CN101710489A - Method and device capable of encoding and decoding audio by grade and encoding and decoding system - Google Patents

Abstract

The invention discloses method and device capable of encoding and decoding audio by grade and an encoding and decoding system, belonging to the technical field of source encoding and signal processing. The method comprises the following steps: obtaining a base layer code stream and NMR of each subband by compressing a code on an audio signal; obtaining an enhancing layer bit plane of the audio signal, wherein the enhancing horizon plane consists of various bit planes of each subband; obtaining the bit plane displacement of each subabnd according to the NMR of each subband; according to the bit plane displacement of each subabnd, translating the bit plane of each subband towards low bit to obtain the bit plane priority of each subband; according to the bit plane priority of each subband, encoding the bit plane of each subband to obtain an enhancing layer code stream; and packaging the base layer code stream and the enhancing layer code stream to obtain an encoded audio code stream. The invention can reduce the expenditure of extra code rate of the system and improve the tone quality of the decoded audio signal.

Description

The method and apparatus of scalable audio coding, decoding and coding/decoding system

Technical field

The present invention relates to information source coding and signal processing technology field, particularly the method and apparatus of scalable audio coding, decoding and coding/decoding system.

Background technology

The method of scalable audio coding can be carried out high code check compression to sound signal at coding side, obtain audio code stream, and can from audio code stream, recover different fidelitys according to concrete user's environment for use, content demand, equipment performance (for example bandwidth) etc. in decoding end.In addition, the method for scalable audio coding need not be preserved corresponding audio file respectively, can reduce memory space, in addition, in addition not reciprocity protection of the audio code stream after the classification, thereby can provide continual audio service in the network that bandwidth constantly changes.Method based on above-mentioned advantage scalable audio coding has been widely used in the audio compression coding.

Prior art provides the method for following two kinds of scalable audio codings, comprising:

First kind, include nuclear and seedless two kinds of patterns.Having under the kernel normal form, adopting AAC (AdvancedAudio Coding, Advanced Audio Coding) core encoder that sound signal is carried out compressed encoding earlier, obtaining basic layer bit stream, shining upon the residual error that obtains each conversion coefficient through residual error simultaneously; Under no kernel normal form, the residual error of each conversion coefficient is the frequency spectrum of original signal.The residual error of each conversion coefficient under above two kinds of patterns is done the bit cutting, obtain bit plane, wherein, this bit plane is made up of the bit plane of subband, JND (just noticeable distortion according to each subband, critical audible distortion) calculates the bit plane shift amount of each subband, and according to the bit plane shift amount of each subband, to the translation that makes progress of the bit plane of each subband, obtain the bit plane after the translation, the bitplanes layer bit stream that is enhanced of encoding encapsulates the audio code stream after obtaining encoding to basic layer bit stream and enhancement layer bitstream then.

Second kind, adopt the AAC core encoder that sound signal is carried out high code check compression, obtain the quantized value of each conversion coefficient, the quantized value of each conversion coefficient is done the bit plane that the bit cutting obtains each frequency band, the full range band is divided into some coding bands, from low frequency coding band, in each coding is with by from the low frequency to the high frequency, encode to lowest bit bit-by-bit plane from the highest-order bit.Wherein, for guaranteeing basic layer tonequality, when coding, defined the spectral range of basic layer; For raising the efficiency, adopted during coding based on contextual Bit-Plane Encoding.

In realizing process of the present invention, the inventor finds that there is following problem in prior art:

Above-mentioned first kind of technology, having under the kernel normal form, when enhancement layer is encoded, do not consider of the influence of actual quantization noise simultaneously to tonequality, the subband that promptly can not priority encoding can improve tonequality, enhancement layer coding does not satisfy human hearing characteristic, the gradable requirement of quality of service frequently of discontented footsteps; When enhancement layer is encoded, also need to write down the bit plane shift amount of each subband, so, increase system's expenditure of extra code rate.

Above-mentioned second kind of technology, when coding, carry out classification according to frequency band, do not consider the impression difference of people's ear to the subjective tonequality of different sub-band, promptly do not pay the utmost attention to the bigger subband of subjective tonequality influence, scrambler does not satisfy human hearing characteristic, can not satisfy the gradable requirement of audio service quality.

Summary of the invention

In order to reduce system's expenditure of extra code rate, improve the tonequality of decoded audio signal under the limited code check, the embodiment of the invention provides the method and apparatus and the coding/decoding system of scalable audio coding, decoding.Described technical scheme is as follows:

A kind of method of scalable audio coding, described method comprises:

By to Audio Signal Compression Coding, the noise that obtains basic layer bit stream and each subband covers than NMR (Noise to Mask Ratio, noise covers ratio);

Obtain the enhancement layer bit plane of described sound signal, described enhancement layer bit plane is made up of the bit plane of each subband;

Obtain the bit plane shift amount of each subband according to the NMR of each subband;

According to the bit plane shift amount of each subband, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband;

According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced;

Described basic layer bit stream and described enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.

The enhancement layer bit plane of the described sound signal of described acquisition specifically comprises:

According to each conversion coefficient of described sound signal and the scale factor of its place subband, calculate the normalization residual error of each conversion coefficient;

The normalization residual error of each conversion coefficient is done the quantification of default value bit, obtain described enhancement layer bit plane.

Described NMR according to each subband obtains the bit plane shift amount of each subband, specifically comprises:

From the NMR of all subbands, search maximum NMR;

The NMR of each subband and the NMR of described maximum are compared, obtain the NMR difference of each subband;

Determine the NMR difference range at the NMR difference place of each subband;

According to the NMR difference range of each subband, from the corresponding relation of bit plane shift amount and NMR difference range, search the bit plane shift amount of each subband.

The method that a kind of audio code stream that the method for utilizing described scalable audio coding is obtained is decoded, described method comprises:

By described audio code stream is decoded, obtain the quantized value of each conversion coefficient, the MSB of the bit plane of each subband (Most Significant Bit, the highest-order bit), the scale factor of each subband and enhanced layer information;

MSB according to the bit plane of enhanced layer information and each subband sets up the enhancement layer bit plane;

According to the scale factor of described enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals.

