ES2218148T3 - USE OF ADAPTABLE GAIN QUANTIFICATION AND NON-UNIFORM LENGTHS OF SYMBOLS FOR AUDIO CODING. - Google Patents

Abstract

A method for encoding an input signal comprising: receiving the input signal and generating a block of subband signals of subband signal components representing a frequency subband of the input signal; compare the magnitudes of the subband signal block components with a threshold, place each component in one of two or more classes, according to the magnitude of the component, and obtain a gain factor; apply the gain factor to the components placed in one of the classes to modify the magnitudes of some of the components in the subband signal block; quantify the components in the subband block; and assembling in a coded signal the control information that drives the classification of the components and the symbols of non-uniform length representing the components of the quantized subband signal.

Description

Use of adaptive gain quantification and non-uniform lengths of symbols for audio coding.

Technical field

The present invention relates generally to coding and decoding signals. The present invention can used advantageously for coding and decoding in band heading, in which the frequency subband signals are encoded separately. The present invention is particularly Useful in perceptual audio coding systems.

Background in the art

There is a continuing interest in coding signals digital audio in a way that imposes low requirements of information capacity in the transmission channels and in the storage media, and that however can drive the Audio signals encoded with a high level of subjective quality. Perceptual coding systems try to achieve these contradictory objectives using for this a process by which encode and quantify the audio signals so that it is done use of larger spectral components within the audio signal to mask or make the quantification noise inaudible resulting. In general, it is advantageous to control the shape and amplitude of the quantification noise spectrum so that the it is just below the masking threshold psychoacoustic signal to be encoded.

A perceptual coding process can be carried out by means of a so-called split band encoder, which applies a bank of analysis filters to the audio signal to get subband signals that have bandwidths that are commensurable with the critical bands of the human auditory system, estimate the masking threshold of the audio signal by applying for this a perceptual model to the subband signals or to some another measure of the spectral content of the audio signal, establishes quantization step sizes to quantify the signals of subband that are just small enough for the resulting quantization noise is just below the threshold estimated masking of the audio signal, quantifies the Subband signals according to the step sizes of established quantification, and assembles in a coded signal a plurality of symbols representing subband signals quantified A process of perceptual decoding Complementary can be done using a decoder split band that extracts the encoded signal symbols and recover quantified subband signals from it, obtain an unquantified representation of subband signals quantified, and applies a bank of synthesis filters to unquantified representations to generate an audio signal which is, ideally, perceptually indifferent from the signal original audio

In the coding processes in these systems a uniform length symbol is usually used for coding represent the elements or components of the signal quantified in Each subband signal. Unfortunately, the use of symbols of uniform length imposes an information capacity greater than the That is necessary. The required information capacity can be reduced by using symbols of non-uniform length to represent the quantified components in each signal of Subband

A technique to provide symbols of non-uniform length to encode subband signals has been described in US Pat. No. 5,924,064. According to this technique, the calculated power of each subband signal is used to categorize the subband signal, which determines a size of the respective quantification step and a coding table of entropy for Huffman coding of the components of the quantified subband signal. Typically, code tables of Huffman have been designed using "training signals" that have been selected to represent the signals to be encoded in real applications. Huffman coding can provide very good coding gain if the function of average probability density (PDF) of training signals is reasonably close to the PDF of the actual signal to be encoded, and if the PDF is not flat.

If the PDF of the actual signal to be encoded does not is close to the average PDF of the training signals, the Huffman coding will not make a profit of the coding, but may incur a lack of coding, increasing signal information capacity requirements coded This problem can be minimized by using it. multiple code books corresponding to the different signals from PDF; however, an additional space of storage to store code books and a Additional processing to encode the signal according to each code book and then choose the one that provides the best results.

The need for a technique of encoding that can represent blocks of signal components of subbands quantified using symbols of non-uniform length within each subband, that is, not dependent on any PDF particular values of the components, and that may be carried out effectively using a minimum of calculation resources and of memory

Exhibition of the invention

An object of the present invention is that of provide the advantages that can be obtained by using non-uniform length symbols to represent components of quantified signal, such as subband signal components within a respective frequency subband in a system of split band coding.

