be adjusted;
  • the adjustment output unit 53 is configured to perform spectrum adjustment on the decoded signal by using the predicted value obtained by the weighting unit 52, and output the adjusted decoded signal.
  • the device further comprises: a prediction point determining unit 50, configured to determine a spectral coefficient to be adjusted according to a quantization coding precision of the spectral coefficient, where the determined spectral coefficient to be adjusted includes: an unquantized spectral coefficient, and/or a spectrum whose quantization precision is lower than a quantization precision threshold coefficient.
  • the selecting unit 51 includes:
  • the weighting mode selection module 511 is configured to select one of three weighting modes: high inter-frame correlation, low inter-frame correlation, or medium inter-frame correlation according to the inter-frame correlation of the frame in which the spectral coefficient to be adjusted is located.
  • Weighting mode high inter-frame correlation, low inter-frame correlation, or medium inter-frame correlation according to the inter-frame correlation of the frame in which the spectral coefficient to be adjusted is located.
  • the correlation spectrum selection module 512 is configured to determine, according to the weighting mode selected by the weighting mode selection module 51 1 , at least two spectral coefficients having high correlation with the spectral coefficient to be adjusted.
  • the weighting unit 52 includes any of the following modules:
  • the high correlation weighting module 521 is configured to obtain a prediction value of the to-be-adjusted spectral coefficient according to a weighting value of at least one of the following information: (1) a predicted value of the previous frame; (2) The quantized spectral coefficient of the frame; (3) the bandwidth spread spectrum coefficient of the previous frame; or, the low correlation weighting module 522, the weighting mode for the correlation between the low frames, obtaining the to-be-tuned according to the weighting value of at least one of the following information
  • the predicted value of the spectral coefficient (1) the quantized spectral coefficient of the current frame; (2) the bandwidth spread spectrum coefficient of the current frame; (3) the predicted value existing in the current frame; or, the correlation weighting module 523, used for centering a weighting mode of inter-frame correlation, obtaining a predicted value of the spectral coefficient to be adjusted according to a weighting value of at least one of the following information: (1) a predicted value of a previous frame or a current frame; (2) a quantization frequency of a previous
  • the weighting weight of the spectrum information used in each of the above correlation weighting modules is controlled by the quantization precision of the spectrum coefficient to be adjusted, and the higher the quantization precision of the frequency information is
  • the weighting weight should be larger, and its weighting weight is proportional to the quantization precision of the spectral coefficient.
  • the product of the spectral coefficient and its corresponding weighted weight is the weighted value of the spectral coefficient.
  • the weighting unit 52 described above further includes:
  • the weight control module 520 is configured to control the weighting weight of the spectrum information according to the quantization precision of the spectrum coefficient to be adjusted. The higher the quantization precision of the spectrum information, the larger the weighting weight corresponding to the corresponding information.
  • the above adjustment output unit 53 further includes:
  • the correction module 530 is configured to generate a correction value of the to-be-adjusted spectral coefficient according to the upper threshold value of the spectral coefficient energy to be adjusted and the obtained predicted value, and perform spectrum adjustment on the decoded signal by using the modified value; wherein, the correction of the spectral coefficient to be adjusted The energy of the value is less than or equal to the upper threshold of the energy of the spectral coefficient to be adjusted.
  • the apparatus for denoising a signal by using a weighting unit to weight the at least two related spectral coefficients selected by the selecting unit to adjust the spectral coefficient, to obtain a predicted value of the spectral coefficient to be adjusted, and the adjusted output unit according to the After the predicted value of the spectrum coefficient is adjusted, the spectrum of the decoded signal is adjusted, and the adjusted decoded signal is output; the predicted spectral coefficient (ie, the predicted value of the spectral coefficient to be adjusted) is adapted to other related spectral coefficients, thereby enabling different quantization.
  • the spectral coefficients obtained by the precision are mutually adapted, the smoothness of the spectrum of the decoded signal is increased, the noise of the synthesized frequency band after decoding is reduced, and the audio signal after the frequency band synthesis can achieve a better hearing effect.
  • an embodiment of the present invention further provides an audio decoding system.
  • a core decoder 61 a bandwidth extension decoder 62, a dequantization decoder 63, and a signal denoising device 60 are included, wherein the core decoder 61 is configured to decode a low frequency first layer code stream.
  • Information the bandwidth extension decoder 62, configured to decode information of the bandwidth-spread second layer code stream; the dequantization decoder 63, configured to decode the third layer code stream of the de-quantized high-band remaining bits information;
  • the signal denoising device 60 may be the signal denoising device provided by the embodiment of the present invention, for receiving the decoded information output by the bandwidth extension decoder and the dequantization decoder, according to the decoded second
  • the information of the layer code stream and the layer 3 code stream determines the spectrum coefficient to be adjusted, and adjusts the spectrum coefficient in the information of the decoded third layer code stream according to the obtained predicted value of the spectrum coefficient to be adjusted.
  • the method in the embodiment of the present invention may be implemented in the form of a software function module, and when the software function module is sold or used as a stand-alone product, it may also be stored in a computer readable storage medium.
  • the above-mentioned storage medium may be a read only memory, a magnetic disk or an optical disk or the like.
  • the functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist physically separately, or two or more units may be integrated into one module.
  • the above integrated modules can be implemented in the form of hardware or in the form of software functional modules.
  • the integrated modules, if implemented in the form of software functional modules and sold or used as separate products, may also be stored in a computer readable storage medium.
  • the storage medium mentioned above may be a read only memory, a magnetic disk or an optical disk or the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Description

