Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making the every other embodiment that is obtained under the creative work prerequisite.
Embodiment one
Referring to Fig. 3, the embodiment of the invention provides a kind of method of signal denoising, comprising:
Step 31 according to the height of the frame-to-frame correlation of spectral coefficient to be adjusted place frame, is selected and two high spectral coefficients of spectral coefficient correlativity to be adjusted at least;
Step 32 adopts at least two spectral coefficients and the spectral coefficient to be adjusted selected to be weighted, and obtains the predicted value of spectral coefficient to be adjusted;
Step 33 utilizes the predicted value of obtaining that decoded signal is carried out the frequency spectrum adjustment, exports adjusted decoded signal.
The method of the signal denoising that the embodiment of the invention provides, by at least two relevant spectral coefficients of spectral coefficient weighting to be adjusted, obtain the predicted value of spectral coefficient to be adjusted, and adjust the frequency spectrum of decoded signal according to the predicted value of this spectral coefficient to be adjusted, make spectral coefficient (being the predicted value of spectral coefficient to be adjusted) the spectral coefficient phase coadaptation relevant of prediction with other, thereby make the spectral coefficient that obtains by different quantified precisions adaptive mutually, increased the smoothness of decoded signal frequency spectrum, reduced the synthetic noise of decoding back frequency band, made the sound signal after frequency band synthesizes to reach better auditory effect.
Embodiment two
Referring to Fig. 4, the method for a kind of signal denoising that the embodiment of the invention provides comprises:
Step 41 is determined spectral coefficient to be adjusted in the decoded signal according to the quantified precision of spectral coefficient.
In decoding end, when core decoder, BWE demoder with after going the quantization decoder device respectively the coded signal that receives to be decoded, the output decoder signal, this decoded signal is made up of the low frequency signal of core decoder output, the bandwidth extended high frequency signal of BWE demoder output and other high-frequency signals that go the quantization decoder device to export, wherein, the bandwidth extended high frequency signal of BWE demoder output is frequency-region signals with other high-frequency signals that go the output of quantization decoder device.The spectral coefficient of determining to be adjusted can comprise: the spectral coefficient that is not quantized, and/or quantified precision is lower than the spectral coefficient of certain quantified precision threshold value, the quantified precision threshold value here can be set as required.
For example, for scalar quantization, if the least bits rate of decoded signal is the 1bit/ frequency, then when a frequency only (bit rate of this frequency is the 1bit/ frequency) during the spectral coefficient of corresponding 1bit, this 1bit can only represent the symbolic information of frequency, the amplitude information that does not have bit position (being 0bit) expression frequency, so being the frequency of 1bit/ frequency, bit rate do not have amplitude size information (can think that the quantified precision of this frequency is 0), this frequency is not quantized, and determines that then this bit rate is the frequency of frequency for needing to adjust of 1bit/ frequency.For vector quantization, can at first determine the average quantization precision of frequency place vector, if quantified precision less than certain lower threshold, as the 0.5bit/ frequency, determines that then all frequencies all need to adjust in this vector; If the average quantization precision greater than certain upper limit threshold, as the 2bit/ frequency, determines that then all frequencies do not need to adjust in this vector; If the average quantization precision falls between, whether as between 0.5bit/ frequency and 2bit/ frequency, then further judging has frequency not by vector quantization, if having in this vector, then these needing not to be defined as adjusting by the frequency of vector quantization, otherwise do not need to adjust.
Step 42, according to the height of the frame-to-frame correlation of spectral coefficient to be adjusted place frame, from high frame-to-frame correlation, low frame-to-frame correlation, select a kind of weighting pattern in three kinds of weighting patterns of frame-to-frame correlation.
The height of frame-to-frame correlation can judge that for example, a kind of BWE algorithm, this algorithm are to utilize frame type to characterize its frame-to-frame correlation size: the transient type frame represents frame-to-frame correlation low according to the parameter relevant with correlativity; The harmonic wave type frame is represented the frame-to-frame correlation height; The general type frame is then represented in the frame-to-frame correlation.In above-mentioned BWE algorithm, frame type is the parameter relevant with correlativity, can determine the height of frame-to-frame correlation according to frame type, thereby determines weighting pattern.