The MSB of described bit plane according to enhanced layer information and each subband sets up the enhancement layer bit plane, specifically comprises:

According to the MSB of the bit plane of each subband, set up blank enhancement layer bit plane, described enhancement layer bit plane comprises the bit plane of all subbands;

From enhanced layer information, read the data of the bit plane of each subband, the described data that read are inserted in proper order the position of described enhancement layer bit plane correspondence.

Described according to the scale factor of described enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals specifically comprises:

From described enhancement layer bit plane, read the bit plane of each conversion coefficient,, calculate the normalization residual error of each conversion coefficient according to the bit plane of each conversion coefficient;

Utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised, obtain the quantized value of revised each conversion coefficient;

According to the quantized value of revised each conversion coefficient and the scale factor of its place subband, acquire each conversion coefficient, each conversion coefficient is carried out inverse transformation obtain described sound signal.

A kind of device of scalable audio coding, described device comprises:

First obtains module, is used for by to Audio Signal Compression Coding, and the noise that obtains basic layer bit stream and each subband covers and compares NMR;

Second obtains module, is used to obtain the enhancement layer bit plane of described sound signal, and described enhancement layer bit plane is made up of the bit plane of each subband;

The 3rd obtains module, is used for obtaining according to the NMR of each subband the bit plane shift amount of each subband;

The translation module is used for the bit plane shift amount according to each subband, and the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband;

Coding module is used for the bit plane priority according to each subband, and the bit plane of each subband is encoded, and layer bit stream is enhanced;

Package module is used for described basic layer bit stream and described enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.

Described second obtains module specifically comprises:

First computing unit is used for calculating the normalization residual error of each conversion coefficient according to each conversion coefficient of described sound signal and the scale factor of its place subband;

Quantifying unit is used for the normalization residual error of each conversion coefficient is done the quantification of default value bit, obtains described enhancement layer bit plane.

The described the 3rd obtains module specifically comprises:

First searches the unit, is used for the NMR from all subbands, searches maximum NMR;

Comparing unit is used for the NMR of each subband and the NMR of described maximum are compared, and obtains the NMR difference of each subband;

Determining unit is used for determining the NMR difference range at the NMR difference place of each subband;

Second searches the unit, is used for the NMR difference range according to each subband, searches the bit plane shift amount of each subband from the corresponding relation of bit plane shift amount and NMR difference range.

The device that a kind of audio code stream that the device that utilizes described scalable audio coding is obtained is decoded, described device comprises:

Acquisition module is used for obtaining the quantized value of each conversion coefficient by described audio code stream is decoded, the highest-order bit MSB of the bit plane of each subband, the scale factor of each subband and enhanced layer information;

Set up module, be used for MSB, set up the enhancement layer bit plane according to the bit plane of enhanced layer information and each subband;

Rebuilding module is used for according to the scale factor of described enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals.

The described module of setting up specifically comprises:

Set up the unit, be used for the MSB according to the bit plane of each subband, set up blank enhancement layer bit plane, described enhancement layer bit plane comprises the bit plane of all subbands;

Insert the unit, be used for reading the data of the bit plane of each subband, the described data that read are inserted in proper order the position of described enhancement layer bit plane correspondence from enhanced layer information.

Described rebuilding module specifically comprises:

Second computing unit is used for reading from described enhancement layer bit plane the bit plane of each conversion coefficient, according to the bit plane of each conversion coefficient, calculates the normalization residual error of each conversion coefficient;

Amending unit is used to utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised, and obtains the quantized value of revised each conversion coefficient;

Obtain the unit, be used for acquiring each conversion coefficient, each conversion coefficient is carried out inverse transformation obtain described sound signal according to the quantized value of revised each conversion coefficient and the scale factor of its place subband.

By sound signal is carried out compressed encoding, obtain the NMR of basic layer bit stream and each subband, obtain the bit plane shift amount of each subband according to the NMR of each subband, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband; According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced; Basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.Owing to need when coding, not write down the bit plane shift amount of each subband, save system overhead; Obtain the bit plane shift amount of each subband according to the NMR of each subband, according to shift amount with the bit plane of each subband to low bit translation, thereby realize different sub-band is distributed different bit plane priority, bit plane priority according to each subband is encoded, can be to the bit plane priority encoding of the bigger subband of NMR, thus the tonequality of decoded audio signal guaranteed.

Description of drawings

Fig. 1 is the method flow diagram of a kind of scalable audio coding of providing of the embodiment of the invention 1;

Fig. 2 is the method detail flowchart of a kind of scalable audio coding of providing of the embodiment of the invention 1;

Fig. 3 is the method flow block diagram of a kind of scalable audio coding of providing of the embodiment of the invention 1;

Fig. 4 is first kind of bit plane synoptic diagram that the embodiment of the invention 1 provides;

Fig. 5 is second kind of bit plane synoptic diagram that the embodiment of the invention 1 provides;

Fig. 6 is the method flow diagram of a kind of decoding of providing of the embodiment of the invention 2;

Fig. 7 is the method detail flowchart of a kind of decoding of providing of the embodiment of the invention 2;

Fig. 8 is the method flow block diagram of a kind of decoding of providing of the embodiment of the invention 2;

Fig. 9 is a kind of blank enhancement layer bit plane synoptic diagram that the embodiment of the invention 2 provides;

Figure 10 is the apparatus structure synoptic diagram of a kind of scalable audio coding of providing of the embodiment of the invention 3;

Figure 11 is the apparatus structure synoptic diagram of a kind of decoding of providing of the embodiment of the invention 4;

Figure 12 is a kind of coding/decoding system structural representation that the embodiment of the invention 5 provides.

Embodiment

For making the purpose, technical solutions and advantages of the present invention clearer, embodiment of the present invention is described further in detail below in conjunction with accompanying drawing.

Embodiment 1

As shown in Figure 1, the embodiment of the invention provides a kind of audio coding method, comprising:

Step 101: by perceptual audio coder sound signal is carried out compressed encoding, obtain the NMR of basic layer bit stream and each subband;

Wherein, sound signal comprises a plurality of subbands, because the code check restriction, perceptual audio coder obtains basic layer bit stream and includes only the part subband, and these subband codebook number are non-vanishing entirely.Basic layer bit stream comprises the not scale factor of zero subband of the quantized value of each conversion coefficient that the non-vanishing subband of codebook number comprises and each codebook number.