一种信号去噪的方法和装置及音频解码系统 本申请要求了 2009年 3月 31 日提交的, 申请号为 200910133808. 6 , 发明名称为 "一种信号去噪的方法和装置及音频解码系统" 的中国专利申 请的优先权, 其全部内容通过引用结合在本申请中。 技术领域
本发明涉及音频编解码技术领域, 具体而言是涉及一种信号去噪的方 法和装置及音频解码系统。 背景技术
许多宽带或宽带以上的音频编解码器, 当码率较低的时候, 宽带或者 超宽带部分的频谱使用 BWE ( Band Width Extension, 带宽扩展) 参数编 码, BWE 参数编码的特点是比特使用少, 带宽有保证, 质量可接受; 当 码率较高的时候, 将宽带或者超宽带部分的频谱进行量化编码, 量化编码 的特点是比特使用较多, 精度较高, 质量较好。
现有技术的支持宽带或者超宽带的音频编解码系统的结构图参考图 1 和图 2。 图 1为现有技术的一种支持宽带或者超宽带的音频编码系统的结 构图, 如图 1所示, 编码系统采用了分层结构: 核心编码器编码低频的信 息, 输出第一层码流; BWE 编码器使用较少的比特编码高频带频谱, 输 出第二层码流; 量化编码器使用剩下的比特量化编码高频带频谱, 输出第 三层码流。
图 2 为现有技术的一种支持宽带或者超宽带的音频解码系统的结构 图, 如图 2所示, 解码系统也采用了分层结构: 核心解码器用于解码低频 的第一层码流的信息; BWE 解码器用于解码带宽扩展的第二层码流的信 息; 去量化解码器用于解码去量化高频带剩下比特的第三层码流的信息; 最后解码系统将三层码流的频带合成, 输出频带合成后的音频信号, 由于 一般核心解码器输出的信号为时域信号, BWE 解码器、 去量化解码器输 出的信号为频域信号, 所以在频带合成时, 会把第二、 第三层码流的频域 信号转换为时域信号, 以便输出频带合成后的时域的音频信号。
在解码过程中, 对于高频带频谱信号, 解码系统在码率较低的情况下 可以只解码出第二层码流, 得到 BWE编码的信息, 保证基本的高频带质 量; 在码率较高的情况下, 可以进一步解码出第三层码流, 获得更好的高 频带质量。
在这种分层结构中, 很多情况下, 由于第三层码流留给频谱量化编码 的比特数不足, 量化器会进行比特分配, 向一些重要的频带分配较多的比 特数进行高精度量化, 而向一些不太重要的频带分配较少的比特数进行精 度较低的量化, 甚至向一些更不重要的频带不分配比特, 也就是说, 对这 部分更不重要的频带量化器不进行量化。
这部分未被量化的频带的频谱, 在现有技术中, 有几种处理方法: 1. 保留 BWE 的频谱; 2.复制一部分去量化得到的频谱, 并经能量调整后, 填充在未量化的部分; 3.将未量化的频谱设置为零或直接用噪声填充。
在实现本发明过程中, 发明人发现现有技术由于以下一种或几种原因 将引起明显的噪声和较差的听觉效果:
1、如果在未被量化的频带的频谱上保留 BWE的频谱, 会导致量化的 频谱与未量化的频带的频谱上保留下来的 BWE的频谱, 在位置信息和 /或 能量信息上不匹配, 从而引入噪声; 2、 如果大量频谱未量化, 而将其置 为零或者用噪音填充, 会在未量化的频带的频谱上直接引入噪声。 由于以 上的不匹配或者置零和噪声填充, 都会在解码后频带合成时引入一些噪 声, 降低音频信号的听觉效果。 发明内容
本发明实施例提供了一种信号去噪的方法和装置及音频解码系统, 能 够减少解码后频带合成的噪声, 提高听觉效果。
具体地, 本发明实施例提供的信号去噪的方法, 包括: 根据待调整频 谱系数所在帧的帧间相关性的高低, 选择与待调整频谱系数相关性高的至 少两个频谱系数; 采用选择的至少两个频谱系数与所述待调整频谱系数进 行加权, 获取待调整频谱系数的预测值; 利用获取的预测值对解码信号进 行频谱调整, 输出调整后的解码信号。
本发明实施例提供的信号去噪的装置, 包括: 选择单元, 用于根据待 调整频谱系数所在帧的帧间相关性的高低, 选择与待调整频谱系数相关性 高的至少两个频谱系数; 加权单元, 用于采用所述选择单元选择的至少两 个频谱系数与所述待调整频谱系数进行加权, 获取待调整频谱系数的预测 值; 调整输出单元, 用于利用所述加权单元获取的预测值对解码信号进行 频谱调整, 输出调整后的解码信号。
本发明实施例提供的音频解码系统, 包括核心解码器、 带宽扩展解码 器、 去量化解码器和上述的信号去噪装置, 其中, 所述核心解码器用于解 码低频的第一层码流的信息; 所述带宽扩展解码器用于解码带宽扩展的第 二层码流的信息; 所述去量化解码器用于解码去量化高频带剩下比特的第 三层码流的信息; 所述信号去噪装置, 用于接收所述带宽扩展解码器和所 述去量化解码器输出的解码后的信息, 在解码后的信息中, 确定待调整频 谱系数, 并根据获取的待调整频谱系数的预测值, 调整解码后信息中的频 谱系数。
由以上本发明实施例提供的技术方案可知, 通过对待调整频谱系数加 权至少两个相关的频谱系数, 来获取待调整频谱系数的预测值, 并根据该 待调整频谱系数的预测值调整解码信号的频谱, 使预测的频谱系数(即待 调整频谱系数的预测值)与其他相关的频谱系数相互适应, 从而使由不同 量化精度得到的频谱系数相互适配, 增加了解码信号频谱的平滑度, 减少 了解码后频带合成的噪声, 使频带合成后的音频信号能够达到更好的听觉 效果。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案, 下面将对 实施例或现有技术描述中所需要使用的附图作简单地介绍, 显而易见地, 下面描述中的附图仅仅是本发明的一些实施例, 对于本领域普通技术人员 来讲, 在不付出创造性劳动性的前提下, 还可以根据这些附图获得其他的 附图。
图 1为现有的音频编码系统的结构图;
图 2为现有的音频解码系统的结构图;
图 3为本发明实施例一提供的一种信号去噪的方法流程示意图; 图 4为本发明实施例二提供的一种信号去噪的方法流程示意图; 图 5为本发明实施例四提供的一种信号去噪的装置结构示意图; 图 6为本发明实施例五提供的一种音频解码系统的结构图。 具体实施方式
下面将结合本发明实施例中的附图, 对本发明实施例中的技术方案进 行清楚、完整地描述, 显然, 所描述的实施例仅仅是本发明一部分实施例, 而不是全部的实施例。 基于本发明中的实施例, 本领域普通技术人员在没 有作出创造性劳动前提下所获得的所有其他实施例, 都属于本发明保护的 范围。
实施例一
参见图 3, 本发明实施例提供一种信号去噪的方法, 包括:
步骤 31,根据待调整频谱系数所在帧的帧间相关性的高低,选择与待 调整频谱系数相关性高的至少两个频谱系数;
步骤 32, 采用选择的至少两个频谱系数与待调整频谱系数进行加权, 获取待调整频谱系数的预测值;
步骤 33, 利用获取的预测值对解码信号进行频谱调整,输出调整后的 解码信号。
本发明实施例提供的信号去噪的方法, 通过对待调整频谱系数加权至 少两个相关的频谱系数, 来获取待调整频谱系数的预测值, 并根据该待调 整频谱系数的预测值调整解码信号的频谱, 使预测的频谱系数(即待调整 频谱系数的预测值)与其他相关的频谱系数相互适应, 从而使由不同量化 精度得到的频谱系数相互适配, 增加了解码信号频谱的平滑度, 减少了解 码后频带合成的噪声, 使频带合成后的音频信号能够达到更好的听觉效 果。