The present invention achieves this objective using a technique that does not depend on any particular PDF of values of the components to achieve a good coding gain and which can be effectively implemented using a minimum of Calculation and memory resources. In some applications, coding systems can advantageously use features of the present invention, together with other techniques, such as Huffman coding.

According to the principles according to one aspect of the present invention, a method for encoding an input signal comprises receiving the input signal and generating a block of signals subband of subband signal components representing a frequency subband of the input signal; compare the magnitudes of the subband signal block components with a threshold, put each component in one of two or more kinds of according to the magnitude of the component, and get a factor of gain; apply the gain factor to the components located in one of the classes to modify the magnitudes of some of the components in the subband signal block, quantify the components in the subband signal block; and assemble in a encoded signal the control information leading to the classification of components and symbols of non-uniform length representing the components of the subband signal quantified

According to the principles according to another aspect of the present invention, a method for decoding a signal encoded comprises receiving the encoded signal and obtaining from it control information and symbols of non-uniform length, and obtain of the non-uniform length symbols signal components of quantified subband representing a frequency subband of an input signal; Unquantify the signal components of subband to obtain unquantified components of the signal from subband; apply a gain factor to modify magnitudes of some of the unquantified components, according to the control information; and generate an output signal in response to the unquantified components of the subband signal.

These methods can be incorporated into a medium. as an instructional program that can be executed by a device for practicing the present invention.

According to the principles according to another aspect of the present invention, an apparatus for qualifying an input signal it comprises an analysis filter that has an input that receives the input signal and that has an output through which it provides a block of subband signals of signal components subband representing a subband frequency signal input a coupled subband signal block analyzer to the analysis filter, which compares magnitudes of the components in the subband signal block with a threshold, puts each component in one of two or more classes, according to the magnitude of the component, and obtains a gain factor; a subband signal component processor coupled to subband signal block analyzer that applies the factor of gain to the components placed in one of the classes to modify the magnitudes of some of the components in the block of subband signals; a first quantifier coupled to Subband signal processor that quantifies the components of the block of subband signals that have the magnitudes modified according to the profit factor; and a formatter coupled to the first quantifier that assembles symbols of non-uniform length representing the components of the subband signal quantified and control information leading to the classification of the components in an encoded signal.

According to the principles according to yet another aspect of the present invention, in an apparatus for decoding a coded signal, the apparatus comprises a deformer that receives the encoded signal and obtains from the same control information and symbols of non-uniform length, and obtains from the symbols of non-uniform length subband signal components quantified; a first quantifier coupled to deformer, which quantifies some of the components of the Subband signal in the block, according to the information of control, to obtain first unquantified components; a subband signal block processor coupled to the first quantifier, which applies a gain factor to modify the magnitudes of some of the first components unquantified in the subband signal block, according to control information; and a synthesis filter that has a input coupled to the processor of the subband signals and that has an output through which a signal of exit.

According to the principles according to yet another aspect of the present invention, a medium conducts (1) symbols of non-uniform length representing signal components of quantified subband, in which the signal components of quantified subband correspond to elements of a block of subband signals representing a frequency subband of a audio signal; (2) control information, which indicates a classification of subband signal components quantified, according to the magnitudes of the corresponding elements of the subband signal block; and (3) an indication of a gain factor that belongs to magnitudes of the components of the quantified subband signal, according to the information of control.

The various features of the present invention and your preferred embodiments may be better understood if done reference to the study that follows and the accompanying drawings, in which reference numbers that are equal make reference to the same elements in the various figures. The content of the study that follows, and the drawings, are exposed only as examples, and should not be understood to represent limitations regarding the scope of the present invention.

Brief description of the drawings

Fig. 1 is a block diagram of an encoder of split wave that incorporates quantification adaptable by the gain.

Fig. 2 is a block diagram of a split wave decoder incorporating quantification adaptable for profit.

Fig. 3 is an organization chart in which they are illustrated the steps of a repetitive bit allocation process.

Figs. 4 and 5 are graphic illustrations of hypothetical blocks of subband signal components and of the effects of applying the gain to the components.

Fig. 6 is a block diagram of stages of cascade gain, for adaptive quantification by gain.

Figs. 7 and 8 are graphic illustrations of quantification functions

Figs. 9A to 9C illustrate how it can be done a split interval quantification function using a Planning transformation

Figs. 10 to 12 are graphic illustrations of quantification functions

Fig. 13 is a block diagram of an apparatus which can be used to implement various aspects of the present invention.

