CN101371297A - Apparatus and method for encoding and decoding signal - Google Patents

Abstract

Encoding and decoding apparatuses and encoding and decoding methods are provided. The decoding method includes extracting a plurality of encoded signals from an input bitstream, determining which of a plurality of decoding methods is to be used to decode each of the encoded signals, decoding the encoded signals using the determined decoding methods, and synthesizing the decoded signals. Accordingly, it is possible to encode signals having different characteristics at an optimum bitrate by classifying the signals into one or more classes according to the characteristics of the signals and encoding each of the signals using an encoding unit that can best serve the class where a corresponding signal belongs. In addition, it is possible to efficiently encode various signals including audio and speech signals.

Description

The equipment and the method that are used for the Code And Decode signal

Technical field

The present invention relates to Code And Decode equipment and Code And Decode method, and more specifically, relating to can be according to the characteristic of signal Code And Decode equipment and the Code And Decode method with best bit rate coding or decoded signal.

Background technology

Conventional audio coder can provide high-quality audio signal with 48kbps or bigger high bit rate, but is inefficient for processes voice signals.On the other hand, conventional sound decorder can be with 12kbps or littler low bit rate encoding speech signal efficiently, but is inefficient for the various sound signals of coding.

Summary of the invention

Technical matters

The invention provides and to have the Code And Decode equipment and the Code And Decode method of the signal (for example, voice and sound signal) of different qualities with best bit rate coding or decoding.

Technical scheme

According to an aspect of the present invention, provide a kind of coding/decoding method, having comprised: from incoming bit stream, extracted a plurality of coded signals; Determine that in a plurality of coding/decoding methods which will be used to each coded signal of decoding; Use determined coding/decoding method decoding and coding signal; Decoded signal is synthesized individual signals; With passing through individual signals is carried out post-processing operation recovery original signal.

According to a further aspect in the invention, provide a kind of decoding device, having comprised: the position parse module, it extracts a plurality of coded signals from incoming bit stream; The demoder determination module, it determines that in a plurality of decoding units which will be used to each coded signal of decoding; Decoder module, it comprises decoding unit, and uses determined decoding unit decodes coded signal; Synthesis module, its synthetic decoded signal; And post-processing module, it recovers original signal by individual signals is carried out post-processing operation.

According to a further aspect in the invention, provide a kind of coding method, having comprised: input signal is carried out pretreatment operation, and making that input signal can be exchanged into can be by encoded signals efficiently; With pretreated signal segmentation is a plurality of splitting signals; Based on each characteristic of splitting signal, determine that in a plurality of coding methods which will be used to each splitting signal of encoding; Use determined coding method code division signal; With produce bit stream based on the splitting signal of having encoded.

According to a further aspect in the invention, provide a kind of encoding device, having comprised: pretreatment module, it carries out pretreatment operation to input signal, and making that input signal can be exchanged into can be by encoded signals efficiently; The signal segmentation module, it is a plurality of splitting signals with pretreated signal segmentation; The scrambler determination module, it is based on each characteristic of splitting signal, determines that in a plurality of coding units which will be used to each splitting signal of encoding; Coding module, it comprises coding unit, and uses determined coding unit code division signal; With position package module, it produces bit stream based on the splitting signal of having encoded.

Advantageous effects

Therefore, by according to the characteristic of signal signal being categorized as one or more classifications and using the coding unit that can best satisfy the classification under the corresponding signal each signal of encoding, can have the signal of different qualities with best bit rate coding.In addition, can encode efficiently and comprise the various signals of audio frequency and voice signal.

Description of drawings

Fig. 1 is the block diagram of encoding device according to an embodiment of the invention;

Fig. 2 is the block diagram of the embodiment of the sort module shown in Fig. 1;

Fig. 3 is the block diagram of the embodiment of the pretreatment unit shown in Fig. 2;

Fig. 4 is the block diagram of equipment that is used for calculating the perceptual entropy of input signal according to an embodiment of the invention;

Fig. 5 is the block diagram of another embodiment of the sort module shown in Fig. 1;

Fig. 6 is the block diagram of the embodiment of the signal segmentation unit shown in Fig. 5;

Fig. 7 and 8 is the views that are used for illustrating the method that merges a plurality of splitting signals according to an embodiment of the invention;

Fig. 9 is the block diagram of another embodiment of the signal segmentation unit shown in Fig. 5;

Figure 10 is used for illustrating the view that according to an embodiment of the invention input signal is divided into the method for a plurality of splitting signals;

Figure 11 is the block diagram of the embodiment of order unit really shown in Fig. 5;

Figure 12 is the block diagram of the embodiment of the coding unit shown in Fig. 1;

Figure 13 is the block diagram of another embodiment of the coding unit shown in Fig. 1;

Figure 14 is the block diagram of encoding device according to another embodiment of the present invention;

Figure 15 is the block diagram of decoding device according to an embodiment of the invention; And

Figure 16 is the block diagram of the embodiment of the synthesis unit shown in Figure 15.

Embodiment

The present invention is described below with reference to the accompanying drawings more fully, exemplary embodiment of the present invention shown in the drawings.

Fig. 1 is the block diagram of encoding device according to an embodiment of the invention.With reference to figure 1, encoding device comprises sort module 100, coding module 200 and position package module 300.

Coding module 200 comprises a plurality of coding units of first coding unit 210 of execution different coding method to m coding unit 220.

Sort module 100 is divided into a plurality of splitting signals with input signal, and each of splitting signal is matched first coding unit 210 to the m coding unit 220.First coding unit 210 some to the m coding unit 220 can be mated two or more splitting signals or do not mated splitting signal.

Sort module 100 can divide the coordination amount with each splitting signal or the definite order of wanting the code division signal of encoding.

Use the coding module 200 of any one encode each splitting signal of first coding unit 210 to the m coding unit 220 to be matched corresponding splitting signal.Sort module 100 is analyzed the characteristic of each splitting signal, and select first coding unit 210 can be the most efficiently to the m coding unit 220 according to encode of each splitting signal of analysis result.

The coding unit of code division signal can be believed to realize the highest compression efficiency the most efficiently.

For example, the splitting signal that can easily be modeled as coefficient and surplus can be encoded efficiently by sound decorder, and the splitting signal that can not easily be modeled as coefficient and surplus can be encoded efficiently by audio coder.

If less than predefined threshold value, then splitting signal can be thought the signal of modeling easily to the energy of the surplus that obtains by the modeling splitting signal to the ratio of the energy of splitting signal.

Because the splitting signal that presents highly redundant on time shaft can use wherein linear prediction method based on previous signal estimation current demand signal by modeling well, therefore, use the sound decorder of linear prediction interpretation method this splitting signal of can encoding the most efficiently.

Position package module 300 produces the bit stream that will be transmitted based on the splitting signal of having encoded that is provided by coding module 200 with about the additional coding information of the splitting signal of having encoded.Position package module 300 can be used unformatted (bit-plain) method in position or bit slice (bit sliced) arithmetic coding method and produce the bit stream with variable bit rate.