实施例二 参见图 4, 本发明实施例提供的一种信号去噪的方法, 包括: 步骤 41, 根据频谱系数的量化精度确定解码信号中的待调整频谱系 数。
在解码端, 当核心解码器、 BWE解码器和去量化解码器分别对接收 的编码信号进行解码后, 输出解码信号, 该解码信号是由核心解码器输出 的低频信号、 B WE解码器输出的带宽扩展高频信号和去量化解码器输出 的其他高频信号组成, 其中, BWE解码器输出的带宽扩展高频信号和去 量化解码器输出的其他高频信号是频域信号。 确定的待调整频谱系数可以 包括: 未被量化的频谱系数, 和 /或量化精度低于某个量化精度阈值的频谱 系数, 这里的量化精度阔值可以根据需要设定。
例如, 对于标量量化, 如果解码信号的最少比特率为 1 bit/频点, 则 当一个频点只对应 lbit的频语系数时 (即该频点的比特率为 1 bit/频点 ) , 这 lbit只能表示频点的符号信息, 没有 bit位 (即 O bit )表示频点的幅度 信息, 所以比特率为 1 bit/频点的频点没有幅度大小信息 (可以认为该频 点的量化精度为 0 ) , 该频点未被量化, 则确定该比特率为 l bit/频点的频 点为需要调整的频点。 对于矢量量化, 可以首先确定频点所在矢量的平均 量化精度, 如果量化精度小于某下限阔值, 如 0.5 bit/频点, 则确定该矢量 内所有频点都需要调整; 如果平均量化精度大于某上限阔值, 如 2 bit/频 点, 则确定该矢量内所有频点都不需要调整; 如果平均量化精度介于两者 之间, 如介于 0.5 bit/频点和 2 bit/频点之间, 则进一步判断该矢量内是否 有频点未被矢量量化, 如果有, 则这些未被矢量量化的频点确定为需要调 整, 否则不需要调整。 步骤 42,根据待调整频谱系数所在帧的帧间相关性的高低,从高帧间 相关性、低帧间相关性、中帧间相关性三种加权模式中选择一种加权模式。
帧间相关性的高低可根据与相关性有关的参数来判断, 例如, 一种 BWE算法, 这种算法是利用帧类型表征其帧间相关性大小: 瞬态类型帧 代表帧间相关性低; 谐波类型帧代表帧间相关性高; 普通类型帧则代表帧 间相关性中。 在上述的 BWE算法中, 帧类型是与相关性有关的参数, 根 据帧类型即可确定帧间相关性的高低, 从而确定加权模式。
当然, 也可以通过计算确定帧间相关性的高低, 例如, 首先 ^一定 的相关性计算方法计算待调整频谱系数所在帧与相邻帧的相关性, 如果相 关性大于上限阔值, 则待调整频谱系数所在帧的帧间相关性高; 如果相关 性小于下限阔值,则待调整频谱系数所在帧的帧间相关性低;其他情况下, 例如相关性介于上限阔值和下限阔值之间, 则待调整频谱系数所在帧的帧 间相关性中。
在步骤 42中, 是根据帧间相关性的高低选用不同的加权模式: 当帧 间相关性高时, 选择高帧间相关性加权模式; 当帧间相关性低时, 选择低 帧间相关性加权模式; 当帧间相关性中时, 选择中帧间相关性加权模式。 不同的加权模式对应不同的加权权重, 用于对帧间频谱系数和帧内频谱系 数进行加权。一般的, 帧间相关性越高, 则帧间频谱系数的加权权重越高, 帧内频语系数的加权权重越低; 帧间相关性越低, 则帧间频谱系数的加权 权重越低, 帧内频谱系数的加权权重越高。
也即是说, 帧间频谱系数的加权权重与帧间相关性成正比, 帧内频谱 信息的加权权重与帧间相关性成反比。 对于高帧间相关性的帧, 其帧间频 谱系数权重较大,帧内频谱系数权重较小或置零;对于低帧间相关性的帧, 其帧内频谱系数权重较大, 帧间频谱系数权重较小或置零; 对于中帧间相 关性的帧, 其帧内和帧间频语系数权重的大小可以由帧间和帧内相关性的 高低比较确定。
步骤 43,根据选择的加权模式, 确定与待调整频谱系数相关性高的至 少两个频谱系数。
当步骤 42选择了加权模式, 根据加权模式确定与待调整频谱系数相 关性高的至少两个频谱系数可以为: 当选择了高帧间相关性加权模式, 则 表明帧间相关性高, 此时可以从与待调整频谱系数所在帧相邻的帧中确定 至少两个频谱系数; 当选择了低帧间相关性加权模式, 则表明帧间相关性 低, 此时可以从待调整频谱系数所在帧中确定至少两个频谱系数; 当选择 了中帧间相关性加权模式, 则表明帧间相关性中, 此时可以同时从待调整 频谱系数所在帧中和与待调整频谱系数所在帧相邻的帧中确定至少两个 频谱系数。
步骤 44, 采用确定的至少两个频谱系数与待调整频谱系数进行加权, 获取待调整频谱系数的预测值。
采用确定的至少两个频谱系数与待调整频谱系数进行加权的方法可 以是利用以下至少一种信息的加权值来预测: 1.去量化解码器输出的量化 频谱系数; 2.BWE解码器输出的带宽扩展频谱系数; 3.已有预测所得到的 频语系数的预测值。 其中, 频语系数与其对应的加权权重的乘积即是频语 系数的加权值; 由于待调整频谱系数可以是未量化的频点对应的频谱系 数, 因此步骤 44中采用至少两个频谱系数与待调整频谱系数进行加权时, 待调整频谱系数的加权值可以为 0, 即只采用确定的至少两个频谱系数的 加权值来获取待调整频谱系数的预测值。
具体地, 对高帧间相关性加权模式, 根据以下至少一种信息的加权值 来预测频语系数: (1)以前帧的预测值; (2)以前帧的量化频语系数; (3)以 前帧的带宽扩展频谱系数。
对低帧间相关性加权模式, 根据以下至少一种信息的加权值来预测频 谱系数: (1)当前帧的量化频谱系数; (2)当前帧的带宽扩展频谱系数; (3) 当前帧已有的预测值。
对中帧间相关性加权模式, 根据以下至少一种信息的加权值来预测频 谱系数: (1)以前帧或当前帧已有的预测值; (2)以前帧或当前帧的量化频 谱系数; (3)以前帧或当前帧的带宽扩展频谱系数。
需要说明的是, 上述每种频谱信息的加权权重也可根据待调整频点的 量化精度做相应的调整。做加权预测时,如果待调整频谱系数有量化结果, 则对此量化结果仍可进行加权预测, 且其加权权重与该频谱系数的量化精 度成正比。
步骤 45, 控制获取的预测值的能量, 对解码信号进行频谱调整。
本步骤首先确定出待调整频谱系数能量的上限阔值, 然后控制调整后 的频谱系数的能量在小于或等于所述上限阔值的范围内。 上限阔值可以根 据待调整频谱系数所在范围的量化误差或最小非零量化值来确定, 其中量 化误差或最小非零量化值可以通过现有技术获得, 这里不再赘述。
控制获取的预测值的能量, 对解码信号进行频谱调整可以是: 根据上 限阔值, 修正待调整频谱系数的预测值, 获得待调整频谱系数的修正值, 该修正值的能量在小于或等于所述上限阔值的范围内, 采用修正值对解码 信号进行频谱调整, 其中, 当预测值小于或等于上限阔值时修正值等于预 测值, 当预测值大于上限阔值时修正值等于上限阔值。
具体的, 如果预测后的频点的频谱系数能量大于待调整频谱系数能量 的上限阔值, 提取(或估计)该频点的量化误差大小 min— D或最小量化值 (指量化频谱系数中不包括零点的最小的幅度值) min— Q, 作为上限阔值 thr, 并根据实际情况确定阔值系数 a (a<=l)。 如果待调整频谱系数的预测 值的能量大于 axthr, 则将预测值的能量调整至小于或等于 axthr。在这里, 阔值系数 a的确定可以使用根据实验统计出来的经验值, 也可以由量化精 度来控制 a的大小。
