CN101436407A - Method for encoding and decoding audio - Google Patents

Abstract

The invention discloses a method for coding and decoding an audio, which mainly solves the problems that the prior method for coding and decoding the audio has low compression ratio and poor reconstruction audio quality. The method comprises the following steps: adopting a time frequency translation and frequency domain filtering method or a time domain filtering and time frequency translation method to analyze an audio signal so as to obtain a frequency domain residual error signal; dividing the frequency domain residual error signal into a low frequency residual error signal and a high frequency residual error signal, and performing direct coding on the low frequency residual error signal and performing parameter coding on the high frequency residual error signal respectively; reconstructing the high frequency residual error signal through the decoded low frequency residual error signal and the decoded high frequency residual error; recombining the decoded low frequency residual error signal and the reconstructed high frequency residual error signal to obtain a reconstructed frequency domain residual error signal; and finally adopting a frequency domain inverse filtering and time frequency inverse transform method or a time frequency inverse transform and time domain inverse filtering method to obtain a reconstructed audio signal. The method eliminates the redundancy in the frequency domain residual error signal and improves the compression ratio of audio coding, the channel utilization rate and the audio transmission quality, and is applied to multimedia communication and consumer electronics.

Description

Audio encoding and decoding method

Technical field

The present invention relates to the audio encoding and decoding technique field, be specifically related to a kind of signal reconfiguring method, be used for multimedia communication and field of consumer electronics based on the reconstruct of high frequency residual error.

Background technology

In field of multimedia communication, comprise that the audio frequency of voice, especially wideband audio become one of main communication service gradually.But sound signal frequency band broad, amount of coded data are bigger, this give real-time Transmission of sound signal and effectively storage bring very big difficulty.Though audio coding algorithms such as MP3, AAC, EAAC and EAAC+ preferably sound signal carry out compressed encoding, satisfy the requirement of certain application, but also can't be competent at business such as present mobile multimedia communication that is developing and various mobile platforms preferably.So be necessary to study the audio coding algorithm of the higher and better quality of efficient.In recent years, in the research field of audio compression coding, the processing of sound signal radio-frequency component, compression and reconstruct become one of gordian technique of correlative study, and how utilizing the signal of low frequency frequency range to rebuild high-frequency signal is important research contents.

The method that prior art utilizes the frequency domain low frequency to carry out the frequency domain high-frequency reconstruction mainly contains two kinds, simply is described below:

Prior art 1

The low frequency signal of audio frequency or voice is handled by a digital filter bank, obtained one group of low frequency sub-band signal; Again this group low frequency sub-band is carried out duplicating of high-frequency signal as a monoblock signal.The clone method of whole high-frequency band signal is that high-frequency signal is divided into some frequency ranges from low to high according to frequency, and every section bandwidth with above-mentioned monoblock low frequency signal is roughly the same; Then with monoblock low frequency sub-band group continuous compound rate each section to high-frequency band.Like this, monoblock low frequency sub-band group can periodically be used several times in high-frequency band, up to the whole high-frequency band that needs to recover all be replicated finish till.Concrete mode has two kinds: the one, monoblock low frequency sub-band group is moved to corresponding high-frequency band, and the 2nd, folding earlier monoblock low frequency sub-band group, promptly put upside down subband and put in order, again monoblock low frequency sub-band group is moved to corresponding high-frequency band; In reproduction process, this dual mode may intersect use.Like this, monoblock low frequency sub-band group can periodically be used, up to the whole high-frequency band of need recovering all be replicated finish till.

Prior art 2

Low frequency signal is handled by the low-pass filter group, obtained one group of low frequency sub-band signal.Here no longer as prior art 1, the low frequency sub-band group of choosing to be done as a whole, whole section HFS that the ground continuous compound rate need be recovered, but utilize subband in the low frequency sub-band group recovers the high-frequency sub-band of some Discrete Distribution respectively accordingly.If at HFS profuse harmonic component is arranged, then the frequency of its harmonic component may be exactly the integral multiple of its corresponding fundamental frequency.Under the guidance of this thought, prior art 2 proposes, if the sub-band serial number of some subband of HFS is natural integral multiples such as 2,3,4,5, it is the corresponding relation that has multiple between some high-frequency sub-band and the low frequency sub-band, there is abundant harmonic components probably in these subbands, need emphasis to recover.Like this, the process of recovering the high-frequency sub-band of Discrete Distribution with continuous low frequency sub-band group has just been finished.At last,, also to choose the close with it low frequency sub-band of waveform, the high-frequency sub-band of omitting be recovered, thereby finished duplicating of all high-frequency sub-band for the high-frequency sub-band that this method is omitted.

No matter low frequency sub-band is carried out periodically shift copy or folding duplicating as a monoblock in above-mentioned two kinds of prior aries by prior art 1, still carrying out frequency multiplication by prior art 2 duplicates, all be mechanically to recover harmonic wave, do not consider the diversity and the variability of audio speech signal, be to extract successively according to sub-band serial number to duplicate when duplicating in addition, because the waveform of low frequency sub-band and high-frequency sub-band is original just different, so the high-frequency sub-band that is replicated is compared with original high-frequency sub-band, to there be big shape differences or peak value difference, therefore the high-frequency signal accuracy of rebuilding is not too high, influences the quality of reconstructed audio signal.

Summary of the invention

One of purpose of the present invention provides a kind of audio encoding and decoding method based on frequency domain filtering and the reconstruct of high frequency residual error, two of purpose provides a kind of audio encoding and decoding method based on time-domain filtering and the reconstruct of high frequency residual error, to avoid the deficiency of above-mentioned prior art, improve the accuracy of reconstruction high frequency residual error and the ratio of compression of audio coding decoding, guarantee the reconstructed audio signal quality.

Technical scheme of the present invention is achieved in that

Technical scheme one:

Audio encoding and decoding method based on frequency domain filtering and the reconstruct of high frequency residual error comprises the steps:

1) at coding side, the original time-domain signal of audio frequency is handled through time-frequency conversion, obtains original frequency-region signal;

2) original frequency-region signal obtains original frequency domain residual error signal through frequency domain perception Filtering Processing;

3) original frequency domain residual error signal obtains low frequency residual signals coding and high frequency residual error parameter coding, and outputs to transmission channel or storage medium through frequency-domain residual analysis and encoding process.