Owing to having the splitting signal of coding or bandwidth, the bit rate restriction can not recover from the decoded signal or the bandwidth that provide by the demoder that uses interpolation, extrapolation or clone method.And, can be included in the bit stream that will be transmitted about the compensated information of the splitting signal that is not encoded.

With reference to figure 1, sort module 110 can comprise a plurality of taxons of first taxon 110 to n taxon 120.First taxon 110 each to the n taxon 120 input signal can be divided into a plurality of splitting signals, switching signal the territory, extract input signal characteristic, input signal is classified or input signal is matched first coding unit 210 to the m coding unit 220 according to the characteristic of input signal.

First taxon 110 to the n taxon 120 can be a pretreatment unit, this pretreatment unit to input signal carry out pretreatment operation make that input signal can be converted into can be by encoded signals efficiently.Pretreatment unit can be divided into input signal a plurality of components, for example coefficient component and component of signal, and can before other taxon is carried out their operation, carry out pretreatment operation to input signal.

Can be according to characteristic, external environmental factor and the target bit rate of input signal, pre-service input signal optionally, and pre-service some from a plurality of splitting signals that input signal obtains optionally only.

Sort module 100 can be according to the apperceive characteristic information of the input signal that is provided by psychologic acoustics MBM 400 and input signal is classified.The example of apperceive characteristic information comprises masking threshold, signal to noise ratio (S/N ratio) (SMR) and perceptual entropy.

In other words, apperceive characteristic information according to input signal, the for example masking threshold of input signal and SNR, sort module 100 can be divided into input signal a plurality of splitting signals maybe can match each splitting signal first coding unit 210 one or more to m coding unit 220.

In addition, sort module 100 can receive tone, zero crossing rate (ZCR) and the information of linear predictor coefficient and the classified information of previous frame such as input signal, and can classify to input signal according to the information that is received.

With reference to figure 1, the object information of being exported by coding module 200 of having encoded can be fed back to sort module 100.

In case input signal be classified module 100 be divided into a plurality of splitting signals and determined will by first coding unit 210 to the m coding unit 220 which, use what amount, what order to come the code division signal with, just come the code division signal according to determined result.In fact the position amount of each splitting signal of being used to encode can be same as the position amount of being distributed by sort module 100.

The information of the difference between the position amount that describes the actual position amount of using in detail and distributed can be fed back to sort module 100, and making sort module 100 can be other splitting signal increases the position amount of being distributed.If the actual bit amount is greater than the position amount of being distributed, then sort module 100 can be the position amount that other splitting signal reduces to be distributed.

The coding unit of actual coding splitting signal can be same as the coding unit that is matched splitting signal by sort module 100.In this case, signal can be fed back to sort module 100, and the coding unit of indication actual coding splitting signal is different from the coding unit that is matched splitting signal by sort module 100.Then, sort module 100 can match splitting signal the coding unit except the coding unit that before matched splitting signal.

Sort module 100 can be divided into input signal a plurality of splitting signals once more according to the object information of having encoded that feeds back to it.In this case, sort module 100 can obtain to have a plurality of splitting signals of the structure different with the structure of the previous splitting signal that obtains.

If sort module 100 selected encoding operations are different from the encoding operation of actual execution, then the information about the difference between them can feed back to sort module 100, makes sort module 100 can determine the information that encoding operation is relevant once more fully.

Fig. 2 is the block diagram of the embodiment of the sort module 100 shown in Fig. 1.With reference to figure 2, the first taxons can be pretreatment unit, and this pretreatment unit is carried out pretreatment operation to input signal, makes input signal to be encoded efficiently.

Can comprise first pretreater 111 of carrying out different preprocess methods a plurality of pretreaters to n pretreater 112 with reference to figure 2, the first taxons 110.First taxon 110 can use first pretreater 111 to the n pretreater 112 to come input signal is carried out pre-service with characteristic, external environmental factor and target bit rate according to input signal.And first taxon 110 can use first pretreater 111 to 112 pairs of input signals of n pretreater to carry out two or more pretreatment operation.

Fig. 3 is the block diagram of first pretreater 111 shown in Fig. 2 to the embodiment of n pretreater 112.With reference to figure 3, pretreater comprises coefficient extraction apparatus 113 and surplus extraction apparatus 114.

Coefficient extraction apparatus 113 is analyzed input signal and is extracted the coefficient of the characteristic of representing input signal from input signal.Surplus extraction apparatus 114 extracts from input signal and has used the coefficient that is extracted to remove the surplus of redundant component from it.

Pretreater can be carried out the linear prediction decoded operation to input signal.In this case, coefficient extraction apparatus 113 passes through input signal is carried out linear prediction analysis and extracted linear predictor coefficient from input signal, and the linear predictor coefficient that surplus extraction apparatus 114 coefficient of performance extraction apparatuss 113 provide extracts surplus from input signal.Removed redundant surplus from it and can have the form identical with white noise.

To describe linear prediction analysis method according to an embodiment of the invention in detail below.

Can form by the linear combination of previous input signal by the prediction signal that linear prediction analysis obtains, represented as equation (1):

Mathematic graph 1

$\hat{x}\left(n\right)=\underset{j=1}{\overset{p}{\mathrm{\Σ}}}{\mathrm{\α}}_{j}x(n-j)$

Wherein, p represents the linear prediction exponent number, ₁Arrive _pThe linear predictor coefficient that expression obtains by the mean square deviation (MSE) that minimizes between input signal and the estimated signal.

The transport function P (z) that is used for linear prediction analysis can be represented by equation (2):

Mathematic graph 2

$P\left(z\right)=\underset{k=1}{\overset{p}{\mathrm{\Σ}}}{\mathrm{\α}}_k{z}^{-k}$

With reference to figure 3, pretreater can use linear predictive interpretation (WLPC) method of curling to extract linear predictor coefficient and surplus from input signal, this linear prediction analysis that linear predictive interpretation method is another type of curling.Have about unit delay Z by replacement ^-1The all-pass filter of transport function A (z), can realize the WLPC method.Transport function A (z) can be represented by equation (3):

Mathematic graph 3

$A\left(z\right)=\left[\frac{z^{-1}-\mathrm{\λ}}{1-\mathrm{\λ}{z}^{-1}}\right]$

Wherein, λ represents the all-pass coefficient.By changing the all-pass coefficient, can change the resolution of the signal that will analyze.For example, if analyzed signal height concentrates on a certain frequency band, for example, if analyzed signal is the sound signal of high concentration in low-frequency band, then making the resolution of low band signal to be increased and can encode efficiently by setting all-pass coefficient will analyzed signal.

In the WLPC method, compare with high-frequency signal, with higher resolution analysis low frequency signal.Like this, the WLPC method can realize high estimated performance and can carry out modeling to low frequency signal better for low frequency signal.

The all-pass coefficient can be according to characteristic, external environmental factor and the target bit rate of input signal and is changed along time shaft.If the all-pass coefficient changes in time, the then significantly distortion of sound signal that obtains by decoding.Like this, when the all-pass index variation, the smooth method may be used on the all-pass coefficient makes the all-pass coefficient little by little to change, and makes distorted signals to minimize.The scope that can be defined as the value of current all-pass coefficient value can be determined by previous all-pass coefficient value.