且, 量化精度越低, 阔值系数 a的值越大, 当量化精度高于某一频点 时, 控制阔值系数 a的值由 1至某一小于 1的数值。 例如, 量化精度高于 1.5 bit/频点时, 设 thr=min— D, a=0.7; 量化精度低于 0.5 bit/频点时, 设 thr=min_Q , a=l ; 量化精度大于 0.5 bit/频点, 小于 1.5 bit/频点时, 设 thr=min_D, a= 1。
通过本发明提供的信号去噪的方法, 通过频谱系数的量化精度确定待 调整频谱系数, 根据待调整频谱系数所在帧的帧间相关性的高低选择不同 的加权模式, 根据选择的加权模式, 确定与待调整频谱系数相关性高的至 少两个频谱系数, 对待调整频谱系数加权来获取待调整频谱系数的预测 值, 并控制获取的预测值的能量, 对解码信号进行频谱调整, 使得预测的 频谱系数(即待调整频谱系数的预测值)与其他相关的频谱系数相互适应, 从而使由不同量化精度得到的频谱系数相互适配, 增加了解码信号频谱的 平滑度, 减少了解码后频带合成的噪声, 使频带合成后的音频信号能够达 到更好的听觉效果。
实施例三
本实施例提供了对待调整频谱系数进行加权预测的方法, 对不同加权 模式下可使用的频语信息进行说明, 包括:
帧内频谱信息为 f— inner[n], 帧内的加权权重为 w— inner [n], 帧 间频谱信息为 f— inter[n], 帧间的加权权重为 w— inter[n], 其中 0 < n < N, N 为一帧具有的最大频点数; 若频点 n的频谱系数为待调整频谱系数, 则频 点 n的频语系数的预测值 f[n], 使用公式表示如式 1 :
f[n] = w_inner[0] xf_inner[0] + w inner [ 1 ] x f inner [ 1 ] + ... +
w_inner[N] x f_inner[N] + w— inter [0] xf— inter [0] + w_inter[ 1 ] x f inter [ 1 ] + ... + w— inter [N] xf—inter[N] 式 1
其中, 帧内的力口权权重 w— inner [n]与帧内相关性成正比; 帧间的加权 权重 w— inter[n]与帧间相关性成正比; 且所有加权权重之和为 1。
下面以一个具体例子说明如何对待调整频语系数进行加权预测。
假设当前帧中频点 n的量化频谱系数 fQ [n]被确定为待调整频谱系数, 当前帧中频点 n的带宽扩展频谱系数为 ffi[n] ; 当前帧的上一帧中频点 n 的量化频谱系数表示为 fS[l][n], 上上一帧中频点 n的量化频谱系数表示 为 fS[0][n]; 当前帧中频点 n的量化频语系数的预测为 f[n]。 以上频语系数 或者预测值都可以是 0或非零数, 当 fQ[n]为零时则表示频点 n未量化。
如果依据实施例二中的步骤 41确定一频点 17需要做调整, 并对该频 点所在帧依据步骤 42选用不同的加权模式, 针对不同的加权模式情况可 进行如下的处理, 其中频点 16、 频点 18为频点 17的相邻频点:
A、 对于低帧间相关性加权模式:
如果 fQ[17]未量化, f[17] = ( fB[17]+fQ[16]+fQ[18] )/3 , 此时, ffi[17]、 fQ[16]、 fQ[18]为确定的与待调整频谱系数相关性高的频谱系数, B[17]、 fQ[16]、 fQ[18]的加权权重分别为 1/3、 1/3、 1/3 , 以下其他加权预 测公式中的含义与此处类似, 不再赞述;
如果 fQ[17]量化精度很低, 则 f[17] = (0.4 X fB[17]+fQ[17]+0.8 χ fQ[16]+0.8 x fQ[18] )/3;
B、 对于高帧间相关性加权模式:
如果 fQ[17]未量化, 贝' J f[17] = ( fS[0][17]+fS[l][17] )12;
如果 fQ[17]量化精度很低, 贝 "J
f[17] = ( 0.3 χ fS[0][17]+0.7 χ fS[l][17]+fQ[17] )12;
C、 对于中帧间相关性加权模式:
如果 fQ[17]未量化, 则 f[17] = ( fB[17]+fQ[16]+fQ[18]+ fS[l][16]+ fS[l][17]+ fS[l][18] )/6;
如果 fQ[17]量化精度很低, 则 f[17] =
( 2.5xfB[17]+fQ[16]+fQ[18]+0.5xfS[l][16]+0.5xfS[l][17]+0.5xfS[l][18] )/6 上述示例中的加权权重与取值频点范围均来自实验结果, 即经验值, 而且在不同场景的实际应用中, 加权权重和取值频点的选择会由于场景不 同而不同, 比如不同的核心编码器将具有不同的带宽扩展范围。 因此上述 的帧间频谱信息、 帧内频谱信息的取值范围和加权权重的具体数值可以根 据不同场景的实验来确定。
实施例三提供的对待调整频谱系数进行加权预测的方法, 采用了具体 的加权权重、 频语系数和计算公式进行说明, 这些具体的加权权重、 频语 系数和计算公式只是一种根据经验值得出的比较好的实现方式, 而不构成 对本发明保护范围的限定, 在实际中可以根据具体情况灵活的调整这些具 体的加权权重、 频谱系数和计算公式, 这些都为不背离本发明的扩展和变 形, 都属于本发明保护的范围。 实施例三提供的对待调整频谱系数进行加 权预测的方法可以应用于本发明各实施例中, 对待调整频语系数进行加权 预测, 并获取待调整频谱系数的预测值。
本发明提供的另一个实施例中,提供一种信号去噪方法,在此以 BWE 算法与 8维格形矢量量化适配为例进行说明, 但并不局限于此, 本发明实 施例提供的方法也可以适用于其他的矢量量化, 如 4维量化等。
首先计算 8维矢量内的需调整频谱系数的幅度上限阔值 thr[i], 其中 i 代表第 i个 8维矢量。 如果第 i个 8维矢量是全零矢量, 则 thr[i]等于权值 域系数的幅度值的加权和或均值等, 加权系数可以由窗函数求得, 也可以 由其它算术公式求得; 如果第 i个 8维矢量不是全零矢量, 则 thr[i]等于权 值乘以该矢量内的最小非零量化值。 此处两个权值可以为根据实验所得经 验值。
为叙述方便, 以下将待调整频谱系数所在帧称为当前帧。
如果当前帧和前一帧都是谐波帧, 即具有高帧间相关性。 则当前一帧 矢量有频谱系数被解码出, 而当前帧相应频段的矢量没有频谱系数被解码 出时, 待调整频谱系数的恢复方法可以为: 如果前一帧的前一帧量化频谱 系数幅度比前一帧对应的量化频谱系数的幅度大给定的倍数(如两倍)时, 待调整的频语系数的幅度为当前帧 BWE频谱系数的幅度与前一帧对应的 量化频谱系数的幅度的加权和, 符号为当前帧 BWE频谱系数的符号; 否 则, 即如果前一帧的前一帧对应的量化频语系数幅度没有比前一帧对应的 量化频谱系数的幅度大给定的倍数时, 待调整频谱系数的幅度为前一帧的 前一帧对应的量化频语系数的幅度、 前一帧对应的量化频语系数的幅度, 及当前帧 BWE频谱系数的幅度的加权和, 符号为当前帧 BWE频谱系数 的符号。