4) in decoding end, receive from the low frequency residual signals coding and the high frequency residual error parameter coding that output to transmission channel or storage medium, and it is carried out frequency-domain residual decoding and reconstruction processing, obtain the reconstructed frequency domain residual signals;

5) the reconstructed frequency domain residual signals is handled through frequency domain perception liftering, obtains the reconstructed frequency domain signal;

6) the reconstruct frequency-region signal is carried out the time-frequency inverse transformation and handle, obtain the audio reconstruction time-domain signal.

Technical scheme two:

Audio encoding and decoding method based on time-domain filtering and the reconstruct of high frequency residual error comprises the steps:

T1) at coding side, the original time-domain signal of audio frequency obtains original time domain residual signals through time domain perception Filtering Processing;

T2) original time domain residual signals is handled through time-frequency conversion, obtains original frequency domain residual error signal;

T3) original frequency domain residual error signal obtains low frequency residual signals coding and high frequency residual error parameter coding, and outputs to transmission channel or storage medium through frequency-domain residual analysis and encoding process.

T4) in decoding end, receive from the low frequency residual signals coding and the high frequency residual error parameter coding that output to transmission channel or storage medium, and it is carried out frequency-domain residual decoding and reconstruction processing, obtain the reconstructed frequency domain residual signals;

T5) the reconstruct frequency domain residual error signal is carried out the time-frequency inversion process, obtain reconstruct time domain residual signals;

T6) reconstruct time domain residual signals is carried out the time domain liftering and handle, obtain the audio reconstruction time-domain signal.

The audio encoding and decoding method of above-mentioned two kinds of schemes, wherein said frequency-domain residual analysis and encoding process comprise the steps:

(A1) according to equiband frequency band or critical band or sound interval frequency band band division methods frequently, earlier original frequency domain residual error signal is carried out frequency band division, the code rate of setting according to audio coder is selected a frequency band division end points then, original frequency domain residual error signal is divided into original low frequency residual signals and original high frequency residual error two parts, makes these two parts respectively have several frequency bands;

(A2) original low frequency residual signals is encoded, obtain the output of low frequency residual signals coding; Again the low frequency residual signals is coded in coding side and carries out local decode, obtain the decoded low frequency residual signals;

(A3) according to the similarity or the correlativity of decoded low frequency residual signals and original high frequency residual error, high frequency residual error parameter is analyzed, promptly select a frequency band matching strategy that decoded low frequency residual signals and original high frequency residual error are carried out the frequency band coupling, and the high frequency residual error parameter of calculating optimum coupling, obtain comprising the original high frequency residual error parameter of frequency band division method, optimum matching frequency band position, energy matching attribute, sound channel coupling parameter and interframe spreading parameter;

(A4) original high frequency residual error parameter is encoded, obtain the output of high frequency residual error parameter coding.

The audio encoding and decoding method of above-mentioned two kinds of schemes, wherein said frequency-domain residual decoding and reconstruction processing comprise the steps:

(C1) the low frequency residual signals coding that receives is decoded, obtain the decoded low frequency residual signals;

(C2) the high frequency residual error parameter coding that receives is decoded, high frequency residual error parameter obtains decoding;

(C3) utilize optimum matching frequency band position and energy matching attribute in decoded low frequency residual signals, the decoding high frequency residual error parameter, reconstructed high frequency residual error frequency band R _H(q ₀, l);

(C4) frequency band division method, sound channel coupling parameter and the interframe spreading parameter in the utilization decoding high frequency residual error parameter makes up the reconstructed high frequency residual error frequency band that obtains, and obtains reconstructed high frequency residual error;

(C5) the frequency-domain residual height frequency range boundary frequency in the utilization decoding high frequency residual error parameter makes up decoded low frequency residual signals and reconstructed high frequency residual error, obtains the reconstructed frequency domain residual signals.

The audio encoding and decoding method of above-mentioned two kinds of schemes, wherein said frequency band matching strategy comprises: the high frequency residual error frequency band mates in low frequency residual signals zone, the high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof, carry out the high frequency residual error coupling and carry out the high frequency residual error coupling by the interframe expansion by sound channel coupling.

The present invention is owing to taken into full account the albefaction characteristic of frequency domain residual error signal frequency spectrum and correlativity and the similarity between high frequency residual error and the low frequency residual signals, by selecting the matching strategy of high frequency residual error and low frequency residual signals, the reconstruction parameter of estimation high frequency residual error, and with high frequency residual error parameter reconstructed high frequency residual error accurately, and based on this, realize the high efficient coding and the decoding of sound signal, thereby ratio of compression or the code efficiency of audio sources coding have been improved, save the required transmission bandwidth of transmitting audio signal and saved the required memory capacity of stored audio signal, improved the audio compression coding quality simultaneously.

Description of drawings

Fig. 1 is based on the audio encoding and decoding method process flow diagram of frequency domain filtering;

Fig. 2 is based on the audio encoding and decoding method process flow diagram of time-domain filtering;

Fig. 3 frequency domain residual error signal is analyzed and the coding method process flow diagram;

Decoding of Fig. 4 frequency domain residual error signal and reconstructing method process flow diagram;

Fig. 5 low-and high-frequency frequency domain residual error signal and high frequency residual error reconstruct synoptic diagram;

Fig. 6 high frequency residual error frequency band carries out the matching strategy synoptic diagram in low frequency residual signals zone;

Fig. 7 high frequency residual error frequency band in low frequency residual signals zone and extended area carry out the matching strategy synoptic diagram;

Fig. 8 carries out high frequency residual error matching strategy synoptic diagram by the sound channel coupling;

Fig. 9 carries out high frequency residual error matching strategy synoptic diagram by the interframe expansion.

Embodiment

Embodiment one

Present embodiment provides a kind of method of audio coding decoding, and this method is based on the audio coding method of frequency domain filtering and the reconstruct of high frequency residual error.High frequency residual error is the HFS of frequency domain residual error signal, and the analysis of high frequency residual error and reconstruct are in order effectively to compress the data volume of frequency domain residual error signal, to improve audio-frequency signal coding and transfer efficiency.The analysis of high frequency residual error and reconstruct are according to certain rule frequency domain residual error signal to be divided into low frequency residual signals and high frequency residual error two parts, utilize the correlativity or the similarity of high frequency residual error and low frequency residual signals, extract high frequency residual error reconstruct parameters needed, abandon high frequency residual error, come reconstructed high frequency residual error with high frequency residual error parameter then, thus the whole frequency domain residual error signal of reconstruct.