Replace original signal, masking threshold can be used as the input that is used for estimating linear predictor coefficient.More specifically, masking threshold can be exchanged into time-domain signal, and can use time-domain signal to carry out WLPC as input.Also can use surplus to carry out the prediction of linear predictor coefficient as input.In other words, linear prediction analysis can be carried out and surpass once, therefore obtains the surplus of further albefaction.

Can comprise with reference to figure 2, the first taxons 110: first pretreater 111, it carries out top linear prediction analysis with reference to equation (1) and (2) description; With the second pretreater (not shown), it carries out WLPC.First taxon 100 can be selected in the first processor 111 and second pretreater, perhaps can input signal not carried out linear prediction analysis according to characteristic, external environmental factor and the decision of target bit rate of input signal.

If the value of all-pass coefficient is 0, then second pretreater can be identical with first pretreater 111.In this case, first taxon 110 can only comprise second pretreater, and selects in linear prediction analysis method and the WLPC method one according to the value of all-pass coefficient.And first taxon 110 can be carried out linear prediction analysis, and any method in perhaps linear prediction analysis method and the WLPC method is selected in frame unit.

Indicate whether to carry out the information of linear prediction analysis and which the selecteed information in linear prediction analysis method and the WLPC side of indicating can be included in the bit stream that will be transmitted.

Position package module 300 receives linear predictor coefficient, indicates whether to carry out the information of linear prediction decoding and the information of the actual Linear Predictive Coder that uses of identification from first taxon 110.Then, position package module 300 is inserted the bit stream that will be transmitted with the information of all receptions.

Being used for that input signal is encoded to the required position amount of signal that the tonequality of tonequality and original input signal almost can not be distinguished can be determined by the perceptual entropy of calculating input signal.

Fig. 4 is the block diagram that is used for calculating the equipment of perceptual entropy according to an embodiment of the invention.With reference to figure 4, this equipment comprises bank of filters 115, linear prediction unit 116, psychologic acoustics modeling unit 117, first computing unit 118 and second computing unit 119.

The perceptual entropy PE of input signal can use equation (4) to calculate:

Mathematic graph 4

$\mathrm{PE}=\frac{1}{2\mathrm{\π}}\underset{0}{\overset{\mathrm{\π}}{\∫}}\max [0,{\log }_2\frac{X\left(e^{\mathrm{jw}}\right)}{T\left(e^{\mathrm{jw}}\right)}]\mathrm{dw}(\mathrm{bit}/\mathrm{sample})$

Wherein, X (e ^Jw) expression original input signal energy level, T (e ^Jw) the expression masking threshold.

In relating to the WLPC method of using all-pass filter, can use the ratio of the masking threshold of the energy of surplus of input signal and surplus to calculate the perceptual entropy of input signal.More specifically, use the encoding device of WLPC method can use equation (5) to calculate the perceptual entropy PE of input signal:

Mathematic graph 5

$\mathrm{PE}=\frac{1}{2\mathrm{\π}}\underset{0}{\overset{\mathrm{\π}}{\∫}}\max [0,{\log }_2\frac{R\left(e^{\mathrm{jw}}\right)}{T\′\left(e^{\mathrm{jw}}\right)}]\mathrm{dw}(\mathrm{bit}/\mathrm{sample})$

Wherein, R (e ^Jw) energy of surplus of expression input signal, T ' (e ^Jw) expression surplus masking threshold.

Masking threshold T ' (e ^Jw) can represent by equation (6):

Mathematic graph 6

T′(e ^jw)＝T(e ^jw)/|H(e ^jw)| ²

Wherein, T (e ^Jw) expression original signal masking threshold, H (e ^Jw) expression is used for the transport function of WLPC.Psychologic acoustics modeling unit 320 is used transfer function H (e ^Jw) and scale factor band territory in masking threshold T (e ^Jw) can calculate masking threshold T ' (e ^Jw).

Receive surplus that obtains by linear prediction unit 116 performed WLPC and the masking threshold of exporting by psychologic acoustics modeling unit 117 with reference to 4, the first computing units of figure 118.Bank of filters 116 can be carried out frequency inverted to original signal, and the result of frequency inverted can be input to psychologic acoustics modeling unit 117 and second computing unit 119.Bank of filters 115 can be carried out Fourier transform to original signal.

First computing unit 118 can use the ratio of the energy of the masking threshold of the original signal of being removed by the spectrum of the transport function of WLPC composite filter and surplus to calculate perceptual entropy.

The curling perceptual entropy WPE that is divided into the signal of 60 or more a plurality of inhomogeneous dividing strips with different bandwidth can use WLPC to be calculated, shown in equation (7):

Mathematic graph 7

$\mathrm{WPE}=-\underset{b=1}{\overset{b\max }{\mathrm{\Σ}}}(w_{\mathrm{high}}\left(b\right)-w_{\mathrm{low}}\left(b\right))\·{\log }_{10}\left(\frac{n{b}_{\mathrm{res}}\left(b\right)}{e_{\mathrm{res}}\left(b\right)}\right)$

$e_{\mathrm{res}}\left(b\right)=\underset{w=w_{\mathrm{low}}\left(b\right)}{\overset{w_{\mathrm{high}}\left(b\right)}{\mathrm{\Σ}}}\mathrm{res}{\left(w\right)}^2$ ${\mathrm{nb}}_{\mathrm{res}}\left(b\right)=\underset{w=w_{\mathrm{low}}\left(b\right)}{\overset{w_{\mathrm{high}}\left(b\right)}{\mathrm{\Σ}}}\frac{{\mathrm{nb}}_{\mathrm{linear}}\left(w\right)}{h{\left(w\right)}^2}$

Wherein, b represents the index of the dividing strip that the applied mental acoustic model obtains, e _Res(b) energy of the surplus among the expression dividing strip b and, w_low (b) and w_high (b) represent the minimum and highest frequency among the dividing strip b, nb respectively _Linear(w) masking threshold of the dividing strip of expression linear mapping, h (w) ²Linear prediction decoding (LPC) energy spectrum of expression frame, nb _Res(w) expression is corresponding to the linear masking threshold of surplus.