Certainly, also can for example, at first calculate the correlativity of spectral coefficient to be adjusted place frame and consecutive frame by calculating the height of determining frame-to-frame correlation according to certain correlation calculations method, if correlativity is greater than upper limit threshold, the frame-to-frame correlation height of spectral coefficient then to be adjusted place frame; If correlativity is less than lower threshold, the frame-to-frame correlation of spectral coefficient then to be adjusted place frame is low; Under other situations, for example correlativity is between upper limit threshold and lower threshold, in the frame-to-frame correlation of spectral coefficient then to be adjusted place frame.
In step 42, be to select different weighting patterns for use: when frame-to-frame correlation is high, select high frame-to-frame correlation weighting pattern according to the height of frame-to-frame correlation; When frame-to-frame correlation is low, select low frame-to-frame correlation weighting pattern; When in the frame-to-frame correlation, frame-to-frame correlation weighting pattern in the selection.The weighting weight that different weighting patterns is corresponding different is used for spectral coefficient in interframe spectral coefficient and the frame is weighted.General, frame-to-frame correlation is high more, and then the weighting weight of interframe spectral coefficient is high more, and the weighting weight of spectral coefficient is low more in the frame; Frame-to-frame correlation is low more, and then the weighting weight of interframe spectral coefficient is low more, and the weighting weight of spectral coefficient is high more in the frame.
That is to say that the weighting weight of interframe spectral coefficient is directly proportional with frame-to-frame correlation, the weighting weight and the frame-to-frame correlation of spectrum information are inversely proportional in the frame.For the frame of high frame-to-frame correlation, its interframe spectral coefficient weight is bigger, the less or zero setting of spectral coefficient weight in the frame; For the frame of low frame-to-frame correlation, the spectral coefficient weight is bigger in its frame, the less or zero setting of interframe spectral coefficient weight; For the frame of middle frame-to-frame correlation, can be more definite in its frame by the height of interframe and in-frame correlation with the size of interframe spectral coefficient weight.
Step 43 according to the weighting pattern of selecting, is determined and two high spectral coefficients of spectral coefficient correlativity to be adjusted at least.
When step 42 has been selected weighting pattern, according to weighting pattern determine with high at least two spectral coefficients of spectral coefficient correlativity to be adjusted can for: when having selected high frame-to-frame correlation weighting pattern, then show the frame-to-frame correlation height, can determine at least two spectral coefficients this moment from the frame adjacent with spectral coefficient to be adjusted place frame; When having selected low frame-to-frame correlation weighting pattern, show that then frame-to-frame correlation is low, can determine at least two spectral coefficients this moment from the frame of spectral coefficient to be adjusted place; When having selected middle frame-to-frame correlation weighting pattern, show then in the frame-to-frame correlation that can determine at least two spectral coefficients simultaneously this moment from the spectral coefficient to be adjusted place frame neutralization frame adjacent with spectral coefficient to be adjusted place frame.
Step 44 adopts at least two spectral coefficients and the spectral coefficient to be adjusted determined to be weighted, and obtains the predicted value of spectral coefficient to be adjusted.
The method that adopts at least two definite spectral coefficients and spectral coefficient to be adjusted to be weighted can be to utilize the weighted value of following at least a information to predict: 1. the quantization spectral coefficient that goes the output of quantization decoder device; 2.BWE the bandwidth spread-spectrum coefficient of demoder output; 3. the predicted value of the resulting spectral coefficient of existing prediction.The product of the weighting weight that wherein, spectral coefficient is corresponding with it promptly is the weighted value of spectral coefficient; Because spectral coefficient to be adjusted can be the spectral coefficient of non-quantized frequency correspondence, when therefore adopting at least two spectral coefficients and spectral coefficient to be adjusted to be weighted in the step 44, the weighted value of spectral coefficient to be adjusted can be 0, and promptly the weighted value of at least two spectral coefficients that employing is determined is obtained the predicted value of spectral coefficient to be adjusted.