Modes for the implementation of the invention A. Coding system

The present invention is oriented towards performance improvement of information representation quantified, such as audio information, and finds advantageous application in the coding systems they use Splitband encoders and splitband decoders. In Figs. 1 and 2, respectively, embodiments have been illustrated of a split band encoder and a band decoder item incorporating various aspects of the present invention.

1. Encoder a) Analysis filtering

In Fig. 1, the analysis filter bank 12 receive an input signal by way 11, divide the signal from entry into subband signals representing subbands of frequency of the input signal, and passes the subband signals to along paths 13 and 23. For clarity of illustration, the embodiments depicted in Figs. 1 and 2 illustrate components for only two subbands; however, it is common that an encoder and a decoder split into a band perceptual coding system process many more subbands than have bandwidths that are commensurate with the widths of Critical band of the human auditory system.

The analysis filter bank 12 can be performed in a variety of ways, including those of polyphase filters, mesh filters, quadrature mirror filters (QMF), various block transformations from time domain to frequency domain, including transformations of the Fourier series type, transformations of filter banks modulated according to the cosine function and transformations of small waves. In preferred embodiments, the filter bank is performed by weighting or modulating overlapping blocks of digital audio samples with an analysis window function and applying a particular Modified Discrete Cosine Function Transformation (MDCT) to particular window-weighted blocks. This MDCT is called the "Aliasing" Cancellation in Time Domain (TDAC) transformation (Aliasing = Serrated diagonals when printing or on the video monitor) and has been described by Princen, Johnson and Bradley in "Subband / Transform Coding Using Filter Bank Designs Based on Time Domain Aliasing Cancellation ", Proc. Inc. Conf. Acoust. Speech and Signal Proc . May 1987, p. 2161-2164. Although the choice of the form of the embodiment can have a profound effect on the actions of the coding system, no particular embodiment of the analysis filter bank is important in terms of its concept for the present invention.

Subband signals passed along the paths 13 and 23 each comprise signal components of Subband that are arranged in blocks. In one embodiment preferred, each subband signal block is represented in a form of scaled block, in which the components are climbs with respect to a scale factor. It can be used, by example, a floating point block (BFF) form.

If the analysis filter bank 12 is performed in a block form, for example, subband signals are generated applying for this the transformation to a block of samples of input signals to generate a block of coefficients of transformation, and then grouping one or more coefficients of adjacent transformation, to form the signal blocks of Subband If the analysis filter bank 12 is performed by another type of digital filter, such as a QMF, for example, is generate the subband signals by applying the filter to a sample sequence of input signals to generate a sequence of subband signal samples for each subband of frequency and then grouping the subband signal samples into blocks The components of the subband signals for these two examples are transformation coefficients and signal samples of subband, respectively.

b) Perceptual modeling

In a preferred embodiment of a perceptual coding system, the encoder uses a perceptual model to establish a respective quantization step size to quantify each subband signal. A method in which a perceptual model is used to assign bits for adaptation is illustrated in Fig. 3. According to this method, in step 51 a perceptual model is applied to information representing characteristics of the input signal to establish a desired quantization noise spectrum. In many embodiments, the noise levels in this spectrum correspond to the estimated psychoacoustic masking threshold of the input signal. In step 52, the sizes of the quantization step of the proposed quantization steps are established to quantify the block components of the subband signals. In step 53, the bit assignments that are required to obtain the proposed quantization step sizes for all subband signal components are determined. Preferably, a margin is left to take into account the effects of noise extension of the synthesis filter bank in the split-band decoder to be used to decode the encoded signal. In US Pat. No. 5,623,577 and in US Pat. Serial No. 09 / 289,865 by Ubale, et al., Entitled "Quantization in Perceptual Audio Coders with Compensation for Synthesis Filter Noise Spreading"("Quantification in Perceptual Audio Encoders with Compensation for the Extension of Synthesis Filter Noise", presented with dated April 12, 1999. that both are hereby incorporated by reference, several methods have been described to determine such
margin.