On the other hand, be divided into the curling perceptual entropy WPE of the signal of 60 or more a plurality of even dividing strips with same band again _SubCan use WLPC to be calculated, shown in equation (8):

Mathematic graph 8

${\mathrm{nb}}_{\mathrm{sub}}\left(s\right)=\underset{s_{\mathrm{low}}\left(s\right)<w<s_{\mathrm{high}}\left(s\right)}{\min }\left(\frac{{\mathrm{nb}}_{\mathrm{linear}}\left(w\right)}{h{\left(w\right)}^2}\right)$

${\mathrm{WPE}}_{\mathrm{sub}}=-\underset{s=1}{\overset{s\max }{\mathrm{\&Sigma;}}}(s_{\mathrm{high}}\left(s\right)-s_{\mathrm{low}}\left(s\right))\&CenterDot;{\log }_{10}\left(\frac{{\mathrm{nb}}_{\mathrm{sub}}\left(s\right)}{e_{\mathrm{sub}}\left(s\right)}\right)$

$e_{\mathrm{sub}}\left(s\right)=\underset{w=s_{\mathrm{low}}\left(s\right)}{\overset{s_{\mathrm{high}}\left(s\right)}{\mathrm{\&Sigma;}}}\mathrm{res}{\left(w\right)}^2$

Wherein, s represents the index of the linear subband of separating, s _Low(w) and s _High(w) the minimum and highest frequency among the subband s of the separation of expression linearity respectively, nb _Sub(s) masking threshold of the linear subband s that separates of expression, e _Sub(s) energy of the linear subband s that separates of expression, the frequency among the promptly linear subband s that separates and.Masking threshold nb _Sub(s) be the minimum value of a plurality of masking thresholds among the linear subband s that separates.

For have same band and have be higher than input spectrum and the band of threshold value, can not calculate perceptual entropy.Like this, the curling perceptual entropy WPE of equation (8) _SubCan be lower than the curling perceptual entropy WPE that high-resolution equation (7) is provided for low-frequency band.

Use WLPC to calculate the perceptual entropy WPE that curls for scale factor band with different bandwidth _Sf, represented as equation (9):

Mathematic graph 9

${\mathrm{nb}}_{\mathrm{sf}}\left(s\right)=\underset{{\mathrm{sf}}_{\mathrm{low}}\left(s\right)<w<{\mathrm{sf}}_{\mathrm{high}}\left(s\right)}{\min }\left(\frac{{\mathrm{nb}}_{\mathrm{linear}}\left(w\right)}{h{\left(w\right)}^2}\right)$

${\mathrm{WPE}}_{\mathrm{sf}}=-\underset{f=1}{\overset{f\max }{\mathrm{\&Sigma;}}}(s_{\mathrm{high}}\left(f\right)-s_{\mathrm{low}}\left(f\right))\&CenterDot;{\log }_{10}\left(\frac{{\mathrm{nb}}_{\mathrm{sf}}\left(f\right)}{e_{\mathrm{sf}}\left(f\right)}\right)$

$e_{\mathrm{sf}}\left(s\right)=\underset{w={\mathrm{sf}}_{\mathrm{low}}\left(s\right)}{\overset{{\mathrm{sf}}_{\mathrm{high}}\left(s\right)}{\mathrm{\&Sigma;}}}\mathrm{res}{\left(w\right)}^2$

Wherein, f represents the index of scale factor band, nb _Sf(f) the minimum masking threshold of expression scale factor band f, WPE _SfThe ratio of the input signal of expression scale factor band f and the masking threshold of scale factor band f, e _Sf(s) all frequencies among the expression scale factor band f and, i.e. the energy of scale factor band f.

Fig. 5 is the block diagram of another embodiment of the sort module 100 shown in Fig. 1.With reference to figure 5, sort module comprises signal segmentation unit 121 and determining unit 122.

More specifically, signal segmentation unit 121 is divided into a plurality of splitting signals with input signal.For example, signal segmentation unit 121 can use sub-filter that input signal is divided into a plurality of frequency bands.Frequency band can have identical bandwidth or different bandwidth.As mentioned above, by can best satisfying the coding unit of the characteristic of splitting signal, splitting signal can be encoded discretely with other splitting signal.

Signal segmentation unit 121 can be divided into input signal a plurality of splitting signals, and for example, a plurality of band signals make that the interference between the band signal can minimize.Signal segmentation unit 121 can have double filter group structure.In this case, also divisible each splitting signal in signal segmentation unit 121.

About the carve information of the splitting signal that obtained by signal segmentation unit 121, for example the breath of taking a message of the total number of splitting signal and each splitting signal can be included in the bit stream that will be transmitted.The decoding device synthetic decoded signal of splitting signal and reference segmentation information of decoding separably recovers original input signal thus.

Carve information can be stored as form.Bit stream can comprise the identification information of the form that is used to cut apart original input signal.

Can determine the importance of each splitting signal (for example, a plurality of band signals), and can regulate bit rate for each splitting signal according to determined result to tonequality.More specifically, the importance of the splitting signal on-fixed value that may be defined as fixed value or change according to the characteristic of the input signal of each frame.

If voice and sound signal are mixed into input signal, then signal segmentation unit 121 can be divided into voice signal and sound signal with input signal according to the characteristic of voice signal and the characteristic of sound signal.

Determining unit 122 can be determined to the m coding unit 220 which of first coding unit 210 in the coding module 200 each splitting signal of can encoding the most efficiently.

Determining unit 122 is categorized as some groups with splitting signal.For example, determining unit 122 can be categorized as splitting signal N classification, and, determine that first coding unit 210 which to the m coding unit 220 will be used to each splitting signal of encoding by each classification in N the classification being matched first coding unit 210 to the m coding unit 220.

More specifically, suppose that coding module 200 comprises that first coding unit 210 is to m coding unit 220, then determining unit 122 can be categorized as first to the m classification with splitting signal, and this first can be encoded by first coding unit 210 to m coding unit 220 respectively the most efficiently to the m classification.

For this reason, can by first coding unit 210 each to the m coding unit 220 the most efficiently the characteristic of encoded signals can be determined in advance, and can limit first the characteristic to the m classification according to this result who determines.After this, determining unit 122 can be extracted the characteristic of each splitting signal, and each splitting signal is categorized as first the classification to the m classification of sharing identical characteristics with corresponding splitting signal according to the result who is extracted.

First example to the m classification comprises voiced speech classification, unvoiced speech classification, background noise classification, noiseless classification, tone audio categories, non-pitch audio categories and voiced speech/audio mix classification.

By with reference to the apperceive characteristic information that provides by psychologic acoustics MBM 400 about splitting signal, the for example masking threshold of splitting signal, SMR or perceptual entropy level, determining unit 122 can determine that first coding unit 210 which to the m coding unit 220 will be used to each splitting signal of encoding.

By the apperceive characteristic information of reference about splitting signal, determining unit 122 can be determined the position amount so that encode each splitting signal, perhaps definite order of wanting the code division signal.

Can comprise definite information that is obtained of carrying out by determining unit 122 in the bit stream that is transmitted, for example, the information of the order of the information of the position amount that has by first coding unit 210 to which and each splitting signal that will be encoded of m coding unit 220 of indication and indication code division signal.

Fig. 6 is the block diagram of the embodiment of the signal segmentation unit 121 shown in Fig. 5.With reference to figure 6, the signal segmentation unit comprises dispenser 123 and combiner 124.

Dispenser 123 can be divided into input signal a plurality of splitting signals.Combiner 124 can be merged into individual signals with the splitting signal with similar characteristics.For this reason, combiner 124 can comprise the composite filter group.

For example, dispenser 123 can be divided into input signal 256 bands.In 256 bands, these bands with similar characteristics can be merged into single band by combiner 124.