如果当前帧或者前一帧是瞬态帧, 即具有低帧间相关性。 如果某频点 的频谱系数没有被解码出, 该频点的待调整频谱系数的恢复方法可以为: 求当前频点的 BWE频谱系数的幅度与相邻频点的量化频谱系数的幅度的 加权平均值 En, 作为待调整频谱系数的幅度。 此处当前频点为待调整频 谱系数所在频点, 可以称为待调整频点, 相邻频点可以为同一帧内比待调 整频点频率高或频率低的频点, 可以为一个或几个。 如果 En大于阀值 thr[i] , 则将 En设为 thr[i], 即将待调整频谱系数的幅度设为 thr[i]。 待调 整频谱系数的符号为该频点的 BWE频语系数的符号。 将待调整频谱系数 的幅度乘以待调整频谱系数的符号作为该频点的调整结果。
如果当前帧类型不属于以上两种情况, 即具有中帧间相关性。 如果某 频点的频谱系数没有被解码出, 该频点的待调整频谱系数的恢复方法可以 为: 将当前频点的 BWE频谱系数的幅度、 当前帧中与当前频点相邻频点 的 BWE频谱系数的幅度、 当前帧的前一帧对应频点的量化频语系数的幅 度, 以及前一帧对应频点的相邻频点的量化频谱系数的幅度加权求平均值
En, 作为待调整频谱系数的幅度。 此处当前频点为待调整频谱系数所在频 点, 可以称为待调整频点, 相邻频点可以为同一帧内比待调整频点频率高 或频率低的频点, 可以为一个或几个。 如果 En大于阀值 thr[i], 则将 En 设为 thr[i], 即将待调整频谱系数的幅度设为 thr[i]。 待调整频谱系数的符 号为该频点的 BWE频语系数的符号。 将待调整频谱系数的幅度乘以待调 整频谱系数的符号作为该频点调整结果。
对于全零矢量和非全零矢量中的零点, 加权运算时的加权系数或有所 不同, 以控制频谱系数调整的程度, 使其既不影响量化频谱系数的听觉分 辨率, 又不引入额外噪声。
实施例四
在前述方法实施例的基础上, 本发明还提供一种信号去噪的装置实施 例, 参见图 5, 包括:
选择单元 51, 用于根据待调整频谱系数所在帧的帧间相关性的高低, 选择与待调整频谱系数相关性高的至少两个频谱系数;
加权单元 52, 用于采用所述选择单元 51选择的至少两个频谱系数与 所述待调整频谱系数进行加权, 获取待调整频谱系数的预测值;
调整输出单元 53, 用于利用所述加权单元 52获取的预测值对解码信 号进行频谱调整, 输出调整后的解码信号。
在选择单元 51根据待调整频谱系数所在帧的帧间相关性的高低, 选 择与待调整频谱系数相关性高的至少两个频谱系数之前, 还需根据频谱系 数的量化编码精度确定待调整频谱系数。 因此所述装置还包括: 预测点确定单元 50,用于根据频谱系数的量化编码精度确定待调整频 谱系数, 所述确定的待调整频谱系数包括: 未被量化的频谱系数, 和 /或量 化精度低于量化精度阈值的频谱系数。
一种实施例方式, 所述选择单元 51包括:
加权模式选择模块 511, 用于根据待调整频谱系数所在帧的帧间相关 性的高低, 从高帧间相关性、 低帧间相关性、 或中帧间相关性三种加权模 式中选择一种加权模式;
相关频谱选择模块 512, 用于根据所述加权模式选择模块 51 1选择的 加权模式, 确定与所述待调整频谱系数相关性高的至少两个频谱系数。
所述加权单元 52包括如下任一模块:
高相关加权模块 521, 用于对高帧间相关性的加权模式, 根据以下至 少一种信息的加权值来获取待调整频谱系数的预测值: ( 1 )以前帧的预测 值; (2)以前帧的量化频谱系数; (3)以前帧的带宽扩展频谱系数; 或者, 低相关加权模块 522, 用于对低帧间相关性的加权模式, 根据以下至 少一种信息的加权值来获取待调整频谱系数的预测值:( 1 )当前帧的量化频 谱系数; (2)当前帧的带宽扩展频谱系数; (3)当前帧已有的预测值; 或者, 中相关加权模块 523, 用于对中帧间相关性的加权模式, 根据以下至 少一种信息的加权值来获取待调整频谱系数的预测值:( 1 )以前帧或当前帧 的预测值; (2)以前帧或当前帧的量化频语系数; (3)以前帧或当前帧的带 宽扩展频谱系数。
需要说明的是, 以上各相关加权模块中所使用到的频谱信息的加权权 重, 由待调整频谱系数的量化精度所控制, 频语信息的量化精度越高其对 应的加权权重越大,且其加权权重与该频谱系数的量化精度成正比。其中, 频谱系数与其对应的加权权重的乘积即是频谱系数的加权值。
因此, 上述的加权单元 52中还包括:
权重控制模块 520, 用于根据待调整频谱系数的量化精度控制频谱信 息的加权权重, 频谱信息的量化精度越高其对应的加权权重越大。
如果预测后的频点的频谱系数能量大于待调整频谱系数能量的上限 阔值, 则需要控制调整后的频谱系数的能量在小于或等于所述上限阔值的 范围内。 因此, 上述的调整输出单元 53还包括:
修正模块 530, 用于根据待调整频谱系数能量的上限阔值和获取的预 测值生成待调整频谱系数的修正值, 利用所述修正值对解码信号进行频谱 调整; 其中, 待调整频谱系数的修正值的能量小于或等于所述待调整频谱 系数能量的上限阔值。
本发明实施例提供的信号去噪的装置, 通过加权单元对待调整频谱系 数加权由选择单元选择的至少两个相关的频谱系数, 来获取待调整频谱系 数的预测值, 并由调整输出单元根据该待调整频谱系数的预测值调整解码 信号的频谱后, 输出调整后的解码信号; 使得预测的频谱系数(即待调整 频谱系数的预测值)与其他相关的频谱系数相互适应, 从而使由不同量化 精度得到的频谱系数相互适配, 增加了解码信号频谱的平滑度, 减少了解 码后频带合成的噪声, 使频带合成后的音频信号能够达到更好的听觉效 果。
实施例五
在上述装置实施例的基础上, 本发明实施例还提供一种音频解码系 统, 参见图 6, 包括核心解码器 61、 带宽扩展解码器 62、 去量化解码器 63和信号去噪装置 60, 其中, 所述核心解码器 61, 用于解码低频的第一 层码流的信息; 所述带宽扩展解码器 62, 用于解码带宽扩展的第二层码流 的信息; 所述去量化解码器 63, 用于解码去量化高频带剩下比特的第三层 码流的信息;
所述信号去噪装置 60可以为上述本发明实施例提供的信号去噪装置, 用于接收所述带宽扩展解码器和所述去量化解码器输出的解码后的信息, 根据解码后的第二层码流和第三层码流的信息, 确定待调整频谱系数, 并 根据获取的待调整频谱系数的预测值, 调整解码后的第三层码流的信息中 的频谱系数。 更具体地可以参见上述的装置实施例, 在此不再赘述。
需要说明的是, 本发明实施例中的方法可以软件功能模块的形式实 现, 并且该软件功能模块作为独立的产品销售或使用时, 也可以存储在一 个计算机可读取存储介质中。 上述提到的存储介质可以是只读存储器, 磁 盘或光盘等。
本发明实施例中的各功能单元可以集成在一个处理模块中, 也可以是 各个单元单独物理存在, 也可以两个或两个以上单元集成在一个模块中。 上述集成的模块既可以采用硬件的形式实现, 也可以采用软件功能模块的 形式实现。 所述集成的模块如果以软件功能模块的形式实现并作为独立的 产品销售或使用时, 也可以存储在一个计算机可读取存储介质中。 上述提 到的存储介质可以是只读存储器, 磁盘或光盘等。
上述具体实施例并不用以限制本发明, 对于本技术领域的普通技术人 员来说, 凡在不脱离本发明原理的前提下, 所作的任何修改、 等同替换、 改进等, 均应包含在本发明的保护范围之内。