Particularly, to high frame-to-frame correlation weighting pattern, predict spectral coefficient: the predicted value of frame before (1) according to the weighted value of following at least a information; (2) quantization spectral coefficient of frame before; (3) the bandwidth spread-spectrum coefficient of frame before.
To low frame-to-frame correlation weighting pattern, predict spectral coefficient according to the weighted value of following at least a information: the quantization spectral coefficient of (1) present frame; (2) the bandwidth spread-spectrum coefficient of present frame; (3) the existing predicted value of present frame.
Centering frame-to-frame correlation weighting pattern is predicted spectral coefficient according to the weighted value of following at least a information: frame or the existing predicted value of present frame before (1); (2) quantization spectral coefficient of frame or present frame before; (3) the bandwidth spread-spectrum coefficient of frame or present frame before.
Need to prove that the weighting weight of above-mentioned every kind of spectrum information also can be done corresponding adjustment according to the quantified precision of frequency to be adjusted.When doing weight estimation,, then still can be weighted prediction, and its weighting weight is directly proportional with the quantified precision of this spectral coefficient to this quantized result if spectral coefficient to be adjusted has quantized result.
Step 45, the energy of the predicted value that control is obtained carries out the frequency spectrum adjustment to decoded signal.
This step is at first determined the upper limit threshold of spectral coefficient energy to be adjusted, and the energy of controlling adjusted spectral coefficient then is in being less than or equal to the scope of described upper limit threshold.Upper limit threshold can determine that wherein quantization error or minimum non-zero quantification value can obtain by prior art, repeat no more here according to the quantization error or the minimum non-zero quantification value of spectral coefficient in-scope to be adjusted.
The energy of the predicted value that control is obtained, decoded signal is carried out the frequency spectrum adjustment can be: according to upper limit threshold, revise the predicted value of spectral coefficient to be adjusted, obtain the modified value of spectral coefficient to be adjusted, the energy of this modified value adopts modified value that decoded signal is carried out the frequency spectrum adjustment, wherein in being less than or equal to the scope of described upper limit threshold, modified value equals predicted value when predicted value is less than or equal to upper limit threshold, when predicted value during greater than upper limit threshold modified value equal upper limit threshold.
Concrete, if the spectral coefficient energy of the frequency after the prediction is greater than the upper limit threshold of spectral coefficient energy to be adjusted, extract the quantization error size min_D or minimum quantization value (referring to not comprise in the quantization spectral coefficient range value of the minimum at the zero point) min_Q of (or estimation) this frequency, as upper limit threshold thr, and determine threshold value coefficient a (a<=1) according to actual conditions.If the energy of the predicted value of spectral coefficient to be adjusted is greater than a * thr, then the energy of predicted value is adjusted to and is less than or equal to a * thr.Here, the empirical value that the determining to use of threshold value coefficient a comes out according to experiment statistics also can be controlled the size of a by quantified precision.
And quantified precision is low more, and the value of threshold value coefficient a is big more, when quantizing precision and be higher than a certain frequency, the value of control threshold value coefficient a by 1 to a certain less than 1 numerical value.For example, when quantified precision is higher than the 1.5bit/ frequency, establish thr=min_D, a=0.7; When quantified precision is lower than the 0.5bit/ frequency, establish thr=min_Q, a=1; Quantified precision during less than the 1.5bit/ frequency, is established thr=min_D, a=1 greater than the 0.5bit/ frequency.
Method by signal denoising provided by the invention, determine spectral coefficient to be adjusted by the quantified precision of spectral coefficient, height according to the frame-to-frame correlation of spectral coefficient to be adjusted place frame is selected different weighting patterns, according to the weighting pattern of selecting, determine and at least two high spectral coefficients of spectral coefficient correlativity to be adjusted, spectral coefficient weighting to be adjusted is obtained the predicted value of spectral coefficient to be adjusted, and the energy of the predicted value obtained of control, decoded signal is carried out the frequency spectrum adjustment, spectral coefficient (being the predicted value of spectral coefficient to be adjusted) the spectral coefficient phase coadaptation relevant of feasible prediction with other, thereby make the spectral coefficient that obtains by different quantified precisions adaptive mutually, increased the smoothness of decoded signal frequency spectrum, reduced the synthetic noise of decoding back frequency band, made the sound signal after frequency band synthesizes to reach better auditory effect.