With reference to figure 7, combiner 124 can be merged into single combined signal with a plurality of splitting signals located adjacent one another.In this case, combiner 124 can be merged into single combined signal with a plurality of adjacent splitting signals according to predefined rule, and does not consider the characteristic of adjacent splitting signal.

Whether alternatively, with reference to figure 8, combiner 124 can be merged into single combined signal with a plurality of splitting signals with similar characteristics, and adjacent one another are irrelevant with splitting signal.In this case, combiner 124 can will be able to be merged into single combined signal by a plurality of splitting signals that identical coding unit is encoded efficiently.

Fig. 9 is the block diagram of another embodiment of the signal segmentation unit 121 shown in Fig. 5.With reference to figure 9, the signal segmentation unit comprises first dispenser 125, second dispenser 126 and the 3rd dispenser 127.

More specifically, signal segmentation unit 121 is cut apart input signal scalably.For example, input signal can be divided into two splitting signals by first dispenser 125, one in two splitting signals can be divided into three splitting signals by second dispenser 126, and in three splitting signals one can be divided into three splitting signals by the 3rd dispenser 127.By this way, input signal may be partitioned into 6 splitting signals altogether.Signal segmentation unit 121 is divided into input signal a plurality of bands with different bandwidth scalably.

In the embodiment show in figure 9, cut apart input signal according to 3 fraction levels, but the invention is not restricted to this.In other words, according to 2 grades or 4 grades or more multistage classification, input signal may be partitioned into a plurality of splitting signals.

One in first dispenser, 125 to the 3rd dispensers 127 in the signal segmentation unit 121 can be divided into input signal a plurality of time-domain signals.

Figure 10 illustrates that signal segmentation unit 121 is divided into input signal the embodiment of a plurality of splitting signals.

During the short frame length cycle, voice or sound signal be stable state normally.Yet sometimes, for example during the transient period, voice or sound signal can have the unstable state characteristic.

In order to analyze unstable signal efficiently and to improve the efficient of this unstable signal of coding, can use small echo or empirical mode decomposition (EMD) method according to the encoding device of present embodiment.In other words, the encoding device according to present embodiment can use unfixed transforming function transformation function to analyze the characteristic of input signal.For example, signal segmentation unit 121 can use unfixed frequency band sub-band filter method input signal to be divided into a plurality of bands with bandwidth varying.

To describe the method that input signal is divided into a plurality of splitting signals by EMD below in detail.

In the EMD method, input signal can be decomposed into one or more natural mode functions (IMF).IMF must satisfy following condition: extreme value number and zero crossing number must equate or differ one at the most; The mean value of envelope of being determined by local maximum and the envelope determined by local minimum is zero.

The IMF representative is similar to the simple oscillation pattern of the component in the simple harmonic function, therefore makes it can use the EMD method to decompose input signal efficiently.

More specifically, in order to extract IMF from input signal s (t), envelope can be produced by connecting all local extremums of using the cubic spline interpolating method to determine by the local maximum of input signal s (t), and envelope can be produced down by connecting all local extremums of using the cubic spline interpolating method to determine by the local minimum of input signal s (t).The all values that input signal s (t) can have can be between envelope and the following envelope.

After this, can calculate the mean value m (t) of envelope and following envelope.After this, can calculate the first component h by deducting mean value m (t) from input signal s (t) ₁(t), shown in equation (10):

Mathematic graph 10

s(t)-m ₁(t)＝h ₁(t)

If the first component h ₁(t) do not satisfy above-mentioned IMF condition, then the first component h ₁(t) can be confirmed as identically, and can carry out aforesaid operations once more up to an IMF C who obtains to satisfy above-mentioned IMF condition with input signal s (t) ₁(t) till.

In case obtain an IMF C ₁(t), just by deducting an IMF C ₁(t) obtain surplus r ₁(t), shown in equation (11):

Mathematic graph 11

s(t)-c ₁(t)＝r ₁(t)

After this, can use surplus r ₁(t) carry out above-mentioned IMF once more as new input signal and extract operation, thereby obtain the 2nd IMF C ₂(t) and surplus r ₂(t).

If extract the surplus r that operating period obtains at above-mentioned IMF _n(t) have constant value or monotone increasing function or an extreme value is only arranged or do not have the simply periodic function of extreme value at all, then can stop above-mentioned IMF and extract operation.

Extract the result of operation as above-mentioned IMF, input signal s (t) can be by a plurality of IMF C ₀(t) to C _M(t) and final surplus r _m(t) and the expression, shown in equation (12):

Mathematic graph 12

$s\left(t\right)=\underset{m=0}{\overset{M}{\mathrm{\&Sigma;}}}{C}_m\left(t\right)+r_m\left(t\right)$

Wherein, M represents the total number of the IMF that extracts.Final surplus r _m(t) but the total characteristic of reflected input signal s (t).

Figure 10 illustrates by using the EMD method to decompose 11 IMF and final surplus that original input signal obtains.With reference to Figure 10, the frequency of the IMF that obtains from original input signal in early days that extracts at IMF is higher than the frequency of the IMF that obtains from original input signal in late period that IMF extracts.

Use previous surplus h _{1 (k-1)}With current surplus h _1kBetween standard deviation S D can simplify IMF and extract, shown in equation (13):

Mathematic graph 13

$\mathrm{SD}=\underset{t=0}{\overset{T}{\mathrm{\&Sigma;}}}\left[\frac{{|h_{1(k-1)}\left(t\right)-h_{1k}\left(t\right)|}^2}{h_{1(k-1)}^2\left(t\right)}\right]$

If standard deviation S D is less than for example 0.3 reference value, then current surplus h _1kCan regard IMF as.

Simultaneously, signal x (t) can be transformed to analytic signal by Hilbert transform, shown in equation (14):

Mathematic graph 14

z(t)＝x(t)+jH{x(t)}＝a(t)e ^jθ(t)

Wherein, (t) expression instantaneous amplitude, (t) expression instantaneous phase, H{} represents Hilbert transform.

As the result of Hilbert transform, input signal can be exchanged into the analytic signal of being made up of real component and imaginary component.

By Hilbert transform being applied to mean value is 0 signal, and can obtain to provide high-resolution frequency component for time domain and frequency domain.

To describe shown in Fig. 4 order unit 122 really below in detail and how determine that in a plurality of coding units which will be used for encoding by decomposing each of a plurality of splitting signals that input signal obtains.

Which each splitting signal of can encoding more efficiently that determining unit 122 can be determined sound decorder and audio coder.In other words, what determining unit 122 can determine the splitting signal that uses first coding unit 210 can encode efficiently to any one sound decorder of m coding unit 220 is encoded is sound decorder, and decision encodes to the splitting signal that uses first coding unit 210 can encode efficiently to any one audio coder in the m coding unit 220 is audio coder.

Which the code division signal more efficiently that to describe below how determining unit 122 to determine sound decorder and audio coder in detail.

Determining unit 122 can be measured the variation in the splitting signal, and if the result who measures greater than predefined reference value, determine that then sound decorder can be than audio coder code division signal more efficiently.