Step 102: obtain the enhancement layer bit plane of sound signal, this enhancement layer bit plane comprises the bit plane of each subband;

Step 103: the bit plane shift amount that obtains each subband according to the NMR of each subband;

Step 104:, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband according to the bit plane shift amount of each subband;

Step 105: according to the bit plane priority of each subband that obtains, the bit plane of each subband is encoded, layer bit stream is enhanced;

Wherein, according to the bit plane shift amount each subband bit plane is carried out translation to low bit, make the bit plane priority of the big more subband of NMR high more, thus can be when coding, to the bit plane priority encoding of the big subband of NMR, thus the tonequality of assurance decoded audio signal.

Step 106: basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.

In embodiments of the present invention, utilize the AAC core encoder that sound signal is carried out compressed encoding, obtain the NMR of basic layer bit stream and each subband, and obtain the bit plane shift amount of each subband according to the NMR of each subband, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband; According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced; Basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.Owing to need when coding, not write down the bit plane shift amount of each subband, save system overhead; According to the bit plane shift amount bit plane of each subband is carried out translation to low bit, can be when coding, the subband priority encoding bigger to NMR, thus guaranteed the tonequality of decoded audio signal.

Referring to Fig. 2 and Fig. 3, the above-mentioned audio coding method that the embodiment of the invention provides can specifically comprise:

Step 201: adopt perceptual audio coder that sound signal is carried out compressed encoding, obtain the NMR of basic layer bit stream and each subband, wherein, basic layer bit stream comprises the scale factor of the subband that codebook number is non-vanishing and the quantized value of it each conversion coefficient that comprises;

Wherein, to coding audio signal, sound signal can be converted to a plurality of conversion coefficients, and these conversion coefficients are divided into a plurality of subbands by perceptual audio coder.Perceptual audio coder (Perceptual coding) is meant the audio compression scrambler that utilizes psychoacoustic model, for example, transform coder (transform coding), subband coder (subband coding), sine parameter scrambler (parameter coding) and linear predictive coding (linear predictive coding) etc., for another example: the AAC core encoder, MPEG (Moving PicturesExperts Group)-1 layer 1 (layer I), layer 2 (layer II) and layer 3 (layer III) scrambler, MPEG-4 etc. are multiple, when the conversion coefficient here is meant-result of frequency conversion, the time-the frequency conversion comprises DFT (DiscreteFourier Transform, discrete Fourier transformation), MDCT (Modified Digital CosineTransform, the discrete cosine transform of revising), DWT (Discrete Wavelet Transform, wavelet transform), WPT (Wavelet Packet Transform, wavelet package transforms), PQF (PolyphaseQuadrature Filter, Methods of Subband Filter Banks) etc.Therefore the form of the conversion coefficient that obtains of every kind of scrambler is all inequality, for example, the conversion coefficient that coding audio signal is obtained by the AAC core encoder is the MDCT frequency spectrum, the conversion coefficient that coding audio signal is obtained by MPEG-1 layer 2 (or claiming MUSICAM, Masking-pattern UniversalSub-band Integrated Coding and Multiplexing) scrambler is a sub-band filter.

Because the code check restriction, perceptual audio coder obtains basic layer bit stream and includes only the part subband, and these subband codebook number are non-vanishing entirely.Basic layer bit stream comprises the scale factor of the subband that the quantized value of each conversion coefficient that the non-vanishing subband of codebook number comprises and each codebook number are non-vanishing.Perceptual audio coder obtains the scale factor of the non-vanishing subband of each codebook number earlier, again by formula (1) quantizes each conversion coefficient that the non-vanishing subband of codebook number comprises, obtains the quantized value x_quant of each conversion coefficient that the non-vanishing subband of each codebook number comprises.

$x\_\mathrm{quant}=\mathrm{int}\{{[\mathrm{abs}(\mathrm{mdct}\_\mathrm{line})\×2^{-\frac{1}{4}(\mathrm{sf}\_\mathrm{decoder}-\mathrm{SF}\_\mathrm{OFFSET})}]}^{\frac{3}{4}}+\mathrm{MAGIC}\_\mathrm{NUMBER}\}...\left(1\right)$

Wherein, MAGIC_NUMBER=0.4054, sf_decoder are the scale factor of subband, and SF_OFFSET=100, mdct_line are conversion coefficient.

In audio coding, main error promptly rounds from the quantification in (1) formula.If can be to the correction that link is carried out certain form that rounds in (1) formula at coding side, then can make the error of each conversion coefficient obtain reducing to a certain degree, thereby in when decoding, each conversion coefficient that utilizes the quantized value of each conversion coefficient to recover obtains reduction to a certain degree.

Wherein, perceptual audio coder is not that zero subband is encoded to codebook number, also just is not that the scale factor of zero subband and the quantized value of it conversion coefficient that comprises write basic layer bit stream with codebook number.For codebook number is zero subband, if in decoding end can be that the quantized value of each conversion coefficient of comprising of zero subband is revised to codebook number also, in when decoding, each conversion coefficient that utilizes the quantized value of each conversion coefficient to recover obtains correction to a certain degree.In addition, perceptual audio coder is when coding, and also the codebook number with each subband writes basic layer bit stream.

In addition, perceptual audio coder when sound signal is carried out compressed encoding, can obtain the NMR of each subband in conjunction with psychoacoustic model.Wherein, some perceptual audio coder is when carrying out compressed encoding to sound signal, can access the NMR of each subband, and some perceptual audio coder is had to the SMR (Signal Mask Ratio, signal cover ratio) of each subband when sound signal is carried out compressed encoding, but can not get the NMR of each subband, so need be according to the SMR of each subband, calculate the NMR of each subband by following computing method, concrete steps are as follows:

(1), according to the conversion coefficient that each subband comprises, calculate the signal energy codec_e (n) of each subband by following formula (2):

$\mathrm{codec}\_e\left(n\right)=\underset{i=\mathrm{Low}\left(n\right)}{\overset{\mathrm{High}\left(n\right)}{\mathrm{\Σ}}}{(\mathrm{mdct}\_\mathrm{line}\left(i\right))}^2...\left(2\right)$