Embodiment three
Present embodiment provides spectral coefficient to be adjusted has been weighted forecast method, and spendable spectrum information under the different weights pattern is described, and comprising:
Suppose: spectrum information is f_inner[n in the frame], the weighting weight in the frame is w_inner[n], the interframe spectrum information is f_inter[n], the weighting weight of interframe is w_inter[n], 0≤n≤N wherein, N is the maximum frequency number that a frame has; If the spectral coefficient of frequency n is a spectral coefficient to be adjusted, then the predicted value f[n of the spectral coefficient of frequency n], make to be formulated suc as formula 1:
F[n]=w_inner[0] * f_inner[0]+w_inner[1] * f_inner[1]+... + w_inner[N] * f_inner[N]+w_inter[0] * f_inter[0]+w_inter[1] * f_inter[1]+... + w_inter[N] * f_inter[N] formula 1
Wherein, the weighting weight w_inner[n in the frame] be directly proportional with in-frame correlation; The weighting weight w_inter[n of interframe] be directly proportional with frame-to-frame correlation; And all weighting weight sums are 1.
How spectral coefficient to be adjusted is weighted prediction with an object lesson explanation below.
Suppose the quantization spectral coefficient fQ[n of present frame intermediate-frequeney point n] be confirmed as spectral coefficient to be adjusted, the bandwidth spread-spectrum coefficient of present frame intermediate-frequeney point n is fB[n]; The quantization spectral coefficient of the previous frame intermediate-frequeney point n of present frame is expressed as fS[1] [n], the quantization spectral coefficient of last previous frame intermediate-frequeney point n is expressed as fS[0] [n]; The quantization spectral coefficient of present frame intermediate-frequeney point n be predicted as f[n].Above spectral coefficient or predicted value can be 0 or non-zero number, as fQ[n] represent that then frequency n does not quantize when being zero.
If determine that according to the step 41 among the embodiment two frequency 17 need adjust, and this frequency place frame selected for use different weighting patterns according to step 42, can carry out following processing at different weighting pattern situations, its intermediate-frequeney point 16, frequency 18 are the adjacent frequency of frequency 17:
A, for low frame-to-frame correlation weighting pattern:
If fQ[17] do not quantize, f[17 then]=(fB[17]+fQ[16]+fQ[18])/3, at this moment, fB[17], fQ[16], fQ[18] for determine with the high spectral coefficient of spectral coefficient correlativity to be adjusted, B[17], fQ[16], fQ[18] the weighting weight be respectively 1/3,1/3,1/3, below implication in other weight estimation formula with similar herein, repeat no more;
If fQ[17] quantified precision is very low, f[17 then]=(0.4 * fB[17]+fQ[17]+0.8 * fQ[16]+0.8 * fQ[18])/3;
B, for high frame-to-frame correlation weighting pattern:
If fQ[17] quantize f[17 then]=(fS[0] [17]+fS[1] [17])/2;
If fQ[17] quantified precision is very low, then
f[17]=(0.3×fS[0][17]+0.7×fS[1][17]+fQ[17])/2;
C, for middle frame-to-frame correlation weighting pattern:
If fQ[17] quantize f[17 then]=(fB[17]+fQ[16]+fQ[18]+fS[1] [16]+fS[1] [17]+fS[1] [18])/6;
If fQ[17] quantified precision is very low, f[17 then]=(2.5 * fB[17]+fQ[16]+fQ[18]+0.5 * fS[1] [16]+0.5 * fS[1] [17]+0.5 * fS[1] [18])/6.
Weighting weight in the above-mentioned example and value frequency point ranges are all from experimental result, it is empirical value, and in the practical application of different scenes, the selection meeting of weighting weight and value frequency will have different bandwidth spreading ranges because scene is different and different such as different core encoder.Therefore the span of spectrum information and the concrete numerical value of weighting weight can be determined according to the experiment of different scenes in above-mentioned interframe spectrum information, the frame.