With reference to Fig. 1, the audio coding decoding step of present embodiment is as follows:

Step 101, at coding side, the original time-domain signal of audio frequency is handled through time-frequency conversion, obtains original frequency-region signal.

The original time-domain signal of audio frequency is the morbid sound of the various voice signals that comprise that voice signal, sound signal or anyone ear can be heard; The frequency range of sound signal mainly at 0Hz between the 20kHz, the sample frequency of sound signal is 96kHz, 48kHz, 44.1kHz, 32kHz, 22.05kHz, 16kHz, 11.025kHz and 8kH; The coding of sound signal normally is unit with the audio frame, the size of audio frame commonly used according to practical application generally within 50 milliseconds.Spatial transform adopts but is not limited to revise discrete cosine transform, correction lapped transform and Fast Fourier Transform (FFT) method and carries out conversion.

Step 102, original frequency-region signal obtains original frequency domain residual error signal through frequency domain perception Filtering Processing.

The frequency domain perceptual filter is the frequency domain filter of reflection human hearing characteristic, and it carries out frequency domain filtering to the frequency-region signal that comes step 101, obtain on perception meaning albefaction frequency domain residual error signal, if use H _M(f) transition function of expression frequency domain perceptual filter is with the perception curve of M (f) expression by the perceptual parameters sign, then H _M(f) can be expressed as $H_M\left(f\right)=\frac{1}{M\left(f\right)},$ Wherein f represents frequency, and unit is Hz.

Step 103, frequency-domain residual analysis and coding.

Shown in 3, the concrete steps of frequency-domain residual analysis and coding comprise:

Step 301, frequency band division and low-and high-frequency residual signals are cut apart, promptly according to methods such as equiband frequency band or critical band or sound interval frequency bands, earlier original frequency domain residual error signal is carried out frequency band division, select the end points of one of them divided band that original frequency domain residual error signal is divided into original low frequency residual signals and original high frequency residual error two parts according to the code rate of audio coder then, make these two parts respectively have several frequency bands.

As shown in Figure 5, complete frequency domain residual error signal is represented with R, for effectively to frequency domain residual error signal analyze, coding and reconstruct, frequency domain residual error signal is carried out frequency band division according to methods such as equiband frequency band or critical band or sound interval frequency bands, having certain attribute, be divided in the identical frequency band as the adjacent frequency components of auditory properties and go.If f _Ci, f _LiAnd f _HiBe respectively centre frequency, low side edge frequency and the high-end edge frequency of i frequency band, b _iBe band bandwidth, select the low side edge frequency f of certain frequency band _LiOr high-end edge frequency f _HiBe boundary frequency f _D, frequency domain residual error signal is divided into low frequency residual signals R in frequency domain _LWith high frequency residual error R _HTwo parts are 0 to f _DBetween just have N continuous _LIndividual low frequency residual signals frequency band is at f _DWith f _ZBetween just have N continuous _HIndividual high frequency residual error frequency band is higher than frequency f _ZThe high frequency residual error frequency band in signal be zero.If f represents frequency, unit is H _Z, f _SThe expression sample frequency, normalized frequency is used $\hat{f}=f/f_S$ Expression; If use f _BThe bandwidth of expression time-domain audio signal, then ${\hat{f}}_B=f_B/f_S=0.5.$

Step 302 is encoded to original low frequency residual signals, obtains the output of low frequency residual signals coding.The coding of original low frequency residual signals adopts the various genuine coding methods of losing, and as linear or non-linear scalar quantization coding, vector quantization coding, perhaps adopts various undistorted coding methods simultaneously, as Huffman coding and arithmetic coding;

Step 303 is coded in coding side to the low frequency residual signals and carries out local decode, obtains the coding of decoded low frequency residual signals and low frequency residual signals;

Step 304, similarity or correlativity according to decoded low frequency residual signals and original high frequency residual error, high frequency residual error parameter is analyzed, promptly select a frequency band matching strategy that decoded low frequency residual signals and original high frequency residual error are carried out the frequency band coupling, and the high frequency residual error parameter of calculating optimum coupling, obtain original high frequency residual error parameter.

As shown in Figure 5, frequency band coupling is exactly to seek and the immediate frequency band of high frequency residual error frequency bandwidth characteristics in the low frequency residual signals, and purpose is in decoding end reconstructed high frequency residual error in high quality.Frequency band coupling adopts different frequency band matching strategies, and so-called frequency band matching strategy is meant how at low frequency residual signals R _LIn select one section frequency domain residual error signal, and come reconstructed high frequency residual error R with it _HIn have the method for the high frequency residual error of same frequency band width.In Fig. 5, because frequency f _ZAbove residual signals makes zero, so there is no need this part residual error to be reconstructed again.

Described frequency band matching strategy includes but not limited to following several:

(P1) the high frequency residual error frequency band mates in low frequency residual signals zone

As shown in Figure 6, will be in f _DWith f _ZBetween N _HIndividual high frequency residual error frequency band is being in 0 to f respectively independently _DBetween low frequency residual signals region R _LMate, seek the optimum matching frequency band of error minimum or correlativity maximum.At this moment, frequency band division method, optimum matching frequency band position and energy matching attribute etc. are high frequency residual error reconstruction parameter.

(P2) the high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof

As shown in Figure 7, adopt above-mentioned frequency band matching strategy (P1), will be in f earlier _DWith f _ZBetween the minimum high frequency residual error frequency band R of frequency _H1Be in 0 to f _DBetween low frequency residual signals region R _LMate, seek the optimum matching frequency band of error minimum or correlativity maximum, and come the minimum high frequency residual error frequency band R of reconfiguration frequency with this optimum matching frequency band _H1Again that the frequency of this reconstruct is minimum high frequency residual error frequency band R _H1Add original low frequency residual signals R _L, form a new low frequency residual signals R who forms _L1=R _L+ R _H1Then,, adopt above-mentioned frequency band matching strategy (P1), continue the high frequency residual error frequency band of reconstruct higher frequency based on the low frequency residual signals of this new composition; The rest may be inferred, until finishing whole f _DTo f _ZBetween the coupling and the reconstruct of all high frequency residual error frequency bands.At this moment, frequency band division method, optimum matching frequency band position, energy matching attribute and low frequency residual error area extension parameter are high frequency residual error reconstruction parameter.