Wherein, the n in the formula (2) is a sub-band serial number, and Low (n) and High (n) are respectively the upper and lower borders of n subband, and mdct_line (i) is an i conversion coefficient;

(2), according to the signal energy codec_e and the SMR of each subband, cover thresholding Mask (n) by what following formula (3) calculated each band;

Mask(n)＝codec_e(n)/SMR(n)......(3)

(3), according to the quantized value of each conversion coefficient and the scale factor of its place subband, and reconstruct each conversion coefficient x_invauant (i) by following inverse quantization formula (4);

$x\_\mathrm{invquant}\left(i\right)=x\_\mathrm{quant}{\left(i\right)}^{\frac{4}{3}}\×2^{\frac{1}{4}(\mathrm{sf}\_\mathrm{decoder}-\mathrm{SF}\_\mathrm{OFFSET})}...\left(4\right)$

Wherein, x_quant (i) is the quantized value of i conversion coefficient obtaining by perceptual audio coder, and SF_OFFSET=100, sf_decoder are the scale factor of i conversion coefficient place subband.

(4), the conversion coefficient that comprises according to each subband and the conversion coefficient of reconstruction, calculate the quantizing noise Noise (n) of each subband by following formula (5):

$\mathrm{Noise}\left(n\right)=\underset{\mathrm{Low}\left(n\right)}{\overset{\mathrm{High}\left(n\right)}{\mathrm{\Σ}}}{[\mathrm{mdct}\_\mathrm{line}\left(i\right)-x\_\mathrm{invquant}\left(i\right)]}^2...\left(5\right)$

Wherein, i the conversion coefficient of the mdct_line (i) in the formula (5) for obtaining by perceptual audio coder, x_invauant (i) is for rebuilding i the conversion coefficient that obtains by formula (4).

(5), according to the quantizing noise Noise (n) of each subband with cover thresholding Mask (n) calculates each subband by following formula (6) NMR (n):

NMR(n)＝Noise(n)/Mask(n)......(6)

Step 202: according to each conversion coefficient and quantized value x_quant that the non-vanishing subband of codebook number comprises, by formula (7) calculate the normalization residual error x_enh of each conversion coefficient;

$x\_\mathrm{enh}={(\mathrm{abs}(\mathrm{mdct}\_\mathrm{line})\×2^{-\frac{1}{4}(\mathrm{sf}\_\mathrm{decoder}-\mathrm{SF}\_\mathrm{OFFSET})})}^{\frac{3}{4}}-x\_\mathrm{quant}...\left(7\right)$

Wherein, the normalization residual error x_enh ∈ (1.0,1.0) of each conversion coefficient that the non-vanishing subband of codebook number comprises, mdct_line is each conversion coefficient.

Step 203: for codebook number is those subbands of zero, calculates the scale factor sf_decoder that each codebook number is zero subband by following formula (8);

sf_decoder＝ceil[log2(maxspec ⁴)]+SF_OFFSET......(8)

Maxspec is the maximum conversion coefficient in each subband, and ceil is for being rounded up to function.

Step 204: according to codebook number is the scale factor of the subband at the conversion coefficient that comprises of zero subband and this conversion coefficient place, calculates the normalization residual error x_enh of conversion coefficient by following formula (9);

$x\_\mathrm{enh}={(\mathrm{abs}(\mathrm{mdct}\_\mathrm{line})\×2^{-\frac{1}{4}(\mathrm{sf}\_\mathrm{decoder}-\mathrm{SF}\_\mathrm{OFFSET})})}^{\frac{3}{4}}...\left(9\right)$

Wherein, codebook number is the normalization residual error x_enh ∈ (1.0,1.0) of each conversion coefficient of comprising of zero subband.

So far, obtain non-vanishing subband of codebook number and codebook number and be the normalization residual error x_enh of all conversion coefficients that zero subband comprises, and x_enh ∈ (1.0,1.0).Normalization residual error x_enh is used to revise the quantized value of conversion coefficient, and revised quantized value is used to rebuild conversion coefficient.Utilize normalization residual error x_enh to revise the quantized value of each conversion coefficient in decoding end, utilize revised quantized value to rebuild each conversion coefficient, just can repair each conversion coefficient to greatest extent.

Step 205: the normalization residual error x_enh of each conversion coefficient is done the quantification of M bit, layer bit plane that be enhanced, wherein, this enhancement layer bit plane has M bit, is made up of the bit plane of all subbands;

Particularly, at a conversion coefficient, the normalization residual error x_enh of this conversion coefficient is done the quantification of M bit, M+1 the bit plane that obtains this conversion coefficient is respectively { b (M+1), b (M), b (M-1), ..., b (1) }, wherein, bit plane b (M+1) is used to deposit the sign bit of the normalization residual error of this conversion coefficient, b (M), b (M-1) ..., b (1) is used to store the data division of the normalization residual error of this conversion coefficient, and bit plane b (M) and b (M+1) are merged.As stated above the normalization residual error of each conversion coefficient is carried out the quantification of M bit, acquire M bit plane of each conversion coefficient, the bit plane of the conversion coefficient that each subband is comprised is formed the bit plane of subband, wherein, the bit plane of each subband all has a MSB and LSB (Least SignificantBit, the lowest bit position), promptly be respectively the highest-order bit and the lowest bit position of the data division of subband, bit plane recomposition enhancement layer bit plane by each subband, be illustrated in figure 4 as the enhancement layer bit plane, each subband bit plane that this enhancement layer bit plane comprises has M bit, and the bit plane of each bit is made up of the bit plane of k subband.

Step 206: the NMR difference of calculating each subband according to the NMR of each subband;

Particularly, search maximum NMR from the NMR of all subbands, the NMR of each subband and the NMR of the maximum of searching are compared, the NMR that obtains each subband is with respect to the difference between the NMR of maximum, i.e. the NMR difference of each subband correspondence.

Step 207:, from the corresponding relation of bit plane shift amount and NMR difference range, obtain the bit plane shift amount of each subband correspondence according to the NMR difference of each subband;

Particularly, at a subband, NMR difference according to this subband, determine the NMR difference range at the NMR difference place of this subband, from the corresponding relation of bit plane shift amount and NMR difference range, read the bit plane shift amount of definite NMR difference range correspondence,, obtain the bit plane shift amount of all subbands as stated above the bit plane shift amount that reads bit plane shift amount as this subband.