Alternatively, determining unit 122 can be measured the tonal components in certain part that is included in splitting signal, and if the result who measures greater than predefined reference value, determine that then sound decorder can be than audio coder code division signal more efficiently.

Figure 11 is the block diagram of the embodiment of order unit 122 really shown in Fig. 5.With reference to Figure 11, determining unit comprises audio coding/decoding unit 500, first bank of filters 510, second bank of filters 520, determining unit 530 and psychologic acoustics modeling unit 540.

Really order unit which each splitting signal of can encoding more efficiently that can determine sound decorder and audio coder shown in Figure 11.

With reference to Figure 11, input signal is by audio coding/decoding unit 500 coding, and coded signal is by 500 decodings of audio coding/decoding unit, thereby recovers original input signal.Audio coding/decoding unit 500 can comprise AMR-WB (AMR-WB) sound decorder/demoder, and the AMR-WB speech coders/decoders can have code exciting lnear predict (CELP) structure.

Input signal can be owed sampling (down-sampled) before being input to audio coding/decoding unit 500.The signal of audio coding/decoding unit 500 outputs can be recovered input signal thus by over-sampling (up-sampled).

Input signal can carry out frequency transformation by first bank of filters 510.

Signal by 500 outputs of audio coding/decoding unit is converted to frequency-region signal by second bank of filters 520.First bank of filters 510 or second bank of filters 520 can be carried out cosine transform to the signal that is input to it, for example, revise discrete cosine transform (MDCT).

The frequency component of the input signal of the recovery of the frequency component of the original input signal of first bank of filters, 510 outputs and 520 outputs of second bank of filters all is imported into determining unit 530.Which coded input signal more efficiently that determining unit 530 can be determined sound decorder and audio coder based on the frequency component that is input to it.

More specifically, based on the frequency component that is input to determining unit, calculate the perceptual entropy PE of each frequency component by using equation (15) _i, which coded input signal more efficiently that determining unit 530 can be determined sound decorder and audio coder:

Mathematic graph 15

${\mathrm{PE}}_i=\underset{j=j_{\mathrm{low}}\left(i\right)}{\overset{j_{\mathrm{high}}\left(i\right)}{\mathrm{\&Sigma;}}}N\left(j\right)$

Wherein

$N\left(j\right)=\left\{\begin{array}{c}0,x\left(j\right)=0\\ {\log }_2(2\left|\mathrm{nint}\left(\frac{x\left(j\right)}{\mathrm{\&delta;}}\right)\right|+1),x\left(j\right)\&NotEqual;0\end{array}\right.$

Wherein, the coefficient of x (j) expression frequency component, j represents the index of frequency component, and i represents quantization step, and nint () is the function that immediate integer is turned back to its independent variable, j _{Low (i)}And j _{High (i)}Be respectively the beginning frequency index and the end frequency index of scale factor band.

Determining unit 530 can use equation (15) to calculate the perceptual entropy of frequency component of the input signal of the perceptual entropy of frequency component of original input signal and recovery, and determine for being used for coded input signal audio coder and sound decorder based on result calculated which more efficiently.

For example, if the perceptual entropy of the frequency component of original input signal less than the perceptual entropy of the frequency component of the input signal that recovers, then determining unit 530 can determine that audio coder can be than sound decorder coded input signal more efficiently.On the other hand, if the perceptual entropy of the frequency component of the input signal that recovers less than the perceptual entropy of the frequency component of original input signal, then determining unit 530 can determine that sound decorder can be than audio coder coded input signal more efficiently.

The block diagram of one the embodiment that Figure 12 is first coding unit 210 shown in Fig. 1 to the m coding unit 220.Coding unit shown in Figure 12 can be a sound decorder.

Usually, sound decorder can be carried out LPC to input signal in frame unit, and uses the Levinson-Durbin algorithm to extract LPC coefficient, for example 16 rank LPC coefficients from each frame of input signal.By adaptive codebook search or fixed codebook search, can quantize pumping signal.Use the linear prediction method of algebraic code excitation, can quantize pumping signal.Use has the quantification form of conjugated structure, can carry out vector quantization to the gain of pumping signal.

Sound decorder shown in Figure 12 comprises linear prediction analysis unit 600, pitch estimation unit 610, codebook search unit 620, line spectrum pair (LSP) unit 630 and quantifying unit 640.

Linear prediction analysis unit 600 uses coefficient of autocorrelation that input signal is carried out linear prediction analysis, and this coefficient of autocorrelation obtains by using asymmetric window.If leading (look-ahead) cycle, promptly asymmetric window has the length of 30ms, and then linear prediction analysis unit 600 can use the 5ms leading cycle to carry out linear prediction analysis.

Coefficient of autocorrelation uses the Levinson-Durbin algorithm to be converted to linear predictor coefficient.For quantizing and linear interpolation, LSP unit 630 is converted to LSP with linear predictor coefficient.Quantifying unit 640 quantizes LSP.

Pitch estimation unit 610 is estimated the open loop pitch, so that reduce the complicacy of adaptive codebook search.More specifically, pitch estimation unit 610 uses the voice signal territory of the weighting of each frame to estimate the open loop pitch cycle.After this, use estimated open loop pitch to construct the harmonic noise forming filter.After this, use harmonic noise forming filter, linear prediction synthesis filter and resonance peak perceptual weighting filter to calculate impulse response.Impulse response can be used for producing the echo signal that is used for quantizing pumping signal.

Adaptive codebook search and fixed codebook search are carried out in codebook search unit 620.Calculate adaptive codebook vector by the search of closed loop pitch and by the interpolation of pumping signal in the past, in subframe unit, can carry out adaptive codebook search.The adaptive codebook parameter can comprise the gain of pitch cycle and pitch wave filter.Can produce pumping signal by linear prediction synthesis filter, so that simplify the closed loop search.

The fixed codebook structure forms based on interweaving monopulse displacement (ISSP) design.The codebook vectors that will comprise 64 positions of locating 64 pulses respectively is divided into four tracks, and each track comprises 16 positions.According to transfer rate, the pulse of predetermined number can be positioned at each of four tracks.Because therefore code book index indication track position and impulse code do not need to store code book, and can only use this code book index to produce pumping signal.

Sound decorder shown in Figure 12 can be carried out above-mentioned decode procedure in time domain.And if use linear prediction interpretation method coded input signal by the sort module shown in Fig. 1 100, then linear prediction analysis unit 600 can be chosen wantonly.

The invention is not restricted to the sound decorder shown in Figure 12.In other words, can use within the scope of the invention except the sound decorder shown in Figure 12, the various sound decorders of encoding speech signal efficiently.

The block diagram of another embodiment of one that Figure 13 is first coding unit 210 shown in Fig. 1 to the m coding unit 220.Coding unit shown in Figure 13 can be an audio coder.

With reference to Figure 13, audio coder comprises bank of filters 700, psychologic acoustics modeling unit 710 and quantifying unit 720.