(P3) carry out the high frequency residual error coupling by the sound channel coupling

As shown in Figure 8, establishing audio sources is the multichannel audio information source with the individual main sound channel of C (C 〉=2), just has 5 main sound channels and a low stress to imitate sound channel such as 5.1 channel audios, at this moment has the low frequency residual signals of C sound channel.Because the sound signal of each sound channel has big correlativity, so the frequency domain residual error signal of each sound channel also has big correlativity, therefore any one sound channel not only can utilize the low frequency residual signals of its place sound channel to carry out high frequency residual error reconstruct, but also can utilize the low frequency residual signals of other sound channel to carry out high frequency residual error reconstruct; Can obtain more match selection like this, improve the high frequency residual error reconstruction quality.With the 1st sound channel is that example is given explanation, at first can adopt above-mentioned frequency band matching strategy (P1) and (P2), utilizes being in of place sound channel 0 to arrive f _DBetween low frequency residual signals R _LCoupling and reconstruct are in f _DWith f _ZBetween high frequency residual error.That in addition, can also utilize other sound channel is in 0 to f _DBetween low frequency residual signals R _LBe in f in coupling and reconstruct the 1st sound channel _DWith f _ZBetween high frequency residual error.The same with the 1st sound channel, the frequency domain residual error signal of all C sound channels is many can to adopt identical method to handle.

(P4) carry out the high frequency residual error coupling by the interframe expansion

Above-mentioned frequency band matching strategy (P1), (P2) and (P3) matching strategy of described high frequency residual error be the high frequency residual error of utilizing the low frequency residual signals reconstruct present frame of present frame.Because have very big correlativity between the consecutive frame sound signal usually, so the residual error of consecutive frame also has bigger correlativity.Therefore, the high frequency residual error of current audio frame not only can be reconstructed with the low frequency residual signals of present frame, also can be reconstructed with the reconstruct residual signals that comprises reconstruct low frequency residual signals and reconstructed high frequency residual error of some frames before the present frame.As shown in Figure 9, adopt low frequency residual signals that above-mentioned frequency band matching strategy (P1), (P2) and method (P3) utilize the j frame that the high frequency residual error of j frame is reconstructed, or adopt low frequency residual signals that above-mentioned frequency band matching strategy (P1), (P2) and method (P3) utilize the j-1 frame that the high frequency residual error of j frame is reconstructed, can also adopt reconstructed high frequency residual error that above-mentioned frequency band matching strategy (P1), (P2) and method (P3) utilize the j-1 frame that the high frequency residual error of j frame is reconstructed.In addition, these methods can also expand to the j-2 frame, the j-3 frame goes.

Above-mentioned various frequency band matching strategy all need be sought the optimum matching frequency band between low frequency residual signals and high frequency residual error, said here the best can be expressed with error or distortion minimum, also can represent with the correlativity maximum.The step of the high frequency residual error parameter of concrete calculating optimum coupling comprises:

(B1) calculate normalization low frequency residual error band signal

With normalization high frequency residual error band signal

If the frequency span of high frequency residual error of mating and low frequency residual signals is L frequency, R _L(p, l), l=0,1 ... L-1 is low frequency residual signals frequency band, R _H(q, l), l=0,1 ... L-1 is the high frequency residual error frequency band, and wherein p and q represent R respectively _L(p, l) and R _H(q, the l) reference position of place frequency band, p ∈ [0, fD], q ∈ [f _D, f _Z], calculate With Wherein,

Expression R _L(p, l), l=0,1 ... L-1 is by the maximal value R of himself absolute value _{L max}(p) carry out normalization low frequency residual signals that normalization obtains, Expression R _H(p, l), l=0,1 ... L-1 is by the maximal value R of himself absolute value _{H max}(q) carry out the normalization high frequency residual error that normalization obtains.

(B2) calculate coupling distortion measure d (p, q) or coupling related function r (p, q), if carry out The matching analysis with the distortion minimum, then Pi Pei distortion measure is expressed as:

$d(p,q)=w(p,q)\·\underset{l=0}{\overset{l=L-1}{\mathrm{\Σ}}}w\left(\left|{\hat{R}}_H(q,l)\right|\right)\·{({\hat{R}}_L(p,l)-{\hat{R}}_H(q,l))}^2---\left(1\right)$

Wherein w (p q) is the frequency influence factor,

It is residual error amplitude factor of influence.If w (p, q)=1.0, $w\left(\left|{\hat{R}}_H(q,l)\right|\right)=1.0,$ Formula (14) can be reduced to so:

$d(p,q)=\underset{l=0}{\overset{l=L-1}{\mathrm{\Σ}}}{({\hat{R}}_L(p,l)-{\hat{R}}_H(q,l))}^2---\left(2\right)$

If carry out The matching analysis with the correlativity maximum, normalization low frequency residual signals

With the normalization high frequency residual error

Related function be expressed as:

$r(p,q)=w(p,q)\·\underset{l=0}{\overset{l=L-1}{\mathrm{\Σ}}}w\left(\left|{\hat{R}}_H(q,l)\right|\right)\·({\hat{R}}_L(p,l)\·{\hat{R}}_H(q,l))---\left(3\right)$

Equally, w (p, q) the expression frequency influence factor,

Expression residual error amplitude factor of influence.If w (p, q)=1.0, $w\left(\left|{\hat{R}}_H(q,l)\right|\right)=1.0,$ Formula (16) can be reduced to so:

$r(p,q)=\underset{l=0}{\overset{l=N-1}{\mathrm{\Σ}}}{\hat{R}}_L(p,l){\hat{R}}_H(q,l)---\left(4\right)$

(B3) determine best frequency band matched position p ₀And q ₀, this optimum matching is exactly distortion measure d (p, minimum value q) or related function r (p, the optimum value p of pairing p of maximal value q) and q ₀And q ₀, they have determined the frequency band that high frequency residual error and low frequency residual error are mated.

(B4) utilize following formula calculating optimum energy matching attribute:

$G_R(p_0,q_0)=\frac{R_{H\max }\left(q_0\right)}{R_{\mathrm{Lp}\max }\left(p_0\right)}---\left(5\right)$

Like this, the resulting frequency band division result of step 304, optimum matching frequency band position, frequency-domain residual height frequency range boundary frequency f _D, energy matching attribute, sound channel coupling parameter and interframe spreading parameter, be exactly original high frequency residual error parameter.

Step 305, high frequency residual error parameter coding.

The original high frequency residual error parameter that step 304 produces is encoded, obtain the output of high frequency residual error coding.The coding of original high frequency residual error parameter adopts the various genuine coding methods of losing, and as linear or non-linear scalar quantization coding, vector quantization coding, perhaps adopts various undistorted coding methods simultaneously, as Huffman coding and arithmetic coding.