Wherein, when the NMR of subband is big more, the NMR difference between the NMR of this subband and the maximum NMR is just more little, and the bit plane shift amount of this subband is just more little.As shown in table 1 for the M value is 13 o'clock, bit plane shift amount of being set up and NMR difference range corresponding relation.

Table 1

The bit plane shift amount	The NMR difference range
		??0	??[0，5.62)
??1	??[5.62，11.89)
		??2	??[11.89，17.56)
??3	??[17.56，23.40)
		??4	??[23.40，29.33)
??5	??[29.33，35.31)
		??6	??[35.31，41.33)
??7	??[41.33，47.33)
		??8	??[47.33，53.36)
??9	??[53.36，59.37)
		??10	??[59.37，65.39)
??11	??[65.39，71.40)
		??12	??[71.40，77.43)
??13	??[77.43，+∞)

Step 208: according to the bit plane shift amount of each subband, the bit plane with each subband in the enhancement layer bit plane carries out translation to low bit, obtains the bit plane priority of each subband;

Wherein, the enhancement layer bit plane carried out translation after, for each bit plane, before coding, the NMR of all subbands differs in 6dB, and arbitrarily between adjacent two bit planes, the NMR of any two subbands differs more than 6dB.So, the bit plane after the translation is carried out sequential encoding, can roughly follow the criterion of maximum NMR, the priority of its bit plane of subband that promptly NMR is bigger is higher relatively, thereby guarantees decoded tonequality.For example, after as shown in Figure 4 enhancement layer bit plane was shifted, the enhancement layer bit plane that obtains as shown in Figure 5.In Fig. 5, for the bit plane of the high more subband of MSB, promptly to the more little bit plane of the bit plane shift amount of low level translation, its bit plane priority is high more.For example, the bit plane shift amount of first subband is 3, all bit planes of first subband are to 3 bits of low level translation, the bit plane shift amount of second subband is 0, all not translations of bit plane of second subband, so the priority of the 1st of second subband～3rd bit plane will be higher than the 1st bit plane priority of first subband, and the priority of the 1st～3rd bit plane of second subband changes from high to low.

Step 209: in the enhancement layer bit plane after translation, encode by the bit plane priority of each subband, layer bit stream is enhanced, wherein, enhancement layer bitstream comprises the data division of enhancement layer bit plane, the MSB of each subband, and codebook number is the scale factor of zero subband;

Particularly, enhancement layer bit plane after the translation is scanned by the sequential scan mode from the low frequency to the high frequency, from MSB to LSB, wherein, for the high bit plane of priority is that the higher bit bit plane of the big more subband of MSB can be by priority scan, the data that the bit plane that scans comprises are encoded, obtain code stream, and the MSB of each subband write code stream, for codebook number is zero subband, the scale factor that also will be zero subband with each codebook number writes code stream, and layer bit stream so is enhanced; Wherein, the MSB of each subband, codebook number is that the scale factor of zero subband can be referred to as side information again.

For example, as shown in Figure 5 enhancement layer bit plane is scanned to high frequency, sequential scan mode from MSB to LSB by low frequency, because the MSB of the bit plane of the 2nd subband is the highest, its bit plane priority is the highest, at first Sao Miao bit plane is that (2, M), bitplanes bp (2 again for bit plane bp, M) encode, as stated above till the enhancement layer bit plane has been encoded entirely.

Wherein, for the big more subband of NMR, its corresponding MSB is big more, and bitplanes adopts sequential scan mode to encode, and can make the higher bit bit plane priority encoding of the big more subband of NMR.

Step 210: basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.

Wherein, enhancement layer bitstream is stored the normalization residual error of each conversion coefficient, utilizes the information in the enhancement layer bitstream that each conversion coefficient is revised in decoding end.

Wherein, in the present embodiment, the normalizing residual error of each conversion coefficient is converted to bit plane, rather than converts the quantized value of each conversion coefficient to bit plane, therefore, the bit plane of subband when the low bit translation, only can change the precision of the normalization residual error of conversion coefficient, and the quantized value of conversion coefficient is changed not, therefore, when coding, do not need to write down the bit plane shift amount of each subband, reduced the expenditure of extra code rate of system.

Wherein, people's ear do not hear and is lower than the sound that covers thresholding, and NMR is quantizing noise and the ratio that covers thresholding, if NMR is smaller, is that quantizing noise is lower than and covers thresholding yet, and people's ear is with the existence of imperceptible quantizing noise.When hanging down code check, because the code check restriction of basic layer coding, the quantizing noise of part subband is still greater than covering thresholding, when bitplanes is encoded, pay the utmost attention to the big subband of quantizing noise and will obtain higher tonequality income, the bit plane of the subband that the promptly preferential flat priority of contraposition is high is encoded.On the contrary, pay the utmost attention to those quantizing noises if do not pay the utmost attention to these subbands and be lower than the subband that covers thresholding, though code check has increased so, the audio service quality is not obviously promoted, thereby causes not matching of code check and audio service quality.So gradable auditory properties that requires scrambler will satisfy people's ear of audio service quality.Therefore, utilize the NMR of each subband to coordinate code check classification and subjective audio frequency quality grading, guarantee the consistance of code check classification and subjective tonequality classification most possibly.

In embodiments of the present invention, utilize the AAC core encoder that sound signal is carried out compressed encoding, obtain the NMR of basic layer bit stream and each subband, obtain the bit plane shift amount of each subband according to the NMR of each subband, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband; According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced; Basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.Because in when decoding, do not need to utilize the bit plane shift amount of each subband can recover the quantized value of revised each conversion coefficient, saved system overhead; Obtain the bit plane shift amount of each subband according to the NMR of each subband, bit plane shift amount according to each subband carries out translation with bit plane to low bit, can make the bit plane priority encoding of the bigger subband of NMR, guarantee the tonequality of decoded audio signal.

Embodiment 2

As shown in Figure 6, the embodiment of the invention also provides a kind of coding/decoding method, and the audio code stream that the audio coding method that utilizes embodiment 1 to provide obtains is decoded, and this method comprises:

Step 301: decode by the audio code stream of demoder after, obtain the quantized value of each conversion coefficient, the bit plane MSB and the enhanced layer information of the scale factor of each subband and each subband to compressed encoding;

Wherein, enhanced layer information comprises the data of all bit planes.