Bank of filters 700 is converted to frequency-region signal with input signal.Bank of filters 700 can be carried out cosine transform to input signal, for example, revises discrete cosine transform (MDCT).

Psychologic acoustics modeling unit 710 is calculated the masking threshold of input signal or the SMR of input signal.Quantifying unit 720 uses the masking threshold that is calculated by psychologic acoustics modeling unit 710 to quantize the MDCT coefficient of being exported by bank of filters 700.Alternatively, in order to minimize audible distortion in given bit rate scope, quantifying unit 720 can be used the SMR of input signal.

Audio coder shown in Figure 13 can be carried out above-mentioned cataloged procedure in frequency domain.

The invention is not restricted to the audio coder shown in Figure 13.In other words, can use within the scope of the invention except the audio coder shown in Figure 13, various audio coders of coding audio signal (for example, advanced audio code translator) efficiently.

The advanced audio code translator carry out transient noise be shaped (TNS), intensity/coupling, predict and in/the stereo decoding of side (M/S).TNS be in the bank of filters window suitably distribution time domain quantization noise make the quantization noise inaudible operation that can become.Intensity/coupling is a kind of operation, comes the energy of transmitting audio signal, this operation can reduce the amount of the spatial information that will transmit by coding audio signal and this fact of time scale of only depending primarily on energy based on the perception of audio direction in the high-band.

Prediction is to remove redundant operation by the correlativity between the spectral component that uses frame from the indeclinable signal of statistical property.The stereo decoding of M/S is the operation of standardized and (that is) and poor (that is side) of transmission stereophonic signal rather than a left side and right channel signal.

The signal that carries out TNS, intensity/coupling, prediction and the stereo decoding of M/S is quantized by quantizer, and this quantizer uses the SMR that obtains from psychoacoustic model to carry out synthesis analysis (AbS).

As mentioned above because audio coder uses the modeling method coded input signal such as the linear prediction interpretation method, so shown in Fig. 5 really order unit 122 can determine whether input signal can be by modeling easily according to one group of predetermined rule.After this, if determine that input signal can be by modeling easily, then determining unit 122 can determine to use the sound decorder coded input signal.On the other hand, if determine that input signal can not be by modeling easily, then determining unit 122 can determine to use the audio coder coded input signal.

Figure 14 is the block diagram of encoding device according to another embodiment of the present invention.In Fig. 1 to 14, the identical identical key element of Reference numeral representative, and therefore, will skip its detailed description.

With reference to Figure 14, sort module 100 is divided into first to a plurality of signals of n splitting signal and determine that in a plurality of coding units 230,240,250,260 and 270 which will be used for encoding first to each of n splitting signal with input signal.

With reference to Figure 14, coding unit 230,240,250,260 and 270 can be sequentially encoded to the n splitting signal to first respectively.And, if input signal is split into a plurality of band signals, then can be according to encoding to this band signal of inferior ordered pair of high-frequency band signals from the lowest band signal.

Under the situation of sequential encoding splitting signal, the encoding error of first front signal can be used for the current demand signal of encoding.As a result, can use different coding method code division signals, so in case stop signal distortion and the bandwidth retractility is provided.

With reference to Figure 14, coding unit 230 coding first splitting signal, first splitting signal that decoding has been encoded, and the error between the decoded signal and first splitting signal outputed to coding unit 240.Coding unit 240 uses the error of coding units 230 outputs second splitting signal of encoding.By this way, consider the encoding error of their previous splitting signals separately, coding second is to the m splitting signal.Therefore, can realize error-free encoding and improve tonequality.

Encoding device shown in Figure 14 is by the performed operation of the encoding device shown in the execution graph 1 to 14 can be from the incoming bit stream restoring signal inversely.

Figure 15 is the block diagram of decoding device according to an embodiment of the invention.With reference to Figure 15, decoding device comprises a parse module 800, demoder determination module 810, decoder module 820 and synthesis module 830.

Position parse module 800 extracts one or more coded signals and decodes the required additional information of this coded signal from incoming bit stream.

Decoder module 820 comprises first decoding unit 821 of the carrying out different coding/decoding methods a plurality of decoding units to m decoding unit 822.

Decoding determination module 810 is determined first decoding units 821 which to the m decoding unit 822 each coded signal of can decoding the most efficiently.Demoder determination module 810 can use with the similar method of method of the sort module 100 shown in Fig. 1 determines first decoding unit 821 which to the m decoding unit 822 each coded signal of can decoding the most efficiently.In other words, demoder determination module 810 can be determined first decoding unit 821 which to the m decoding unit 822 each coded signal of can decoding the most efficiently based on the characteristic of each coded signal.Preferably, demoder determination module 810 can be determined first decoding unit 821 which to the m decoding unit 822 each coded signal of can decoding the most efficiently based on the additional information of extracting from incoming bit stream.

Additional information can comprise: classification information identifies the classification under the information encoded that is classified by encoding device; Coding unit information, sign is used to produce the coding unit of this coded signal; With decoding unit information, sign will be used to the to decode decoding unit of this coded signal.

For example, demoder determination module 810 can be based on additional information and determine which classification coded signal belongs to, and selects first decoding unit 821 any one decoding unit corresponding to the classification of coded signal to the m decoding unit 822 for coded signal.In this case, selected decoding unit can have a kind of structure and makes its can decode the most efficiently signal of the classification that belongs to identical with the classification of coded signal.

Alternatively, demoder determination module 810 can be discerned the coding unit that is used to produce coded signal based on additional information, and selects first decoding unit 821 any one decoding unit corresponding to the coding unit of identification to the m decoding unit 822 for coded signal.For example, if produced coded signal by sound decorder, then can to select first decoding unit 821 for coded signal be any one decoding unit of Voice decoder to m decoding unit 822 to demoder determination module 810.

Alternatively, demoder determination module 810 can be discerned the decoding unit of decodable code coded signal based on additional information, and selects first decoding unit 821 any one decoding unit corresponding to the decoding unit of being discerned to the m decoding unit 822 for coded signal.

Alternatively, demoder determination module 810 can obtain the characteristic of decoded signal from additional information, and any one decoding unit of the signal of selecting first decoding unit 821 can decode the most efficiently to the m decoding unit 822 to have the characteristic identical with the characteristic of coded signal.

By this way, each coded signal that extracts from incoming bit stream is by first decoding unit 821 be defined as to the m decoding unit 822 decoding the most efficiently any one decoding unit coding of respective coding signal.Decoded signal is synthetic by synthesis module 830, thereby recovers original signal.

Position parse module 800 extracts the carve information about coded signal, the breath of taking a message of the number of coded signal and each coded signal for example, but and the decoded signal that provides of the synthetic decoder module 820 of synthesis module 830 reference segmentation information.

Synthesis module 830 can comprise a plurality of synthesis units of first synthesis unit 831 to n synthesis unit 832.First synthesis unit 831 each to the n synthesis unit 832 all can be synthesized the decoded signal that decoder module 820 is provided, and perhaps in the decoded signal some or all is carried out territories conversion or additional decodings.