In step 54 it is determined whether the total of the required allocations differs significantly from the total number of bits available for quantification. If the total allocation is too high, the proposed quantization step sizes are increased in step 55. If the total allocation is too low, the sizes of the proposed quantization step are decreased in step 55. The process returns to step 53 and reiterates this process until in step 54 it determines that the total allocation required to obtain the proposed quantization step sizes is sufficiently close to the total number of available bits. Next, in step 56, the components of the subband signals are quantified according to the sizes of the quantization step established.
clarified

c) Adaptive quantification by profit

Adaptive quantification can be incorporated for the gain in the method described above including for this several aspects of the present invention in step 53, for example. Although the method described above is typical of many perceptual coding systems, it is just a example of a coding process that can incorporate the present invention The present invention can be used in systems of coding in which essentially any criteria are used subjective and / or objective to establish the step size for Quantify the signal components. To facilitate your study, simplified embodiments have been used here to explain several Aspects of the present invention.

The subband signal block for a frequency subband is passed along path 13 to subband signal analyzer 14, which compares the magnitude of the components of the subband signals of each block with a threshold, and puts each component in one of two classes, of according to the magnitude of the component. Control information which leads the classification of the components is passed to formatter 19. In a preferred embodiment, the components that they are of a magnitude equal to or less than the threshold are put in a first class. The subband signal analyzer 14 obtains also a gain factor for later use. How will it be explained? in what follows, preferably the value of the gain factor It is related in some way to the threshold level. By For example, the threshold can be expressed as a function of only the gain factor Alternatively, the threshold can be expressed as a function of the profit factor and by others considerations.

The subband signal components that are placed in the first class are passed to profit element 15, which applies the gain factor obtained by the analyzer 14 of subband signals to each component of the first class, and the components modified by profit are then passed to quantifier 17. Quantifier 17 quantifies the components modified by gain according to a first step size of quantification and passes the resulting quantified components to formatter 19. In a preferred embodiment, the first size of the quantification step is established according to a model perceptual and according to the threshold value used by the 14 subband signal analyzer.

The subband signal components that do not they are put in the first class they are passed along the path 16 to quantifier 18, which quantifies those components according with a second size of the quantification step. The second size of the quantization step can be equal to the first step size quantification; however, in a preferred embodiment, the second quantization step size is smaller than the first quantization step size.

The subband signal block for the second frequency subband is passed along path 23 and is processed by the subband signal analyzer 24, the element of profit 25, and quantifiers 27 and 28, in the same way described above for the first subband of frequency. In a preferred embodiment, the threshold used for each frequency subband is adaptable and independent of the threshold used for other frequency subbands.

d) Formatting the encoded signal

The formatter 19 assembles the control information that drives the classification of the components and the symbols of non-uniform length representing the quantized subband signal components in an encoded signal, and passes the encoded signal along the path 20 to be driven by transmission media that include baseband or modulated communication paths across the spectrum that includes from supersonic to ultraviolet frequencies, or by storage media that include magnetic tape, magnetic disk and optical disk, which conduct information using a magnetic recording or recording technology
optical.

The symbols used to represent the quantified components may be identical to the values quantified, or they can be of some type of code derived from Quantified values. For example, the symbols can be obtained directly from a quantifier, or they can be obtained by some process such as Huffman's coding of the quantified values. The quantified values themselves they can be easily used as the symbols of length not uniform, because no bit numbers can be assigned uniform to the subband signal components quantified in a subband

2. Decoder a) Deformation of the encoded signal

In Fig. 2, the deformer 32 receives a signal encoded by path 31 and gets the same symbols as represent quantified subband signal components and control information that drives the classification of components. Decoding processes can be applied in the extent to which it is necessary to deduct the components Quantified symbols. In a preferred embodiment, put the components modified by profit in a first class. The deformer 32 also obtains any information that may be necessary, by any perceptual models or bit allocation processes, for example.

b) Adaptive quantification by profit

The quantifier 33 receives the components for a block of subband signals that are placed in the first class, decrypts them according to a first size of the quantization step, and passes the result to the gain element 35. In a preferred embodiment, the The first size of the quantification step is established according to a perceptual model and according to a threshold that was used to classify the components of the sub-signal.
band.