What embodiment three provided is weighted forecast method to spectral coefficient to be adjusted; adopted concrete weighting weight, spectral coefficient and computing formula to describe; these concrete weighting weight, spectral coefficient and computing formula are a kind of reasonable implementation of of rule of thumb being worth; and do not constitute qualification to protection domain of the present invention; can adjust these concrete weighting weight, spectral coefficient and computing formula as the case may be flexibly in practice; these all for not deviating from expansion of the present invention and distortion, all belong to the scope of protection of the invention.Embodiment three provides is weighted forecast method to spectral coefficient to be adjusted and can be applied in the various embodiments of the present invention spectral coefficient to be adjusted is weighted prediction, and obtains the predicted value of spectral coefficient to be adjusted.
Among another embodiment provided by the invention, a kind of signal antinoise method is provided, is adapted for example at this with BWE algorithm and 8 dimension lattice shape vector quantizations and describes, but be not limited thereto, the method that the embodiment of the invention provides also goes for other vector quantization, quantizes as 4 dimensions etc.
At first calculate the amplitude upper limit threshold thr[i of the need adjustment spectral coefficient in 8 n dimensional vector ns], wherein i represents i 8 n dimensional vector ns.If i 8 n dimensional vector ns are full zero vectors, thr[i then] equal weights and multiply by this frequency range frequency domain envelope value, described frequency domain envelope value can be the weighted sum of the range value of continuous two or more frequency coefficients or average etc., weighting coefficient can be tried to achieve by window function, also can be tried to achieve by other Arithmetic Formula; If i 8 n dimensional vector ns are not full zero vectors, then thr[i] equal weights and multiply by minimum non-zero quantification value in this vector.Two weights can be for according to experiment gained empirical value herein.
For sake of convenience, below spectral coefficient to be adjusted place frame is called present frame.
If present frame and former frame all are the harmonic wave frames, promptly has high frame-to-frame correlation.Then spectral coefficient is decoded to be gone out when last frame vector has, and the vector of present frame corresponding band is not when having that spectral coefficient is decoded to be gone out, the restoration methods of spectral coefficient to be adjusted can for: if the former frame quantization spectral coefficient amplitude of former frame is during than the big given multiple (as twice) of the amplitude of former frame corresponding quantitative spectral coefficient, the amplitude of spectral coefficient to be adjusted is the weighted sum of the amplitude of the amplitude of present frame BWE spectral coefficient and former frame corresponding quantitative spectral coefficient, and symbol is the symbol of present frame BWE spectral coefficient; Otherwise, when promptly if the former frame corresponding quantitative spectral coefficient amplitude of former frame is not than the big given multiple of the amplitude of former frame corresponding quantitative spectral coefficient, the amplitude of spectral coefficient to be adjusted is the amplitude of the former frame corresponding quantitative spectral coefficient of former frame, the amplitude of former frame corresponding quantitative spectral coefficient, and the weighted sum of the amplitude of present frame BWE spectral coefficient, symbol is the symbol of present frame BWE spectral coefficient.
If present frame or former frame are the transient state frames, promptly has low frame-to-frame correlation.If the spectral coefficient of certain frequency does not have decodedly to go out, the restoration methods of the spectral coefficient to be adjusted of this frequency can for: ask the weighted mean value En of amplitude of the quantization spectral coefficient of the amplitude of BWE spectral coefficient of current frequency and adjacent frequency, as the amplitude of spectral coefficient to be adjusted.Current herein frequency is a spectral coefficient to be adjusted place frequency, can be called frequency to be adjusted, and adjacent frequency can be one or several for same frame internal ratio frequency frequency to be adjusted height or the low frequency of frequency.If En is greater than threshold values thr[i], then En is made as thr[i], the amplitude that is about to spectral coefficient to be adjusted is made as thr[i].The symbol of spectral coefficient to be adjusted is the symbol of the BWE spectral coefficient of this frequency.The amplitude of spectral coefficient to be adjusted be multiply by the adjustment result of the symbol of spectral coefficient to be adjusted as this frequency.