Claims

权 利 要 求 书
1、 一种信号去噪的方法, 其特征在于, 包括:
根据待调整频谱系数所在帧的帧间相关性, 选择与待调整频谱系数相 关性高的至少两个频谱系数;
采用选择的至少两个频谱系数与所述待调整频谱系数进行加权, 获取 待调整频谱系数的预测值;
利用获取的预测值对解码信号进行频谱调整, 输出调整后的解码信 号。
2、 根据权利要求 1所述的方法, 其特征在于, 所述根据待调整频谱 系数所在帧的帧间相关性, 选择与待调整频谱系数相关性高的至少两个频 谱系数的步骤之前, 包括:
根据频谱系数的量化编码精度确定待调整频谱系数, 所述确定的待调 整频谱系数包括: 未被量化的频谱系数,和 /或量化精度低于量化精度阔值 的频谱系数。
3、 根据权利要求 1所述的方法, 其特征在于, 所述根据待调整频谱 系数所在帧的帧间相关性, 选择与待调整频谱系数相关性高的至少两个频 谱系数的步骤, 包括:
根据待调整频谱系数所在帧的帧间相关性, 从高帧间相关性、 低帧间 相关性、 或中帧间相关性三种加权模式中选择一种加权模式;
根据选择的加权模式, 确定与所述待调整频谱系数相关性高的至少两 个频谱系数。
4、 根据权利要求 3所述的方法, 其特征在于, 所述采用选择的至少 两个频谱系数与待调整频谱系数进行加权, 获取待调整频谱系数的预测值 的步骤包括:
对高帧间相关性的加权模式, 根据以下至少一种信息的加权值来获取 待调整频谱系数的预测值: 以前帧的预测值; 以前帧的量化频语系数; 以 前帧的带宽扩展频谱系数;
对低帧间相关性的加权模式, 根据以下至少一种信息的加权值来获取 待调整频谱系数的预测值: 当前帧的量化频谱系数; 当前帧的带宽扩展频 谱系数; 当前帧已有的预测值;
对中帧间相关性的加权模式, 根据以下至少一种信息的加权值来获取 待调整频语系数的预测值: 以前帧或当前帧的预测值; 以前帧或当前帧的 量化频谱系数; 以前帧或当前帧的带宽扩展频谱系数。
5、 根据权利要求 4所述的方法, 其特征在于, 所述采用选择的至少 两个频谱系数与待调整频谱系数进行加权, 获取待调整频谱系数的预测值 的步骤还包括:
根据待调整频谱系数的量化精度控制频谱信息的加权权重, 频谱信息 的量化精度越高其对应的加权权重越大。
6、 根据权利要求 1或 4所述的方法, 其特征在于, 所述方法包括: 所述待调整频谱系数所在帧和其前一帧都是谐波帧, 所述待调整频谱 系数所在帧具有高帧间相关性;
如果所述前一帧的前一帧对应的量化频语系数幅度比所述前一帧对 应的量化频谱系数的幅度大给定的倍数, 待调整频谱系数的幅度为所述待 调整频谱系数所在帧的带宽扩展频谱系数的幅度与所述前一帧对应的量 化频谱系数的幅度的加权和, 所述待调整频谱系数的符号为所述待调整频 谱系数所在帧的带宽扩展频谱系数的符号;
如果前一帧的前一帧对应的量化频语系数幅度没有比前一帧对应的 量化频谱系数的幅度大给定的倍数, 所述待调整频谱系数的幅度为所述前 一帧的前一帧对应的量化频语系数的幅度、 所述前一帧对应的量化频语系 数的幅度, 及所述待调整频谱系数所在帧的带宽扩展频谱系数的幅度的加 权和, 所述待调整频谱系数的符号为所述待调整频谱系数所在帧的带宽扩 展频谱系数的符号。
7、 根据权利要求 1或 4所述的方法, 其特征在于, 所述方法包括: 所述待调整频谱系数所在帧或其前一帧是瞬态帧, 所述待调整频谱系 数所在帧具有低帧间相关性;
待调整频谱系数的幅度为待调整频点的带宽扩展频谱系数的幅度与 相邻频点的量化频谱系数的幅度的加权平均值; 若所述加权平均值大于待 调整频谱系数的幅度上限阔值, 则将所述待调整频谱系数的幅度设为所述 上限阔值;
所述待调整频谱系数的符号为所述待调整频点的带宽扩展频谱系数 的符号。
8、 根据权利要求 1或 4所述的方法, 其特征在于, 所述方法包括: 不属于所述待调整频谱系数所在帧和其前一帧都是谐波帧, 或所述待 调整频谱系数所在帧或其前一帧是瞬态帧的情况, 所述待调整频谱系数所 在帧具有中帧间相关性;
所述待调整频谱系数的幅度为所述待调整频点的带宽扩展频谱系数 的幅度、 所述待调整频点相邻频点的带宽扩展频谱系数的幅度、 所述待调 整频点所在帧的前一帧对应频点的量化频语系数的幅度, 以及所述前一帧 对应频点的相邻频点的量化频谱系数的幅度的加权平均值; 若所述加权平 均值大于待调整频谱系数的幅度上限阔值, 则将所述待调整频谱系数的幅 度设为所述上限阔值;
所述待调整频谱系数的符号为所述待调整频点的带宽扩展频谱系数 的符号。
9、 根据权利要求 1所述的方法, 其特征在于, 所述利用获取的预测 值对解码信号进行频谱调整包括:
根据待调整频谱系数能量的上限阔值和获取的预测值生成待调整频 谱系数的修正值, 利用所述修正值对解码信号进行频谱调整; 其中, 待调 整频谱系数的修正值的能量小于或等于所述待调整频谱系数能量的上限 阔值。
10、 一种信号去噪的装置, 其特征在于, 包括:
选择单元, 用于根据待调整频谱系数所在帧的帧间相关性, 选择与待 调整频谱系数相关性高的至少两个频谱系数;
加权单元, 用于采用所述选择单元选择的至少两个频谱系数与所述待 调整频谱系数进行加权, 获取待调整频谱系数的预测值;
调整输出单元, 用于利用所述加权单元获取的预测值对解码信号进行 频谱调整, 输出调整后的解码信号。
11、 根据权利要求 10所述的装置, 其特征在于, 所述装置还包括: 预测点确定单元, 用于根据频谱系数的量化编码精度确定待调整频谱 系数, 所述确定的待调整频谱系数包括: 未被量化的频谱系数, 和 /或量化 精度低于量化精度阈值的频谱系数。
12、 根据权利要求 10所述的装置, 其特征在于, 所述选择单元包括: 加权模式选择模块, 用于根据待调整频谱系数所在帧的帧间相关性, 从高帧间相关性、 低帧间相关性、 或中帧间相关性三种加权模式中选择一 种加权模式;
相关频谱选择模块, 用于根据所述加权模式选择模块选择的加权模 式, 确定与所述待调整频谱系数相关性高的至少两个频谱系数。
13、 根据权利要求 12所述的装置, 其特征在于, 所述加权单元包括 下任一模块:
高相关加权模块, 用于对高帧间相关性的加权模式, 根据以下至少一 种信息的加权值来获取待调整频谱系数的预测值:(1)以前帧的预测值;(2) 以前帧的量化频谱系数; (3)以前帧的带宽扩展频谱系数; 或者,
低相关加权模块, 用于对低帧间相关性的加权模式, 根据以下至少一 种信息的加权值来获取待调整频谱系数的预测值:(1)当前帧的量化频谱系 数; (2)当前帧的带宽扩展频谱系数; (3)当前帧已有的预测值; 或者,
中相关加权模块, 用于对中帧间相关性的加权模式, 根据以下至少一 种信息的加权值来获取待调整频谱系数的预测值:( 1 )以前帧或当前帧的预 测值; (2)以前帧或当前帧的量化频谱系数; (3)以前帧或当前帧的带宽扩 展频谱系数。
14、 根据权利要求 13所述的装置, 其特征在于, 所述加权单元还包 括: 权重控制模块, 用于根据待调整频谱系数的量化精度控制频谱信息的 加权权重, 频谱信息的量化精度越高其对应的加权权重越大。
15、 根据权利要求 10所述的装置, 其特征在于, 所述调整输出单元 包括:
修正模块, 用于根据待调整频谱系数能量的上限阔值和获取的预测值 生成待调整频谱系数的修正值, 利用所述修正值对解码信号进行频谱调 整; 其中, 待调整频谱系数的修正值的能量小于或等于所述待调整频谱系 数能量的上限阔值。
16、 一种音频解码系统, 其特征在于, 包括核心解码器、 带宽扩展解 码器、 去量化解码器和权利要求 10至 15任一项所述的信号去噪装置; 其 中,
所述核心解码器用于解码低频的第一层码流的信息; 所述带宽扩展解码器用于解码带宽扩展的第二层码流的信息; 所述去量化解码器用于解码去量化高频带剩下比特的第三层码流的 信息;
所述信号去噪装置, 用于接收所述带宽扩展解码器和所述去量化解码 器输出的解码后的信息; 在解码后的信息中, 确定待调整频谱系数, 并根 据获取的待调整频谱系数的预测值, 调整解码后的信息中的频谱系数。
PCT/CN2009/076155 2009-03-31 2009-12-28 一种信号去噪的方法和装置及音频解码系统 WO2010111876A1 (zh)