The profit element 35 applies a factor of gain to the unquantified components received from quantifier 33, and passes the components modified by the fusion gain 37. The operation of the gain element 35 reverses the profit modifications provided by the gain element 15 in the companion encoder. How has it explained above, preferably this factor of gain is associated with the threshold that was used to classify the Subband signal components.

The subband signal components that do not are placed in the first class are passed to the quantifier 34, which unquantifies those components according to a second size of the quantification step, and passes the result to the merger 37. The second size of the quantification step may be equal to first size of the quantification step; however, in a preferred embodiment, the second size of the quantification step It is smaller than the first size of the quantification step.

In the fusion 37 a block of signals of subband melting for it the unquantified components modified by the gain received from profit element 35 with the quantified components received from the quantifier 36, and passes the resulting subband signal block along from path 38, to the synthesis filter bank 39.

The quantified components in the block of Subband signals for the second frequency subband are processed by quantifiers 43 and 44, the profit element 45, and fusion 47, in the same manner as described in what precedes for the first frequency subband, and passes the block of resulting subband signals along road 48, to the bank of synthesis filters 39.

c) Synthesis filtering

The synthesis filter bank 39 can be carried out in a great diversity of ways, which with complementary to the ways previously considered to make the filter bank of analysis 12. An output signal is generated along path 40 in response to subband signal component blocks received by roads 38 and 48.

B. Features 1. Classification of signal components subband a) Simplified threshold function

The effects of quantification adaptable by the gain can be seen by referring to Fig. 4, in which hypothetical blocks 111, 112 and 113 have been illustrated of subband signal components. In the illustrated example, each subband signal block comprises eight components numbered 1 to 8. Each component is represented by a vertical line, and the magnitude of each component is represented by the height of the respective line. For example, component 1 in block 111 it has a magnitude slightly greater than the value 0.25, as illustrated on the ordinate axis in the graph.

Line 102 represents a threshold at level 0.50. Each component of block 111 can be placed in one of two classes, comparing the magnitudes of the respective components with the threshold. The components that have a magnitude equal or less than the threshold are put in a first class. The Remaining components are put in a second class. Alternatively, you can get results slightly different if the components are classified putting for it in the first class those components that have a magnitude strictly lower than the threshold. To facilitate the analysis, the threshold comparisons made according to the first example they will be assumed and mentioned here more in particular.

The components in block 112 are obtained by applying a gain factor of two to each component of block 111 that is placed in the first class. For example, the magnitude of component 1 in block 112, which is slightly greater than 0.500, is obtained by multiplying the magnitude of component 1 in block 111 by a gain factor equal to two. Conversely, the magnitude of component 2 in block 112 is equal to the magnitude of component 2 in block 111, due to that that component was placed in the second class and is not modified by the profit factor.

Line 104 represents a threshold at level 0.25. Each component in block 111 can be placed in one of two classes, comparing the magnitudes of the respective components with that threshold and putting the components that have a magnitude equal to or less than the threshold in a first class. The Remaining components are put in a second class.

The components in block 113 are obtained applying a gain factor of four to each component in the block 111 that has been placed in the first class. For example, the magnitude of component 3 in block 113, which is of approximately 0.44, is obtained by multiplying the magnitude of the component 3 of block 111, which is about 0.11, for a profit factor equal to four. Conversely, the magnitude of the component 1 in block 113 is equal to the magnitude of the component 1 in block 111, because this component was put in the second class and is not modified by the factor of gain.

The threshold can be expressed as a function of Only the profit factor. As illustrated by these two examples, the threshold can be expressed as

(1) Th = 1 / G

where

Th = the threshold value; Y

G = gain factor.

b) Alternative threshold function

Unfortunately, a threshold obtained by the expression (1) may be too large, because a component of the subband signal having a magnitude that is slightly smaller than the Th threshold, when modified by the gain factor G, can Overload the quantifier.

It is said that a value overloads a quantifier if the quantization error of that value exceeds half the size of the quantification step. For quantifiers symmetrical that have a uniform quantization step size that quantifies values within a range from about -1 up to +1, the region of positive quantities that overload the quantifier can be expressed as

(2a) Q_ {OL}> Q_ {MAX} + \ frac {\ Delta Q} {2}

and the region of negative values that overload to the quantifier can be expressed how

(2b) Q_ {OL} <- Q_ {MAX} - \ frac {\ Delta Q} {2}

where

Q_ {OL} = a value that overloads the quantifier;

Q_ {MAX} = maximum positive quantified value; Y

ΔQ = step size of quantification.