Step 104, frequency-domain residual decoding and reconstruct.

In decoding end, low frequency residual signals coding that receives and high frequency residual error parameter coding obtain the reconstructed frequency domain residual signals, as shown in Figure 4 through frequency-domain residual decoding and reconstruction processing.Concrete steps comprise:

Step 401 is decoded to the low frequency residual signals coding that receives, and obtains the decoded low frequency residual signals.

Step 402 is decoded to the high frequency residual error parameter coding that receives, and the high frequency residual error parameter that obtains decoding, this decoding high frequency residual error parameter comprise frequency band division result, frequency-domain residual height frequency range boundary frequency f _D, optimum matching frequency band position, energy matching attribute, sound channel coupling parameter and interframe spreading parameter.

Step 403, high frequency residual error reconstruct.

According to decoding high frequency residual error parameter, utilize the decoded low frequency residual signals to duplicate and reconstructed high frequency residual error, concrete steps comprise:

(D1) the decoded low frequency residual signals that obtains according to step 401 and 402, optimum matching frequency band position and energy matching attribute, carry out high frequency residual error with following formula and duplicate with energy flux matched:

R _H(q ₀，l)＝G _R(p ₀，q ₀)·R _L(p ₀，l)，l＝0，1，...L-1 (6)

Obtain reconstructed high frequency residual error frequency band R _H(q ₀, l);

(D2) the frequency band division result who obtains according to step 401 and 402, sound channel coupling parameter and interframe spreading parameter make up all reconstructed high frequency residual error frequency bands, obtain reconstructed high frequency residual error.

Step 404, the frequency-domain residual that obtains according to decoding height frequency range boundary frequency f _D, decoded low frequency residual signals and reconstructed high frequency residual error are made up, obtain the reconstructed frequency domain residual signals.

Step 105, the reconstructed frequency domain residual signals is handled through frequency domain inverse filtering, obtains the reconstructed frequency domain signal; If use H _R(f) expression frequency domain perception inverse filter, then H _R(f) be expressed as $H_R\left(f\right)=\frac{1}{H_M\left(f\right)}=M\left(f\right),$ Wherein f represents frequency, and unit is Hz.

Step 106 is carried out the time-frequency inverse transformation to the reconstruct frequency-region signal and is handled, and obtains the audio reconstruction time-domain signal.Corresponding with time-frequency conversion, the time domain inverse transformation adopts inverse modified discrete cosine transform, the overlapping inverse transformation of inverse modified or inverse FFT method.

Embodiment two

Present embodiment provides a kind of audio encoding and decoding method based on time-domain filtering and the reconstruct of high frequency residual error.

Referring to Fig. 2, this method step is as follows:

Step 201, at coding side, the original time-domain signal of audio frequency obtains original time domain residual signals through time domain perception Filtering Processing.Wherein, the time domain perceptual filter is the time domain filtering of reflection human hearing characteristic, and it carries out time-domain filtering to the original time-domain signal of audio frequency, obtain on perception meaning albefaction the time domain residual signals; The transition function H of time domain perceptual filter _M(z) expression, the time domain perceptual filter adopts but is not limited to linear prediction filter.

Step 202, original time domain residual signals is handled through time-frequency conversion, obtains original frequency domain residual error signal, and wherein, spatial transform adopts but is not limited to revise discrete cosine transform, correction lapped transform and Fast Fourier Transform (FFT) method and carries out conversion.

Step 203, frequency-domain residual analysis and coding.

Shown in 3, the concrete steps of frequency-domain residual analysis and coding comprise:

Step 301, frequency band division and low-and high-frequency residual signals are cut apart, promptly according to methods such as equiband frequency band or critical band or sound interval frequency bands, earlier original frequency domain residual error signal is carried out frequency band division, select the end points of one of them divided band that original frequency domain residual error signal is divided into original low frequency residual signals and original high frequency residual error two parts according to the code rate of audio coder then, make these two parts respectively have several frequency bands.

As shown in Figure 5, complete frequency domain residual error signal is represented with R, for effectively to frequency domain residual error signal analyze, coding and reconstruct, frequency domain residual error signal is carried out frequency band division according to methods such as equiband frequency band or critical band or sound interval frequency bands, having certain attribute, be divided in the identical frequency band as the adjacent frequency components of auditory properties and go.If f _Ci, f _LiAnd f _HiBe respectively centre frequency, low side edge frequency and the high-end edge frequency of i frequency band, b _iBe band bandwidth, select the low side edge frequency f of certain frequency band _LiOr high-end edge frequency f _HiBe boundary frequency f _D, frequency domain residual error signal is divided into low frequency residual signals R in frequency domain _LWith high frequency residual error R _HTwo parts are 0 to f _DBetween just have N continuous _LIndividual low frequency residual signals frequency band is at f _DWith f _ZBetween just have N continuous _HIndividual high frequency residual error frequency band is higher than frequency f _ZThe high frequency residual error frequency band in signal be zero.If f represents frequency, unit is H _Z, f _SThe expression sample frequency, normalized frequency is used $\hat{f}=f/f_S$ Expression; If use f _BThe bandwidth of expression time-domain audio signal, then ${\hat{f}}_B=f_B/f_S=0.5.$

Step 302 is encoded to original low frequency residual signals, obtains the output of low frequency residual signals coding.The coding of original low frequency residual signals adopts the various genuine coding methods of losing, and as linear or non-linear scalar quantization coding, vector quantization coding, perhaps adopts various undistorted coding methods simultaneously, as Huffman coding and arithmetic coding;

Step 303 is coded in coding side to the low frequency residual signals and carries out local decode, obtains the coding of decoded low frequency residual signals and low frequency residual signals;

Step 304, similarity or correlativity according to decoded low frequency residual signals and original high frequency residual error, high frequency residual error parameter is analyzed, promptly select a frequency band matching strategy that decoded low frequency residual signals and original high frequency residual error are carried out the frequency band coupling, and the high frequency residual error parameter of calculating optimum coupling, obtain original high frequency residual error parameter.