Step 302: the MSB according to the bit plane of enhanced layer information and each subband, set up the enhancement layer bit plane;

Wherein, the enhancement layer bit plane comprises the bit plane of all subbands, and the bit plane of each subband comprises the bit plane of a plurality of conversion coefficients, and each subband is made up of one or more bit planes.

Step 303: according to the quantized value of enhancement layer bit plane and each conversion coefficient, reconstructed audio signals.

In the embodiment of the invention, by demoder audio code stream is decoded, obtain the quantized value of each conversion coefficient, the MSB and the enhanced layer information of the scale factor of each subband and each subband bit plane, according to the MSB and the enhanced layer information of the bit plane of each subband, set up the enhancement layer bit plane, according to the quantized value of enhancement layer bit plane and each conversion coefficient, rebuild conversion coefficient, the conversion coefficient of rebuilding is carried out inverse transformation promptly obtain decoded audio signal.Owing to do not need to utilize the bit plane shift amount can recover the quantized value of revised conversion coefficient during decoding, thereby can reduce system's expenditure of extra code rate.

Shown in Fig. 7 and 8, the said method that the embodiment of the invention provides specifically comprises:

Step 401: utilize demoder that audio code stream is decoded, obtain the quantized value of each conversion coefficient, the scale factor of each subband, the MSB of the bit plane of enhanced layer information and each subband;

Wherein, enhanced layer information comprises the data of all bit planes, and it is that the quantized value of each conversion coefficient of comprising of zero subband is 0 that demoder obtains codebook number.

Step 402:, set up the bit plane of enhancement layer according to the MSB and the enhanced layer information of the bit plane of each subband;

Particularly, MSB according to each subband sets up blank enhancement layer bit plane, as shown in Figure 9, this enhancement layer bit plane is empty, to from enhanced layer information, read the data of the bit plane of each subband, and the data of the bit plane that reads will be filled in proper order the position of blank enhancement layer bit plane correspondence.

Wherein, the enhancement layer bit plane of foundation comprises the bit plane of a plurality of subbands, and the bit plane of subband comprises the bit plane of a plurality of conversion coefficients.

Step 403:, calculate the normalization residual error x_enh[k of each conversion coefficient] according to the bit plane of each conversion coefficient;

Particularly, the bit plane with each conversion coefficient is brought into the normalization residual error x_enh[k that following formula (10) calculates each conversion coefficient].

$x\_\mathrm{enh}\left[k\right]=[2b(k,M+1)-1]\×\underset{j=1}{\overset{m}{\mathrm{\Σ}}}b(k,M+1-j)\×2^{-j}...\left(10\right)$

Wherein, x_enh[k] be the normalization residual error of k conversion coefficient, (k M+1) is the sign bit of the normalization residual error of k conversion coefficient to b, and m is the MSB value of the bit plane of k conversion coefficient, and (k M+1-j) is j bit of k conversion coefficient to b.

Step 404: the quantized value of each conversion coefficient that the normalization residual error of utilizing each conversion coefficient obtains after to decoding is revised, and obtains the quantized value of revised each conversion coefficient.

Particularly, with the quantized value of each conversion coefficient and the normalization residual error is got and, obtain the quantized value of revised each conversion coefficient.

Step 405: the scale factor according to the subband at the quantized value of each conversion coefficient and its place, obtain each conversion coefficient, the conversion coefficient after promptly obtaining rebuilding carries out inverse transformation to the conversion coefficient of rebuilding and can obtain decoded audio signal.

Particularly, with the revised quantized value x ' of each conversion coefficient _iBe brought in the following inverse quantization formula (11) with the scale factor sf_decoder of the subband at its place, calculate each the conversion coefficient x_invquant (k) after the reconstruction, each conversion coefficient is carried out inverse transformation obtain decoded audio signal.

$x\_\mathrm{invquant}\left(k\right)={{x^{\&prime;}}_{\mathrm{<figure-callout\; id="4"\; label="k"\; filenames="H2009102374821E0000032.png,H2009102374821E0000041.png"\; state="\{\{state\}\}">k</figure-callout>}}}^{\frac{4}{3}}\&times;2^{\frac{1}{4}(\mathrm{sf}\_\mathrm{decoder}\left(n\right)-\mathrm{SF}\_\mathrm{OFFSET})}...\left(11\right)$

Wherein, x ' _kBe the quantized value of revised k conversion coefficient, x_invquant (k) is k conversion coefficient after rebuilding.Scrambler obtains the quantized value of the conversion coefficient that the non-vanishing subband of codebook number comprises, owing to there is quantization error, make that the quantizing noise of conversion coefficient is bigger, from enhancement layer bitstream, obtain the normalization residual error of conversion coefficient in the present embodiment, utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised, thereby utilize the conversion coefficient of the quantized value acquisition of revised conversion coefficient can access correction to greatest extent.

In addition, in embodiments of the present invention, when audio code stream is decoded, do not utilize the bit plane shift amount of each subband to recover the step of the bit plane of each subband, can recover the revised quantized value of each conversion coefficient, so can reduce the expenditure of extra code rate of system.

In the embodiment of the invention, by demoder audio code stream is decoded, obtain the quantized value of each conversion coefficient, the MSB and the enhanced layer information of the scale factor of each subband and each subband bit plane, MSB and enhanced layer information according to the bit plane of each subband, set up the enhancement layer bit plane, obtain the normalization residual error of each conversion coefficient from the enhancement layer bit plane, utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised, utilize the quantized value of each conversion coefficient of revising to reconstruct each conversion coefficient, each conversion coefficient is carried out inverse transformation obtain sound signal.Owing to do not need to utilize the bit plane shift amount of each subband, just can reconstruct sound signal, so can reduce system's expenditure of extra code rate.