First synthesis unit 831 to the n synthesis unit 832 can carry out post-processing operation to synthetic signal, and this post-processing operation is the inverse operation (inverse) of the pretreatment operation carried out of encoding device.Can extract the decoded information that indicates whether to carry out the information of post-processing operation and be used to carry out post-processing operation from incoming bit stream.

With reference to Figure 16, first synthesis unit 831 to the n synthesis unit 832, particularly, second synthesis unit 833 can comprise a plurality of preprocessors of first preprocessor 834 to n preprocessor 835.First synthesis unit 831 synthesizes individual signals with a plurality of decoded signals, and first preprocessor 834 to the n preprocessor 835 is to by the synthetic individual signals execution post-processing operation that obtains.

Indicate first preprocessor 834 which to the n preprocessor 835 can be included in the incoming bit stream to information by the synthetic individual signals execution post-processing operation that obtains.

First compositor 831 to the n compositor 832 can use the linear predictor coefficient that extracts from incoming bit stream to carrying out the linear prediction decoding by the synthetic individual signals that obtains, thus the recovery original signal.

The present invention can be embodied as the embodied on computer readable code that writes on the computer-readable recording medium.Computer-readable recording medium can be a pen recorder of storing any kind of data in the mode of embodied on computer readable.The example of computer-readable recording medium comprises ROM, RAM, CD-ROM, tape, floppy disk, optical data memories and the carrier wave data transmission of the Internet (for example, by).Computer-readable recording medium can be distributed on a plurality of computer systems that are connected to network, makes the embodied on computer readable code be written on it and in the mode of disperseing and carries out from it.Those skilled in the art can easily construct and realize function program, code and code segment required for the present invention.

Though illustrate and described the present invention particularly with reference to exemplary embodiment of the present invention, it will be understood by those skilled in the art that the various changes that to make form and details here and do not depart from the spirit and scope of the present invention that limit as claim.

Industrial applicibility

As mentioned above, according to the present invention, by the characteristic according to signal signal is categorized as one Or a plurality of classifications are also come whenever with the coding unit that can best satisfy the affiliated classification of corresponding signal Individual signal is encoded, and can have with best bit rate coding the signal of different qualities. Therefore, Can encode to the various signals that comprise audio frequency and voice signal efficiently.

Claims (23)

1. coding/decoding method comprises:

Extract a plurality of coded signals from incoming bit stream;

Determine in a plurality of coding/decoding methods which will be used to decode each of described coded signal;

Use the determined coding/decoding method described coded signal of decoding;

Decoded signal is synthesized individual signals; With

By being carried out post-processing operation, described individual signals recovers original signal.

2. coding/decoding method according to claim 1, the described execution of wherein said post-processing operation comprise carries out the linear prediction decoding to described individual signals.

3. coding/decoding method according to claim 1 also comprises from described incoming bit stream extraction described individual signals is carried out the required post-processing information of described post-processing operation.

4. coding/decoding method according to claim 3, wherein said post-processing information comprises the information about linear predictor coefficient.

5. coding/decoding method according to claim 3, wherein said post-processing information comprises the information about the all-pass filter coefficient.

6. coding/decoding method according to claim 5, wherein said all-pass filter coefficient is variable.

7. coding/decoding method according to claim 5, wherein said all-pass filter coefficient is determined according to the energy level of described coded signal.

8. coding/decoding method according to claim 1 comprises also from described incoming bit stream and extracts post-processing approach information that this post-processing approach message identification will be used for described individual signals is carried out the post-processing approach of described post-processing operation,

The described execution of wherein said post-processing operation comprises with reference to one in a plurality of post-processing approach of described post-processing approach Information Selection.

9. coding/decoding method according to claim 1, wherein said determine to comprise select can to decode the most efficiently in the described coding/decoding method each any one coding/decoding method of described coded signal.

10. decoding device comprises:

The position parse module, it extracts a plurality of coded signals from incoming bit stream;

The demoder determination module, it determines in a plurality of decoding units which will be used to decode each of described coded signal;

Decoder module, it comprises described decoding unit, and uses the described coded signal of determined decoding unit decodes;

Synthesis module, its synthetic described decoded signal; With

Post-processing module, it recovers original signal by individual signals is carried out post-processing operation.

11. decoding device according to claim 10, wherein said post-processing module is carried out the linear prediction decoding to described individual signals.

12. decoding device according to claim 10, wherein said position parse module is from described incoming bit stream extraction at least one relevant information with linear predictor coefficient and all-pass filter coefficient.

13. decoding device according to claim 10, wherein, institute's rheme parse module extracts post-processing approach information from described incoming bit stream, and this post-processing approach message identification will be used for described individual signals is carried out the post-processing approach of described post-processing operation,

Wherein, described post-processing module comprises a plurality of preprocessors, and with reference to one in the described preprocessor of described post-processing approach Information Selection.

14. a coding method comprises:

Input signal is carried out pretreatment operation, and making that described input signal can be exchanged into can be by encoded signals efficiently;

With pretreated signal segmentation is a plurality of splitting signals;

Based on each characteristic of described splitting signal, determine in a plurality of coding methods which will be used to encode each of described splitting signal;

Use the determined coding method described splitting signal of encoding; With

Produce bit stream based on the splitting signal of having encoded.

15. coding method according to claim 14, the described execution of wherein said pretreatment operation comprises: described individual signals is carried out linear prediction decoding.

16. coding method according to claim 14, the described execution of wherein said pretreatment operation comprises: use the pre-service function that comprises all-pass filter.

17. coding method according to claim 14 comprises also based on the energy level of surplus and masking threshold or target bit rate and calculates the position amount with each of the described splitting signal of encoding that described surplus obtains by the described execution of described pretreatment operation.

18. coding method according to claim 14, the described execution of wherein said pretreatment operation comprises:

Select in a plurality of preprocess methods one based in characteristic, external environmental information and the target bit rate of described input signal at least one; With

Use selected preprocess method that described input signal is carried out described pretreatment operation.

19. an encoding device comprises:

Pretreatment module, it carries out pretreatment operation to input signal, and making that described input signal can be exchanged into can be by encoded signals efficiently;

The signal segmentation module, it is a plurality of splitting signals with pretreated signal segmentation;

The scrambler determination module, it is based on each characteristic of described splitting signal, determines in a plurality of coding units which will be used to encode each of described splitting signal;

Coding module, it comprises described coding unit, and uses the determined coding unit described splitting signal of encoding; With

Position package module, it produces bit stream based on the splitting signal of having encoded.

20. encoding device according to claim 19, wherein said pretreatment module is carried out linear prediction decoding to described input signal.

21. encoding device according to claim 19, wherein said pretreatment module comprises all-pass filter.

22. encoding device according to claim 19, wherein, described pretreatment module comprises a plurality of pretreaters, and selects in the described pretreater one based in characteristic, external environmental information and the target bit rate of described input signal at least one.

23. a computer-readable recording medium has and is used for carrying out according to each the described coding/decoding method in the claim 1 to 9 or according to the program of each the described coding method in the claim 14 to 18.

Images