As shown in Figure 5, frequency band coupling is exactly to seek and the immediate frequency band of high frequency residual error frequency bandwidth characteristics in the low frequency residual signals, and purpose is in decoding end reconstructed high frequency residual error in high quality.Frequency band coupling adopts different frequency band matching strategies, and so-called frequency band matching strategy is meant how at low frequency residual signals R _LIn select one section frequency domain residual error signal, and come reconstructed high frequency residual error R with it _HIn have the method for the high frequency residual error of same frequency band width.In Fig. 5, because frequency f _ZAbove residual signals makes zero, so there is no need this part residual error to be reconstructed again.

Described frequency band matching strategy includes but not limited to following several:

(P1) the high frequency residual error frequency band mates in low frequency residual signals zone

As shown in Figure 6, will be in f _DWith f _ZBetween N _HIndividual high frequency residual error frequency band is being in 0 to f respectively independently _DBetween low frequency residual signals region R _LMate, seek the optimum matching frequency band of error minimum or correlativity maximum.At this moment, frequency band division method, optimum matching frequency band position and energy matching attribute etc. are high frequency residual error reconstruction parameter.

(P2) the high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof

As shown in Figure 7, adopt above-mentioned frequency band matching strategy (P1), will be in f earlier _DWith f _ZBetween the minimum high frequency residual error frequency band R of frequency _H1Be in 0 to f _DBetween low frequency residual signals region R _LMate, seek the optimum matching frequency band of error minimum or correlativity maximum, and come the minimum high frequency residual error frequency band R of reconfiguration frequency with this optimum matching frequency band _H1Again that the frequency of this reconstruct is minimum high frequency residual error frequency band R _H1Add original low frequency residual signals R _L, form a new low frequency residual signals R who forms _L1=R _L+ R _H1Then,, adopt above-mentioned frequency band matching strategy (P1), continue the high frequency residual error frequency band of reconstruct higher frequency based on the low frequency residual signals of this new composition; The rest may be inferred, until finishing whole f _DTo f _ZBetween the coupling and the reconstruct of all high frequency residual error frequency bands.At this moment, frequency band division method, optimum matching frequency band position, energy matching attribute and low frequency residual error area extension parameter are high frequency residual error reconstruction parameter.

(P3) carry out the high frequency residual error coupling by the sound channel coupling

As shown in Figure 8, establishing audio sources is the multichannel audio information source with the individual main sound channel of C (C 〉=2), just has 5 main sound channels and a low stress to imitate sound channel such as 5.1 channel audios, at this moment has the low frequency residual signals of C sound channel.Because the sound signal of each sound channel has big correlativity, so the frequency domain residual error signal of each sound channel also has big correlativity, therefore any one sound channel not only can utilize the low frequency residual signals of its place sound channel to carry out high frequency residual error reconstruct, but also can utilize the low frequency residual signals of other sound channel to carry out high frequency residual error reconstruct; Can obtain more match selection like this, improve the high frequency residual error reconstruction quality.With the 1st sound channel is that example is given explanation, at first can adopt above-mentioned frequency band matching strategy (P1) and (P2), utilizes being in of place sound channel 0 to arrive f _DBetween low frequency residual signals R _LCoupling and reconstruct are in f _DWith f _ZBetween high frequency residual error.That in addition, can also utilize other sound channel is in 0 to f _DBetween low frequency residual signals R _LBe in f in coupling and reconstruct the 1st sound channel _DWith f _ZBetween high frequency residual error.The same with the 1st sound channel, the frequency domain residual error signal of all C sound channels is many can to adopt identical method to handle.

(P4) carry out the high frequency residual error coupling by the interframe expansion

Above-mentioned frequency band matching strategy (P1), (P2) and (P3) matching strategy of described high frequency residual error be the high frequency residual error of utilizing the low frequency residual signals reconstruct present frame of present frame.Because have very big correlativity between the consecutive frame sound signal usually, so the residual error of consecutive frame also has bigger correlativity.Therefore, the high frequency residual error of current audio frame not only can be reconstructed with the low frequency residual signals of present frame, also can be reconstructed with the reconstruct residual signals that comprises reconstruct low frequency residual signals and reconstructed high frequency residual error of some frames before the present frame.As shown in Figure 9, adopt low frequency residual signals that above-mentioned frequency band matching strategy (P1), (P2) and method (P3) utilize the j frame that the high frequency residual error of j frame is reconstructed, or adopt low frequency residual signals that above-mentioned frequency band matching strategy (P1), (P2) and method (P3) utilize the j-1 frame that the high frequency residual error of j frame is reconstructed, can also adopt reconstructed high frequency residual error that above-mentioned frequency band matching strategy (P1), (P2) and method (P3) utilize the j-1 frame that the high frequency residual error of j frame is reconstructed.In addition, these methods can also expand to the j-2 frame, the j-3 frame goes.

Above-mentioned various frequency band matching strategy all need be sought the optimum matching frequency band between low frequency residual signals and high frequency residual error, said here the best can be expressed with error or distortion minimum, also can represent with the correlativity maximum.The step of the high frequency residual error parameter of concrete calculating optimum coupling comprises:

(B1) calculate normalization low frequency residual error band signal

With normalization high frequency residual error band signal

If the frequency span of high frequency residual error of mating and low frequency residual signals is L frequency, R _L(p, l), l=0,1 ... L-1 is low frequency residual signals frequency band, R _H(q, l), l=0,1 ... L-1 is the high frequency residual error frequency band, and wherein p and q represent R respectively _L(p, l) and R _H(q, the l) reference position of place frequency band, p ∈ [0, f _D], q ∈ [f _D, f _Z], calculate

With

Wherein, Expression R _L(p, l), l=0,1 ... L-1 is by the maximal value R of himself absolute value _{L max}(p) carry out normalization low frequency residual signals that normalization obtains,

Expression R _H(p, l), l=0,1 ... L-1 is by the maximal value R of himself absolute value _{H max}(q) carry out the normalization high frequency residual error that normalization obtains.