Embodiment 3

As shown in figure 10, the embodiment of the invention provides a kind of device of scalable audio coding, comprising:

First obtains module 501, is used for obtaining the NMR of basic layer bit stream and each subband by to Audio Signal Compression Coding;

Second obtains module 502, is used to obtain the enhancement layer bit plane of sound signal, and this enhancement layer bit plane is made up of the bit plane of each subband;

The 3rd obtains module 503, is used for obtaining according to the NMR of each subband the bit plane shift amount of each subband;

Translation module 504 is used for the bit plane shift amount according to each subband, and the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband;

Coding module 505 is used for the bit plane priority according to each subband, and the bit plane of each subband is encoded, and layer bit stream is enhanced;

Package module 506 is used for basic layer bit stream and enhancement layer bitstream are encapsulated, the audio code stream after obtaining encoding.

Wherein, the second acquisition module 502 specifically comprises:

First computing unit is used for calculating the normalization residual error of each conversion coefficient according to each conversion coefficient of sound signal and the scale factor of its place subband;

Quantifying unit is used for the normalization residual error of each conversion coefficient is done the quantification of default value bit, and a layer bit plane is enhanced;

The 3rd obtains module 503 specifically comprises:

First searches the unit, is used for the NMR from all subbands, searches maximum NMR;

Comparing unit is used for the NMR of each subband and maximum NMR are compared, and obtains the NMR difference of each subband;

Determining unit is used for determining the NMR difference range at the NMR difference place of each subband;

Second searches the unit, is used for the NMR difference range according to each subband, searches the bit plane shift amount of each subband from the corresponding relation of bit plane shift amount and NMR difference range.

In embodiments of the present invention, by sound signal is carried out compressed encoding, obtain basic layer bit stream, obtain the bit plane shift amount of each subband according to the NMR of each subband, according to the bit plane shift amount of each subband with the bit plane of each subband to low bit translation, obtain the bit plane priority of each subband; According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced; Basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.Owing to need when coding, not write down the bit plane shift amount of each subband, save system's expenditure of extra code rate; According to the bit plane shift amount of each subband with each subband bit plane after low bit carries out translation, take the mode of sequential scanning can make the bit plane priority encoding of the bigger subband of NMR, guaranteed the tonequality of decoded audio signal.

Embodiment 4

As shown in figure 11, the device that the embodiment of the invention provides a kind of audio code stream that the device of the scalable audio coding that utilizes embodiment 3 is obtained to decode comprises:

Acquisition module 601 is used for obtaining the quantized value of each conversion coefficient by audio code stream is decoded the highest-order bit MSB of the bit plane of each subband, the scale factor of each subband and enhanced layer information;

Set up module 602, be used for MSB, set up the enhancement layer bit plane according to the bit plane of enhanced layer information and each subband;

Rebuilding module 603 is used for according to the scale factor of enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals.

Wherein, setting up module 602 specifically comprises:

Set up the unit, be used for the MSB according to the bit plane of each subband, set up blank enhancement layer bit plane, described enhancement layer bit plane comprises the bit plane of all conversion coefficients;

Insert the unit, be used for reading the data of the bit plane of each conversion coefficient, the data that read are inserted in proper order the position of enhancement layer bit plane correspondence from enhanced layer information;

Rebuilding module 603 specifically comprises:

Second computing unit is used for reading from the enhancement layer bit plane each bit of each conversion coefficient, according to each bit of each conversion coefficient, calculates the normalization residual error of each conversion coefficient;

Amending unit is used to utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised;

Obtain the unit, be used for acquiring each conversion coefficient, promptly obtain the conversion coefficient rebuild, the conversion coefficient of rebuilding is carried out inverse transformation promptly obtain decoded audio signal according to the quantized value of revised each conversion coefficient and the scale factor of its place subband.

In the embodiment of the invention, by demoder audio code stream is decoded, obtain the quantized value of each conversion coefficient, the MSB and the enhanced layer information of the scale factor of each subband and each subband bit plane, MSB and enhanced layer information according to the bit plane of each subband, set up the enhancement layer bit plane, according to the quantized value of enhancement layer bit plane and each conversion coefficient, reconstructed audio signals.Owing to do not need to utilize the bit plane shift amount of each subband, just the energy reconstructed audio signals can reduce system's expenditure of extra code rate.

Embodiment 5

As shown in figure 12, the embodiment of the invention provides a kind of coding/decoding system, comprising:

The device 701 of scalable audio coding is used for obtaining the NMR of basic layer bit stream and each subband by to Audio Signal Compression Coding; Obtain the enhancement layer bit plane of sound signal, this enhancement layer bit plane is made up of the bit plane of each subband; Obtain the bit plane shift amount of each subband according to the NMR of each subband; According to the bit plane shift amount of each subband, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband; According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced; Basic layer bit stream and enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding;

The device 702 of decoding is used for obtaining the quantized value of each conversion coefficient by audio code stream is decoded the highest-order bit MSB of the bit plane of each subband, the scale factor of each subband and enhanced layer information; MSB according to the bit plane of enhanced layer information and each subband sets up the enhancement layer bit plane; According to the scale factor of enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals.

In embodiments of the present invention, at coding side, behind the NMR that obtains basic layer bit stream and all subbands, obtain the enhancement layer bit plane, obtain the bit plane shift amount of each subband according to the NMR that calculates, again according to the bit plane shift amount, with the bit plane of each subband to low bit translation, obtain the bit plane priority of each subband, according to bit plane priority the bit plane of each subband is encoded, the layer bit stream that is enhanced encapsulates basic layer bit stream and enhancement layer bitstream the audio code stream after obtaining encoding.In decoding end, acquire the quantized value of each conversion coefficient through decoding, the MSB and the enhanced layer information of the scale factor of each subband and each subband bit plane, MSB and enhanced layer information according to the bit plane of each subband, set up the enhancement layer bit plane, according to the quantized value of enhancement layer bit plane and each conversion coefficient, reconstructed audio signals.Owing to when coding, do not need to write down the bit plane shift amount of each subband, saved system overhead; Obtain the bit plane shift amount of each subband according to the NMR of each subband, according to described bit plane shift amount the bit plane of each subband is carried out translation to low bit, when the bit plane after the translation is carried out sequential scanning, can make the bigger subband priority encoding of NMR guarantee the tonequality of decoding back sound signal like this.

All or part of content in the technical scheme that above embodiment provides can realize that its software program is stored in the storage medium that can read by software programming, storage medium for example: the hard disk in the computing machine, CD or floppy disk.