If the present frame type does not belong to above two kinds of situations, promptly has middle frame-to-frame correlation.If the spectral coefficient of certain frequency does not have decodedly to go out, the restoration methods of the spectral coefficient to be adjusted of this frequency can for: with in the amplitude of the BWE spectral coefficient of current frequency, the present frame with the amplitude of the quantization spectral coefficient of the corresponding frequency of former frame of the amplitude of the BWE spectral coefficient of current frequency adjacent frequency, present frame, and the amplitude weighting of the quantization spectral coefficient of the adjacent frequency of the corresponding frequency of the former frame En that averages, as the amplitude of spectral coefficient to be adjusted.Current herein frequency is a spectral coefficient to be adjusted place frequency, can be called frequency to be adjusted, and adjacent frequency can be one or several for same frame internal ratio frequency frequency to be adjusted height or the low frequency of frequency.If En is greater than threshold values thr[i], then En is made as thr[i], the amplitude that is about to spectral coefficient to be adjusted is made as thr[i].The symbol of spectral coefficient to be adjusted is the symbol of the BWE spectral coefficient of this frequency.The amplitude of spectral coefficient to be adjusted be multiply by the symbol of spectral coefficient to be adjusted as this frequency adjustment result.
For the zero point in full zero vector and the non-full zero vector, the weighting coefficient during ranking operation or different, the degree so that the control spectral coefficient is adjusted makes its sense of hearing resolution that neither influences quantization spectral coefficient, does not introduce additional noise again.
Embodiment four
On the basis of preceding method embodiment, the present invention also provides a kind of device embodiment of signal denoising, referring to Fig. 5, comprising:
Selected cell 51 is used for the height according to the frame-to-frame correlation of spectral coefficient to be adjusted place frame, selects and at least two high spectral coefficients of spectral coefficient correlativity to be adjusted;
Weighted units 52, at least two spectral coefficients and the described spectral coefficient to be adjusted that are used to adopt described selected cell 51 to select are weighted, and obtain the predicted value of spectral coefficient to be adjusted;
Adjust output unit 53, the predicted value that is used to utilize described weighted units 52 to obtain is carried out the frequency spectrum adjustment to decoded signal, exports adjusted decoded signal.
At the height of selected cell 51 according to the frame-to-frame correlation of spectral coefficient to be adjusted place frame, select with at least two high spectral coefficients of spectral coefficient correlativity to be adjusted before, also need determine spectral coefficient to be adjusted according to the quantization encoding precision of spectral coefficient.Therefore described device also comprises:
Future position determining unit 50 is used for determining spectral coefficient to be adjusted according to the quantization encoding precision of spectral coefficient, and described definite spectral coefficient to be adjusted comprises: the spectral coefficient that is not quantized, and/or quantified precision is lower than the spectral coefficient of quantified precision threshold value.
A kind of by way of example, described selected cell 51 comprises:
Weighting pattern is selected module 511, is used for the height according to the frame-to-frame correlation of spectral coefficient to be adjusted place frame, from high frame-to-frame correlation, low frame-to-frame correlation or select a kind of weighting pattern in three kinds of weighting patterns of frame-to-frame correlation;
Relevant frequency spectrum is selected module 512, is used for selecting according to described weighting pattern the weighting pattern of module 511 selections, determines and at least two high spectral coefficients of described spectral coefficient correlativity to be adjusted.
Described weighted units 52 comprises following arbitrary module:
High related weighing module 521 is used for the weighting pattern to high frame-to-frame correlation, obtains the predicted value of spectral coefficient to be adjusted according to the weighted value of following at least a information: the predicted value of frame before (1); (2) quantization spectral coefficient of frame before; (3) the bandwidth spread-spectrum coefficient of frame before; Perhaps,
Low related weighing module 522 is used for the weighting pattern to low frame-to-frame correlation, obtains the predicted value of spectral coefficient to be adjusted according to the weighted value of following at least a information: the quantization spectral coefficient of (1) present frame; (2) the bandwidth spread-spectrum coefficient of present frame; (3) the existing predicted value of present frame; Perhaps,
In related weighing module 523, be used for the weighting pattern of centering frame-to-frame correlation, obtain the predicted value of spectral coefficient to be adjusted according to the weighted value of following at least a information: the predicted value of frame or present frame before (1); (2) quantization spectral coefficient of frame or present frame before; (3) the bandwidth spread-spectrum coefficient of frame or present frame before.