Priority Applications (5)

Application Number Priority Date Filing Date Title
JP2012502425A JP5459688B2 (ja) 2009-03-31 2009-12-28 復号信号のスペクトルを調整する方法、装置、および音声復号システム
KR1020117024686A KR101390433B1 (ko) 2009-03-31 2009-12-28 신호 잡음 제거 방법, 신호 잡음 제거 장치, 및 오디오 디코딩 시스템
EP09842532A EP2407965B1 (en) 2009-03-31 2009-12-28 Method and device for audio signal denoising
KR1020137015052A KR101320963B1 (ko) 2009-03-31 2009-12-28 신호 잡음 제거 방법, 신호 잡음 제거 장치, 및 오디오 디코딩 시스템
US13/248,725 US8965758B2 (en) 2009-03-31 2011-09-29 Audio signal de-noising utilizing inter-frame correlation to restore missing spectral coefficients

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN200910133808 2009-03-31
CN200910133808.6 2009-03-31

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US13/248,725 Continuation US8965758B2 (en) 2009-03-31 2011-09-29 Audio signal de-noising utilizing inter-frame correlation to restore missing spectral coefficients

Publications (1)

Publication Number Publication Date
WO2010111876A1 true WO2010111876A1 (zh) 2010-10-07

Family

ID=42827479

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2009/076155 WO2010111876A1 (zh) 2009-03-31 2009-12-28 一种信号去噪的方法和装置及音频解码系统

Country Status (5)

Country Link
US (1) US8965758B2 (zh)
EP (2) EP2555191A1 (zh)
JP (1) JP5459688B2 (zh)
KR (2) KR101390433B1 (zh)
WO (1) WO2010111876A1 (zh)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5997592B2 (ja) * 2012-04-27 2016-09-28 株式会社Nttドコモ 音声復号装置
EP2720222A1 (en) 2012-10-10 2014-04-16 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Apparatus and method for efficient synthesis of sinusoids and sweeps by employing spectral patterns
US9602841B2 (en) * 2012-10-30 2017-03-21 Texas Instruments Incorporated System and method for decoding scalable video coding
CN105976824B (zh) * 2012-12-06 2021-06-08 华为技术有限公司 信号解码的方法和设备
CN106463143B (zh) * 2014-03-03 2020-03-13 三星电子株式会社 用于带宽扩展的高频解码的方法及设备
SG11201609834TA (en) 2014-03-24 2016-12-29 Samsung Electronics Co Ltd High-band encoding method and device, and high-band decoding method and device
EP2980792A1 (en) * 2014-07-28 2016-02-03 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Apparatus and method for generating an enhanced signal using independent noise-filling
KR20200107125A (ko) * 2019-03-06 2020-09-16 삼성전자주식회사 간소화된 상관을 위한 전자 장치, 방법, 및 컴퓨터 판독가능 매체