For a signed quantizer bit symmetric mid - tread b having a size of the uniform quantization step that quantifies values within a range from approximately -1 to +1, the maximum positive quantized value Q_ {MAX} is equal to 1 -2 1-b, the size of the quantization step ΔQ is equal to 2 1-b, and half the size of the quantization step is equal to 2-b. The expression (2a) for positive overload values can also be written as

(3a) Q_ {OL}> 1 - 2 1-b + 2 - b = 1 - 2 <b>

and the expression (2b) for negative values of the overload can also be written how

(3b) Q_ {OL} < - (1-2 ^ 1-b) - 2 ^ - b = -1 + 2 <b>

Line 100 in Fig. 4 represents the limit of positive overload values for a quantifier with 3-bit symmetric half footprint sign. The negative margin of This quantifier has not been represented. The quantified value Maximum positive for this quantifier is 0.75 = (1-2 ^ 1-3 *) and half the size of the quantification step is 0.125 = 2 -3; therefore the limit for positive overload values for this Quantifier is 0.875 = (1 - 2 - 3). The limit for Negative values of the overload is -0.875.

Component 5 in block 111 has a magnitude that is slightly less than the 0.500 value threshold. When a profit factor equal to two, the resulting magnitude exceeds the overload limit of the quantifier A similar problem is planted for component 6 when a threshold equal to 0.250 is used, with a gain factor equal to 4.

A threshold value for positive quantities that avoid overload and plan optimally the domain of values of the positive components in the first class in the positive margin of a quantifier can be expressed as

(4a) Th = \ frac {Q_ {OL}} {G}

The threshold for negative quantities can express yourself as

(4b) Th- \ frac {Q_ {OL}} {G}

In all that remains of this study, only The positive threshold will be considered. With this simplification I don't know loses no generality, because those operations in which component quantities with a positive threshold are compared are equivalent to other operations in which amplitudes of components with positive and negative thresholds.

For the symmetric, half-bit signed quantizer of b bits described above, the threshold function of the expression (4a) can also be written as

(5) Th = \ frac {1-2 ^ - b}} {G}

The effects of quantification adaptable by the Gain using that alternative threshold are illustrated in Fig. 5, in which hypothetical blocks 121, 122, 123 have been illustrated and 124 of the subband signal components. In the example illustrated, each subband signal block comprises eight numbered components from 1 to 8, whose magnitudes are represented by the lengths of the respective lines vertical Lines 102 and 104 represent the thresholds for a 3-bit symmetric half-finger signed quantifier for profit factors equal to 2 and 4, respectively. 100 line represents the limit of positive overload values for This quantifier.

The components of signal block 122 of subband can be obtained by comparing the magnitudes of the components of block 121 with threshold 102 and applying a gain of G = 2 to components that have equal magnitudes or less than the threshold. Similarly, the components of block 123 of subband signals can be obtained by comparing the magnitudes of the components of block 121 with threshold 104 and applying a gain of G = 4 to components that have equal magnitudes or less than that threshold. The components of the signal block 124 of subband can be obtained using a waterfall technique, which describe in the following. Unlike the examples illustrated in Fig. 4 for the first threshold considered above, none of the components modified by the gain represented in Fig. 5 it exceeds the overload limit of the quantifier.

On the one hand, the alternative threshold according to the expression (5) is desirable, because it avoids the overload of the quantifier for small components of the first class and optimally loads the quantifier. On the other hand, this threshold may not be desirable in some embodiments, in which an optimal quantization step size is sought, because the threshold cannot be determined until the quantization step size has been established. In embodiments in which the size of the quantization step is adapted by bit allocation, the size of the quantization step cannot be set until the allocation of bits b for a respective subband signal block is known. This disadvantage is explained in more detail in the following.

2. Quantification

Preferably, the quantization step size of the quantifiers used to quantify components of a subband signal block is adapted in response to the gain factor for that block. In an embodiment in which a process similar to the one considered above and illustrated in Fig. 3 is used, a number of bits b is signaled to each component within a subband signal block, and then the size is adapted of the quantization step, and possibly the bit allocation, for each component, according to the gain factor selected for that block. For this embodiment, the gain factor is selected from four possible values representing gains of 1, 2, 4, and 8. The components within that block are quantified using a quantifier with a half-symmetric footprint sign.