(B2) calculate coupling distortion measure d (p, q) or coupling related function r (p, q), if carry out The matching analysis with the distortion minimum, then Pi Pei distortion measure is expressed as:

$d(p,q)=w(p,q)\·\underset{l=0}{\overset{l=L-1}{\mathrm{\Σ}}}w\left(\left|{\hat{R}}_H(q,l)\right|\right)\·{({\hat{R}}_L(p,l)-{\hat{R}}_H(q,l))}^2---\left(1\right)$

Wherein w (p q) is the frequency influence factor,

It is residual error amplitude factor of influence.If w (p, q)=1.0, $w\left(\left|{\hat{R}}_H(q,l)\right|\right)=1.0,$ Formula (14) can be reduced to so:

$d(p,q)=\underset{l=0}{\overset{l=L-1}{\mathrm{\Σ}}}{({\hat{R}}_L(p,l)-{\hat{R}}_H(q,l))}^2---\left(2\right)$

If carry out The matching analysis with the correlativity maximum, normalization low frequency residual signals

With the normalization high frequency residual error

Related function be expressed as:

$r(p,q)=w(p,q)\·\underset{l=0}{\overset{l=L-1}{\mathrm{\Σ}}}w\left(\left|{\hat{R}}_H(q,l)\right|\right)\·({\hat{R}}_L(p,l)\·{\hat{R}}_H(q,l))---\left(3\right)$

Equally, w (p, q) the expression frequency influence factor,

Expression residual error amplitude factor of influence.If w (p, q)=1.0, $w\left(\left|{\hat{R}}_H(q,l)\right|\right)=1.0,$ Formula (16) can be reduced to so:

$r(p,q)=\underset{l=0}{\overset{l=N-1}{\mathrm{\Σ}}}{\hat{R}}_L(p,l){\hat{R}}_H(q,l)---\left(4\right)$

(B3) determine best frequency band matched position p ₀And q ₀, this optimum matching is exactly distortion measure d (p, minimum value q) or related function r (p, the optimum value p of pairing p of maximal value q) and q ₀And q ₀, they have determined the frequency band that high frequency residual error and low frequency residual error are mated.

(B4) utilize following formula calculating optimum energy matching attribute:

$G_R(p_0,q_0)=\frac{R_{H\max }\left(q_0\right)}{R_{\mathrm{Lp}\max }\left(p_0\right)}---\left(5\right)$

Like this, the resulting frequency band division result of step 304, optimum matching frequency band position, frequency-domain residual height frequency range boundary frequency f _D, energy matching attribute, sound channel coupling parameter and interframe spreading parameter, be exactly original high frequency residual error parameter.

Step 305, high frequency residual error parameter coding.

The original high frequency residual error parameter that step 304 produces is encoded, obtain the output of high frequency residual error coding.The coding of original high frequency residual error parameter adopts the various genuine coding methods of losing, and as linear or non-linear scalar quantization coding, vector quantization coding, perhaps adopts various undistorted coding methods simultaneously, as Huffman coding and arithmetic coding.

Step 204, frequency-domain residual decoding and reconstruct.

In decoding end, low frequency residual signals coding that receives and high frequency residual error parameter coding obtain the reconstructed frequency domain residual signals, as shown in Figure 4 through frequency-domain residual decoding and reconstruction processing.Concrete steps comprise:

Step 401 is decoded to the low frequency residual signals coding that receives, and obtains the decoded low frequency residual signals.

Step 402 is decoded to the high frequency residual error parameter coding that receives, and the high frequency residual error parameter that obtains decoding, this decoding high frequency residual error parameter comprise frequency band division result, frequency-domain residual height frequency range boundary frequency f _D, optimum matching frequency band position, energy matching attribute, sound channel coupling parameter and interframe spreading parameter.

Step 403, high frequency residual error reconstruct.

According to decoding high frequency residual error parameter, utilize the decoded low frequency residual signals to duplicate and reconstructed high frequency residual error, concrete steps comprise:

(D1) the decoded low frequency residual signals that obtains according to step 401 and 402, optimum matching frequency band position and energy matching attribute, carry out high frequency residual error with following formula and duplicate with energy flux matched:

R _H(q ₀，l)＝G _R(p ₀，q ₀)·R _L(p ₀，l)，l＝0，1，...L-1 (6)

Obtain reconstructed high frequency residual error frequency band R _H(q ₀, l);

(D2) the frequency band division result who obtains according to step 401 and 402, sound channel coupling parameter and interframe spreading parameter make up all reconstructed high frequency residual error frequency bands, obtain reconstructed high frequency residual error.

Step 404, the frequency-domain residual that obtains according to decoding height frequency range boundary frequency f _D, decoded low frequency residual signals and reconstructed high frequency residual error are made up, obtain the reconstructed frequency domain residual signals.

Step 205, the reconstructed frequency domain residual signals is handled through the time-frequency inverse transformation, obtains reconstruct time domain residual signals; Corresponding with time-frequency conversion, the time domain inverse transformation adopts inverse modified discrete cosine transform, the overlapping inverse transformation of inverse modified or inverse FFT method.

Step 206 is carried out time domain perception liftering to reconstruct time domain residual signals and is handled, and obtains the output of audio reconstruction time-domain signal; If use H _R(z) expression time domain perception inverse filter, then H _R(f) can be expressed as $H_R\left(z\right)=\frac{1}{H_M\left(z\right)}.$

Audio coding method that the above embodiment of the present invention provides and coding/decoding method can carry out efficient high-quality compressed encoding to the sound signal that comprises voice signal, improve the audio frequency transfer efficiency.

Above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (10)

1. the audio encoding and decoding method based on frequency domain filtering and the reconstruct of high frequency residual error comprises the steps:

1) at coding side, the original time-domain signal of audio frequency is handled through time-frequency conversion, obtains original frequency-region signal;

2) original frequency-region signal obtains original frequency domain residual error signal through frequency domain perception Filtering Processing;

3) original frequency domain residual error signal obtains low frequency residual signals coding and high frequency residual error parameter coding, and outputs to transmission channel or storage medium through frequency-domain residual analysis and encoding process.

4) in decoding end, receive from the low frequency residual signals coding and the high frequency residual error parameter coding that output to transmission channel or storage medium, and it is carried out frequency-domain residual decoding and reconstruction processing, obtain the reconstructed frequency domain residual signals;

5) the reconstructed frequency domain residual signals is handled through frequency domain perception liftering, obtains the reconstructed frequency domain signal;

6) the reconstruct frequency-region signal is carried out the time-frequency inverse transformation and handle, obtain the audio reconstruction time-domain signal.