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1276896A (zh) * 1997-09-18 2000-12-13 马特拉诺泰尔通讯公司 数字语音信号的去噪声方法
CN1463422A (zh) * 2001-06-06 2003-12-24 三菱电机株式会社 噪声抑制装置
CN1892822A (zh) * 2005-05-31 2007-01-10 日本电气株式会社 噪声抑制方法和设备
CN101010727A (zh) * 2004-06-28 2007-08-01 索尼株式会社 信号编码器件和方法与信号解码器件和方法
WO2007099222A1 (fr) * 2006-03-01 2007-09-07 Parrot Procede de debruitage d'un signal audio
CN101046964A (zh) * 2007-04-13 2007-10-03 清华大学 基于重叠变换压缩编码的错误隐藏帧重建方法
CN101067650A (zh) * 2007-06-08 2007-11-07 骆建华 基于部分频谱数据信号重构的信号去噪方法

Family Cites Families (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6014544A (ja) 1983-07-05 1985-01-25 Omron Tateisi Electronics Co スペクトラム拡散通信の受信回路
US5237587A (en) 1992-11-20 1993-08-17 Magnavox Electronic Systems Company Pseudo-noise modem and related digital correlation method
JPH09261065A (ja) * 1996-03-25 1997-10-03 Mitsubishi Electric Corp 量子化装置及び逆量子化装置及び量子化逆量子化システム
JPH1020011A (ja) 1996-07-02 1998-01-23 Oki Electric Ind Co Ltd 方位分析方法及び装置
JP3613303B2 (ja) * 1996-08-08 2005-01-26 富士通株式会社 音声情報圧縮蓄積方法及び装置
DE69712537T2 (de) 1996-11-07 2002-08-29 Matsushita Electric Industrial Co., Ltd. Verfahren zur Erzeugung eines Vektorquantisierungs-Codebuchs
US5940435A (en) 1996-11-21 1999-08-17 Dsp Group, Inc. Method for compensating filtering delays in a spread-spectrum receiver
EP0878790A1 (en) * 1997-05-15 1998-11-18 Hewlett-Packard Company Voice coding system and method
US6548465B2 (en) 2000-03-10 2003-04-15 General Electric Company Siloxane dry cleaning composition and process
US6931373B1 (en) * 2001-02-13 2005-08-16 Hughes Electronics Corporation Prototype waveform phase modeling for a frequency domain interpolative speech codec system
US7447631B2 (en) * 2002-06-17 2008-11-04 Dolby Laboratories Licensing Corporation Audio coding system using spectral hole filling
KR100524065B1 (ko) 2002-12-23 2005-10-26 삼성전자주식회사 시간-주파수 상관성을 이용한 개선된 오디오 부호화및/또는 복호화 방법과 그 장치
JP4311034B2 (ja) * 2003-02-14 2009-08-12 沖電気工業株式会社 帯域復元装置及び電話機
KR100608062B1 (ko) * 2004-08-04 2006-08-02 삼성전자주식회사 오디오 데이터의 고주파수 복원 방법 및 그 장치
WO2006062202A1 (ja) * 2004-12-10 2006-06-15 Matsushita Electric Industrial Co., Ltd. 広帯域符号化装置、広帯域lsp予測装置、帯域スケーラブル符号化装置及び広帯域符号化方法
BRPI0607646B1 (pt) * 2005-04-01 2021-05-25 Qualcomm Incorporated Método e equipamento para encodificação por divisão de banda de sinais de fala
US8112286B2 (en) * 2005-10-31 2012-02-07 Panasonic Corporation Stereo encoding device, and stereo signal predicting method
JP2008033269A (ja) * 2006-06-26 2008-02-14 Sony Corp デジタル信号処理装置、デジタル信号処理方法およびデジタル信号の再生装置
JP4769673B2 (ja) * 2006-09-20 2011-09-07 富士通株式会社 オーディオ信号補間方法及びオーディオ信号補間装置
JP5294086B2 (ja) * 2007-02-28 2013-09-18 日本電気株式会社 重み係数学習システム及び音声認識システム
JP2009047831A (ja) 2007-08-17 2009-03-05 Toshiba Corp 特徴量抽出装置、プログラムおよび特徴量抽出方法
PT2186089T (pt) * 2007-08-27 2019-01-10 Ericsson Telefon Ab L M Método e dispositivo para descodificação espetral percetual de um sinal áudio que inclui preenchimento de buracos espetrais

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1276896A (zh) * 1997-09-18 2000-12-13 马特拉诺泰尔通讯公司 数字语音信号的去噪声方法
CN1463422A (zh) * 2001-06-06 2003-12-24 三菱电机株式会社 噪声抑制装置
CN101010727A (zh) * 2004-06-28 2007-08-01 索尼株式会社 信号编码器件和方法与信号解码器件和方法
CN1892822A (zh) * 2005-05-31 2007-01-10 日本电气株式会社 噪声抑制方法和设备
WO2007099222A1 (fr) * 2006-03-01 2007-09-07 Parrot Procede de debruitage d'un signal audio
CN101046964A (zh) * 2007-04-13 2007-10-03 清华大学 基于重叠变换压缩编码的错误隐藏帧重建方法
CN101067650A (zh) * 2007-06-08 2007-11-07 骆建华 基于部分频谱数据信号重构的信号去噪方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP2407965A4 *

Also Published As

Publication number Publication date
EP2407965B1 (en) 2012-12-12
KR20120000091A (ko) 2012-01-03
KR20130086634A (ko) 2013-08-02
KR101320963B1 (ko) 2013-10-23
US8965758B2 (en) 2015-02-24
EP2555191A1 (en) 2013-02-06
EP2407965A1 (en) 2012-01-18
JP5459688B2 (ja) 2014-04-02
US20120022878A1 (en) 2012-01-26
JP2012522272A (ja) 2012-09-20
KR101390433B1 (ko) 2014-04-29
EP2407965A4 (en) 2012-01-18

Similar Documents

Publication Publication Date Title
JP5203929B2 (ja) スペクトルエンベロープ表示のベクトル量子化方法及び装置
WO2010111876A1 (zh) 一种信号去噪的方法和装置及音频解码系统
US9424851B2 (en) Frame error concealment method and apparatus and decoding method and apparatus using the same
US8612219B2 (en) SBR encoder with high frequency parameter bit estimating and limiting
AU2017201872B2 (en) Audio encoder and decoder
JP6452759B2 (ja) 先進量子化器
JP6368029B2 (ja) 雑音信号処理方法、雑音信号生成方法、符号化器、復号化器、並びに符号化および復号化システム
RU2707727C1 (ru) Устройство обработки аудиосигнала, способ обработки аудиосигнала и программа обработки аудиосигнала
WO2015196837A1 (zh) 一种音频编码方法和装置
AU2012361423A1 (en) Method, apparatus, and system for processing audio data
US9131290B2 (en) Audio coding device, audio coding method, and computer-readable recording medium storing audio coding computer program
JP5340965B2 (ja) 定常的な背景雑音の平滑化を行うための方法及び装置
CN101853664B (zh) 一种信号去噪的方法和装置及音频解码系统
US10504531B2 (en) Audio parameter quantization
KR102243217B1 (ko) 오디오 신호 부호화 방법 및 장치
AL-Rawi ADPCM: US Patents from 2010 to 2016
JP2005301002A (ja) 音声符号化情報処理装置および音声符号化情報処理プログラム

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: 09842532

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2012502425

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2009842532

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 20117024686

Country of ref document: KR

Kind code of ref document: A