Components of larger magnitudes that are not classified in the first class and that have not been modified by gain are assigned the same number b of bits that would have been assigned without the benefit of the present invention. In an alternative embodiment, in which a split interval quantification function considered in the following is used, the bit allocation for those components of larger magnitudes may be reduced as corresponding to some gain factors.

To components of magnitudes smaller than have qualified in the first class and are modified by the gain is assigned a number of bits according to the values represented in Table I.

TABLE I

Gain Assignment one b two b-1 3 b-2 4 b-3

A gain factor equal to 1 for a particular subband signal block indicates that the gain modification characteristic of the present invention is not applied to that block; therefore, each component is assigned the same number b of bits as would be assigned without the benefit of the present invention. The use of the gain factor G = 2, 4 and 8 for a particular subband signal block can potentially provide the benefit of a reduced allocation of 1, 2, and 3 bits, respectively, for each component of smaller magnitude in the Subband block

The assignments illustrated in Table I are subject to the limitation that the number of bits assigned to each component cannot be less than one. For example, if the bit allocation process assigned b = 3 bits to the components of a particular subband signal block, and a gain factor G = 8 was selected for that block, the bit allocation for the magnitude components Smaller would be reduced to one bit, instead of zero bits, as Table I. suggests. The effect that is intended with the modification of the gain and the adjustment of the bit allocation is to preserve essentially the same signal ratio to quantization noise using a smaller number of bits. If desired, in one embodiment the selection of any gain factor that does not reduce the number of assigned bits can be avoided.

3. Control information

As explained above, the subband signal analyzer 14 provides information on control to formatter 19 for assembly in the encoded signal. That control information drives the classification for each component of a block of subband signals. This information from control can be included in the signal encoded in a diversity of forms.

One way to include control information is that of putting a bit string into the encoded signal to each block of subband signals, in which a bit corresponds to Each component of the block. A bit set to a value, in the value 1, for example, would indicate that the corresponding component does not it is a component modified by gain, and a set bit in the other value, which is the value 0 in this example, it would indicate that the corresponding component is a component modified by the profit. Another way to include control information is the of putting a special "escape code" in the encoded signal immediately before each component that is modified by the profit, or, alternatively, that is not modified by the gain.

In the preferred embodiment considered above, in which uses a quantifier with a sign of half symmetrical footprint, each component of magnitude greater than not modified by the gain is preceded by an escape code that is equal to a value Quantification not used. For example, the values of quantification for a quantifier with two complement sign 3-bit ranges from a minimum of -0.750, represented by the 3-bit b'101 binary string, up to a maximum of +0.75 represented by binary chain b'011. The b'100 binary chain, which corresponds to -1,000, is not used for quantification and is Available for use as control information. Similarly, the Binary string not used for a quantifier with sign of Two-bit complement of 4 bits is b'1000.

With reference to block 121 of subband signals of Fig. 5, components 4 and 5 are components of large magnitude exceeding threshold 102. If this threshold is used in conjunction with a gain factor G = 2, the bit allocation for all the components of small magnitude placed in the first class is b-1 , as illustrated above, in Table I. If the bit allocation process assigns b = 4 bits to each component of block 121, for example, the allocation for each component of the subband signal would be reduced to 3 = (b-1) bits, and a 3-bit quantizer would be used to quantify the components of small magnitude. Each component of large magnitude, which in this example are components 4 and 5, would be quantified with a 4-bit quantifier and identified by control information that is equal to the unused binary string of the 3-bit quantizer, or b ' 100 This control information for each large magnitude component can be conveniently assembled into the encoded signal that immediately precedes the respective large magnitude component.

It may be instructive to note that the present invention does not provide any benefit in the example considered in the preceding paragraph. The cost or "overhead" that is require to conduct the control information, which in this example is six bits, is equal to the number of bits that are economize by reducing the bit allocation for the components of small magnitude With reference to the previous example, if only a component of block 121 was a component of magnitude large, the present invention would reduce the number of bits by four required to drive that block. Seven bits would be saved through assignments reduced to seven components of magnitude small, and only three bits would be required to drive the control information for the large magnitude component.