Need to prove, more than the weighting weight of the spectrum information that used in each related weighing module, quantified precision by spectral coefficient to be adjusted is controlled, high more its corresponding weighting weight of the quantified precision of spectrum information is big more, and its weighting weight is directly proportional with the quantified precision of this spectral coefficient.The product of the weighting weight that wherein, spectral coefficient is corresponding with it promptly is the weighted value of spectral coefficient.
Therefore, also comprise in the above-mentioned weighted units 52:
Weight control module 520 is used for the weighting weight according to the quantified precision control spectrum information of spectral coefficient to be adjusted, and high more its corresponding weighting weight of the quantified precision of spectrum information is big more.
If the spectral coefficient energy of the frequency after the prediction then needs to control the energy of adjusted spectral coefficient in being less than or equal to the scope of described upper limit threshold greater than the upper limit threshold of spectral coefficient energy to be adjusted.Therefore, above-mentioned adjustment output unit 53 also comprises:
Correcting module 530 is used for generating according to the upper limit threshold of spectral coefficient energy to be adjusted and the predicted value obtained the modified value of spectral coefficient to be adjusted, utilizes described modified value that decoded signal is carried out the frequency spectrum adjustment; Wherein, the energy of the modified value of spectral coefficient to be adjusted is less than or equal to the upper limit threshold of described spectral coefficient energy to be adjusted.
The device of the signal denoising that the embodiment of the invention provides, at least two relevant spectral coefficients spectral coefficient weighting to be adjusted being selected by selected cell by weighted units, obtain the predicted value of spectral coefficient to be adjusted, and, export adjusted decoded signal by after adjusting the frequency spectrum of output unit according to the predicted value adjustment decoded signal of this spectral coefficient to be adjusted; Spectral coefficient (being the predicted value of spectral coefficient to be adjusted) the spectral coefficient phase coadaptation relevant of feasible prediction with other, thereby make the spectral coefficient that obtains by different quantified precisions adaptive mutually, increased the smoothness of decoded signal frequency spectrum, reduced the synthetic noise of decoding back frequency band, made the sound signal after frequency band synthesizes to reach better auditory effect.
Embodiment five
On the basis of said apparatus embodiment, the embodiment of the invention also provides a kind of audio decoding system, referring to Fig. 6, comprise core decoder 61, bandwidth extension decoder 62, remove quantization decoder device 63 and signal denoising device 60, wherein, described core decoder 61 is used for the information of the ground floor code stream of decoded low frequency; Described bandwidth extension decoder 62 is used for the information of the second layer code stream of decode bandwidth expansion; The described quantization decoder device 63 that goes, the information of the 3rd layer bit stream that quantizes the remaining bit of high frequency band of being used to decode;
The signal denoising device that described signal denoising device 60 can provide for the invention described above embodiment, be used to receive described bandwidth extension decoder and described decoded information of going the output of quantization decoder device, information according to decoded second layer code stream and the 3rd layer bit stream, determine spectral coefficient to be adjusted, and, adjust the spectral coefficient in the information of decoded the 3rd layer bit stream according to the predicted value of the spectral coefficient to be adjusted that obtains.More specifically can not repeat them here referring to above-mentioned device embodiment.
Need to prove that the form that the method in the embodiment of the invention can software function module realizes, and this software function module can be stored in also in the computer read/write memory medium as production marketing independently or when using.The above-mentioned storage medium of mentioning can be a ROM (read-only memory), disk or CD etc.
Each functional unit in the embodiment of the invention can be integrated in the processing module, also can be that the independent physics in each unit exists, and also can be integrated in the module two or more unit.Above-mentioned integrated module both can adopt the form of hardware to realize, also can adopt the form of software function module to realize.If described integrated module realizes with the form of software function module and during as independently production marketing or use, also can be stored in the computer read/write memory medium.The above-mentioned storage medium of mentioning can be a ROM (read-only memory), disk or CD etc.
Above-mentioned specific embodiment is not in order to restriction the present invention; for those skilled in the art; all under the prerequisite that does not break away from the principle of the invention, any modification of being done, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.