2. encode/decode audio signal method according to claim 1, wherein described frequency-domain residual analysis of step 3) and encoding process comprise the steps:

(A1) according to the frequency band division method of equiband frequency band or critical band or sound interval frequency band, earlier original frequency domain residual error signal is carried out frequency band division, the code rate of setting according to audio coder is selected a frequency band division end points then, original frequency domain residual error signal is divided into original low frequency residual signals and original high frequency residual error two parts, makes these two parts respectively have several frequency bands;

(A2) original low frequency residual signals is encoded, obtain the output of low frequency residual signals coding; Again the low frequency residual signals is coded in coding side and carries out local decode, obtain the decoded low frequency residual signals;

(A3) according to the similarity or the correlativity of decoded low frequency residual signals and original high frequency residual error, high frequency residual error parameter is analyzed, promptly select a frequency band matching strategy that decoded low frequency residual signals and original high frequency residual error are carried out the frequency band coupling, and the high frequency residual error parameter of calculating optimum coupling, obtain comprising the just original high frequency residual error parameter of frequency range boundary frequency, optimum matching frequency band position, energy matching attribute, sound channel coupling parameter and interframe spreading parameter of frequency band division result, frequency-domain residual;

(A4) original high frequency residual error parameter is encoded, obtain the output of high frequency residual error parameter coding.

3. encode/decode audio signal method according to claim 2, wherein the described frequency band matching strategy of step (A3) comprises: the high frequency residual error frequency band mates in low frequency residual signals zone, the high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof, carry out the high frequency residual error coupling and carry out the high frequency residual error coupling by the interframe expansion by sound channel coupling.

4. encode/decode audio signal method according to claim 3, wherein said high frequency residual error frequency band mates in low frequency residual signals zone, be to mate in low frequency residual signals zone, seek the optimum matching frequency band of error minimum or correlativity maximum with each high frequency residual error frequency band.

5. encode/decode audio signal method according to claim 3, wherein said high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof, be that the high frequency residual error frequency band that frequency is minimum mates in low frequency residual signals zone earlier, seek the optimum matching frequency band of error minimum or correlativity maximum, and with the minimum high frequency residual error frequency band of this optimum matching frequency band reconfiguration frequency; Again that the frequency of this reconstruct is minimum high frequency residual error frequency band adds original low frequency residual signals, forms a new low frequency residual signals of forming; Then based on the low frequency residual signals of this new composition, the high frequency residual error frequency band of higher frequency is mated, until the coupling of finishing all high frequency residual error frequency bands.

6. encode/decode audio signal method according to claim 3, the wherein said coupling by sound channel carried out the high frequency residual error coupling, be meant under the multi-channel audio signal situation, the high frequency residual error frequency band of any one sound channel adopts the high frequency residual error frequency band to mate in low frequency residual signals zone or the high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof, utilize the low frequency residual signals of place sound channel that the high frequency residual error of place sound channel is mated, and employing high frequency residual error frequency band mates in low frequency residual signals zone or the high frequency residual error frequency band mates in low frequency residual signals zone and extended area thereof, utilizes the low frequency residual signals of other sound channel that the high frequency residual error of place sound channel is mated.

7. encode/decode audio signal method according to claim 3, the wherein said expansion by interframe carried out the high frequency residual error coupling, the high frequency residual error frequency band that is meant present frame adopts the high frequency residual error frequency band to mate in low frequency residual signals zone, the high frequency residual error frequency band is in low frequency residual signals zone and extended area mates or carry out the high frequency residual error coupling by the sound channel coupling, utilize the low frequency residual signals of present frame that the high frequency residual error of present frame is mated, and adopt the high frequency residual error frequency band to mate in low frequency residual signals zone, the high frequency residual error frequency band is in low frequency residual signals zone and extended area mates or carry out the high frequency residual error coupling by the sound channel coupling, utilize the reconstructed frequency domain residual signals of former frame or preceding some frames, comprise reconstruct low frequency residual signals and reconstructed high frequency residual error, the high frequency residual error of present frame is mated.

8. encode/decode audio signal method according to claim 2, wherein the high frequency residual error parameter of the described calculating optimum coupling of step (A3) comprises the steps:

(B1) p ∈ [0, f _D], q ∈ [f _D, f _Z] scope, calculate normalization low frequency residual error band signal

With normalization high frequency residual error band signal

Wherein, p and q represent respectively

With

The reference position of place frequency band, f _DBe low frequency residual signals and high frequency residual error boundary frequency, l is a frequency pointer in the frequency band.

(B2) p ∈ [0, f _D], q ∈ [f _D, f _Z] scope, calculate coupling distortion measure d (p, q) or coupling related function r (p, q);

(B3) p ∈ [0, f _D], q ∈ [f _D, f _Z] scope, with distortion measure d (p, minimum value q) or related function r (p, the value p of pairing p of maximal value q) and q ₀And q ₀Be best frequency band matched position;

(B4) calculating optimum energy matching attribute G _R(p ₀, q ₀).

9. encode/decode audio signal method according to claim 1, wherein described frequency-domain residual decoding of step 4) and reconstruction processing comprise the steps:

(C1) the low frequency residual signals coding that receives is decoded, obtain the decoded low frequency residual signals;

(C2) the high frequency residual error parameter coding that receives is decoded, high frequency residual error parameter obtains decoding;

(C3) utilize optimum matching frequency band position and energy matching attribute in decoded low frequency residual signals, the decoding high frequency residual error parameter, reconstructed high frequency residual error frequency band R _H(q ₀, l);

(C4) frequency band division method, sound channel coupling parameter and the interframe spreading parameter in the utilization decoding high frequency residual error parameter makes up the reconstructed high frequency residual error frequency band that obtains, and obtains reconstructed high frequency residual error;

(C5) the frequency-domain residual height frequency range boundary frequency in the utilization decoding high frequency residual error parameter makes up decoded low frequency residual signals and reconstructed high frequency residual error, obtains the reconstructed frequency domain residual signals.

10. the audio encoding and decoding method based on time-domain filtering and the reconstruct of high frequency residual error comprises the steps:

T1) at coding side, the original time-domain signal of audio frequency obtains original time domain residual signals through time domain perception Filtering Processing;

T2) original time domain residual signals is handled through time-frequency conversion, obtains original frequency domain residual error signal;

T3) original frequency domain residual error signal obtains low frequency residual signals coding and high frequency residual error parameter coding, and outputs to transmission channel or storage medium through frequency-domain residual analysis and encoding process.

T4) in decoding end, receive from the low frequency residual signals coding and the high frequency residual error parameter coding that output to transmission channel or storage medium, and it is carried out frequency-domain residual decoding and reconstruction processing, obtain the reconstructed frequency domain residual signals;

T5) the reconstruct frequency domain residual error signal is carried out the time-frequency inversion process, obtain reconstruct time domain residual signals;

T6) reconstruct time domain residual signals is carried out the time domain liftering and handle, obtain the audio reconstruction time-domain signal.

Images