The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (12)

1. the method for a scalable audio coding is characterized in that, described method comprises:

By to Audio Signal Compression Coding, the noise that obtains basic layer bit stream and each subband covers and compares NMR;

Obtain the enhancement layer bit plane of described sound signal, described enhancement layer bit plane is made up of the bit plane of each subband;

Obtain the bit plane shift amount of each subband according to the NMR of each subband;

According to the bit plane shift amount of each subband, the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband;

According to the bit plane priority of each subband, the bit plane of each subband is encoded, layer bit stream is enhanced;

Described basic layer bit stream and described enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.

2. the method for claim 1 is characterized in that, the enhancement layer bit plane of the described sound signal of described acquisition specifically comprises:

According to each conversion coefficient of described sound signal and the scale factor of its place subband, calculate the normalization residual error of each conversion coefficient;

The normalization residual error of each conversion coefficient is done the quantification of default value bit, obtain described enhancement layer bit plane.

3. the method for claim 1 is characterized in that, described NMR according to each subband obtains the bit plane shift amount of each subband, specifically comprises:

From the NMR of all subbands, search maximum NMR;

The NMR of each subband and the NMR of described maximum are compared, obtain the NMR difference of each subband;

Determine the NMR difference range at the NMR difference place of each subband;

According to the NMR difference range of each subband, from the corresponding relation of bit plane shift amount and NMR difference range, search the bit plane shift amount of each subband.

4. method that the audio code stream that utilizes the described method of claim 1 to obtain is decoded is characterized in that described method comprises:

By described audio code stream is decoded, obtain the quantized value of each conversion coefficient, the highest-order bit MSB of the bit plane of each subband, the scale factor of each subband and enhanced layer information;

MSB according to the bit plane of enhanced layer information and each subband sets up the enhancement layer bit plane;

According to the scale factor of described enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals.

5. method as claimed in claim 4 is characterized in that, the MSB of described bit plane according to enhanced layer information and each subband sets up the enhancement layer bit plane, specifically comprises:

According to the MSB of the bit plane of each subband, set up blank enhancement layer bit plane, described enhancement layer bit plane comprises the bit plane of all subbands;

From enhanced layer information, read the data of the bit plane of each subband, the described data that read are inserted in proper order the position of described enhancement layer bit plane correspondence.

6. method as claimed in claim 4 is characterized in that, described according to the scale factor of described enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals specifically comprises:

From described enhancement layer bit plane, read the bit plane of each conversion coefficient,, calculate the normalization residual error of each conversion coefficient according to the bit plane of each conversion coefficient;

Utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised, obtain the quantized value of revised each conversion coefficient;

According to the quantized value of revised each conversion coefficient and the scale factor of its place subband, acquire each conversion coefficient, each conversion coefficient is carried out inverse transformation obtain described sound signal.

7. the device of a scalable audio coding is characterized in that, described device comprises:

First obtains module, is used for by to Audio Signal Compression Coding, and the noise that obtains basic layer bit stream and each subband covers and compares NMR;

Second obtains module, is used to obtain the enhancement layer bit plane of described sound signal, and described enhancement layer bit plane is made up of the bit plane of each subband;

The 3rd obtains module, is used for obtaining according to the NMR of each subband the bit plane shift amount of each subband;

The translation module is used for the bit plane shift amount according to each subband, and the bit plane of each subband to low bit translation, is obtained the bit plane priority of each subband;

Coding module is used for the bit plane priority according to each subband, and the bit plane of each subband is encoded, and layer bit stream is enhanced;

Package module is used for described basic layer bit stream and described enhancement layer bitstream are encapsulated the audio code stream after obtaining encoding.

8. device as claimed in claim 7 is characterized in that, described second obtains module specifically comprises:

First computing unit is used for calculating the normalization residual error of each conversion coefficient according to each conversion coefficient of described sound signal and the scale factor of its place subband;

Quantifying unit is used for the normalization residual error of each conversion coefficient is done the quantification of default value bit, obtains described enhancement layer bit plane.

9. device as claimed in claim 1 is characterized in that, the described the 3rd obtains module specifically comprises:

First searches the unit, is used for the NMR from all subbands, searches maximum NMR;

Comparing unit is used for the NMR of each subband and the NMR of described maximum are compared, and obtains the NMR difference of each subband;

Determining unit is used for determining the NMR difference range at the NMR difference place of each subband;

Second searches the unit, is used for the NMR difference range according to each subband, searches the bit plane shift amount of each subband from the corresponding relation of bit plane shift amount and NMR difference range.

10. device that the audio code stream that utilizes the described device of claim 7 to obtain is decoded is characterized in that described device comprises:

Acquisition module is used for obtaining the quantized value of each conversion coefficient by described audio code stream is decoded, the highest-order bit MSB of the bit plane of each subband, the scale factor of each subband and enhanced layer information;

Set up module, be used for MSB, set up the enhancement layer bit plane according to the bit plane of enhanced layer information and each subband;

Rebuilding module is used for according to the scale factor of described enhancement layer bit plane, each subband and the quantized value of each conversion coefficient, reconstructed audio signals.

11. device as claimed in claim 10 is characterized in that, the described module of setting up specifically comprises:

Set up the unit, be used for the MSB according to the bit plane of each subband, set up blank enhancement layer bit plane, described enhancement layer bit plane comprises the bit plane of all subbands;

Insert the unit, be used for reading the data of the bit plane of each subband, the described data that read are inserted in proper order the position of described enhancement layer bit plane correspondence from enhanced layer information.

12. device as claimed in claim 10 is characterized in that, described rebuilding module specifically comprises:

Second computing unit is used for reading from described enhancement layer bit plane the bit plane of each conversion coefficient, according to the bit plane of each conversion coefficient, calculates the normalization residual error of each conversion coefficient;

Amending unit is used to utilize the normalization residual error of each conversion coefficient that the quantized value of each conversion coefficient is revised, and obtains the quantized value of revised each conversion coefficient;

Obtain the unit, be used for acquiring each conversion coefficient, each conversion coefficient is carried out inverse transformation obtain described sound signal according to the quantized value of revised each conversion coefficient and the scale factor of its place subband.

Images