In this last example one aspect is ignored additional. Two bits are required for each signal block of subband in this embodiment that serves as an example, to drive the which uses four gain factors for that block. How has it mentioned above, a profit factor can be used equal to 1 to indicate that the characteristics of the present invention They do not apply to a particular subband signal block.

The present invention does not usually provide advantage to quantify blocks of subband signals with Four or less components. In coding systems perceptuals that generate subband signals that have widths of commensurable band with critical system bandwidths human auditory, the number of components in the signal blocks Subband for low frequency subband signals is low, perhaps only one component per block, but the number of components per block of subband signals increases as the Subband frequency. As a result, in realizations preferred, the processing required to perform the characteristics of the present invention may be limited to subband signals more wide. A piece of additional control information to indicate the subband of lower frequency at which adaptive quantification is used by the profit. The encoder can select adaptively this subband, according to the characteristics of the signal of entry. With this technique the need to provide control information for subband signals that do not use adaptive quantification by profit.

4. Decoder features

A decoder that incorporates the features of the present invention may change adaptively the size of the quantification step of your quantifiers, essentially in any way. For example, a decoder that is intended to decode an encoded signal generated by Encoder embodiments considered above, you can use Adaptive bit allocation to set the step size of quantification. The decoder can operate in a so-called forward adaptable system, in which you can get the bit allocation directly from the encoded signal, can operate in a system called backward adaptable, in which the bit assignments are obtained by repeating the same assignment process that was used in the encoder, or it can operate in a hybrid system of those two systems. The values of the assignments obtained in this way are called assignments of "conventional" bits.

The decoder obtains control information of the encoded signal to identify gain factors and the classification of the components in each block of subband signals. Continuing with the example considered above, the control information that drives a gain factor G = 1 indicates that the adaptive gain characteristic has not been used and that the conventional bit allocation b should be used to quantify the components in that particular subband signal block. For other gain factor values, the conventional bit allocation b for a block is used to determine the value of the "escape code" or control information that identifies the components of large magnitude. In the example cited above, an assignment of b = 4 with a gain factor G = 2 indicates that the control information is the binary string b'100, which has a length equal to 3 = (b-1) bits. The presence of that control information in the encoded signal indicates that it immediately follows a large magnitude component.

The bit allocation for each component modified by profit is adjusted as seen in what above and illustrated in Table I. The quantification is carried out using the size of the appropriate quantification step, and the components modified by the gain are subjected to a factor of profit that is the reciprocal of the profit factor used for carry out the modification of the gain in the encoder. By example, if the components of small magnitude were multiplied by a gain factor G = 2 in the encoder, the decoder apply a reciprocal gain G = 0.5 to the corresponding unquantified components.

C. Additional features

In addition to the variants considered above, Consider several alternatives below.

I. Additional Classifications

According to an alternative, the magnitudes of the components of a subband signal block with two or more thresholds, and they are put in more than two classes. With reference to Fig. 5, for example, the magnitude could be compared of each component of block 121 with thresholds 102 and 104 and set it in one of three classes. Profit factor could be obtained for two of the classes and apply them to the appropriate components. For example, a gain factor G = 4 could be applied to components that were of magnitudes equal to or less than the threshold 104, and a gain factor G = 2 could be applied to the components that were of a magnitude equal to or less than threshold 102, but greater than threshold 104. Alternatively, a gain factor G = 2 to all components of equal magnitudes or less than threshold 102, and a factor could be applied again of gain G = 2 to components of equal or lesser magnitudes than threshold 104.

2. Cascade operation

The profit modification process described above can be carried out multiple times before quantification. Fig. 6 is a block diagram that illustrates an embodiment of two stages of cascade gain. In this embodiment, the subband signal analyzer 61 compares the magnitudes of the components in a signal block of subband with a first threshold, and puts the components in one of Between two classes The profit element 62 applies a first factor of gain to the component placed in one of the classes. The value of the first gain factor is related to the value of First threshold

Subband signal analyzer 64 compares the magnitudes of the components modified by the gain, and possibly of the remaining components in the block, with a second threshold, and puts the components in one of two classes. The profit element 65 applies a second gain factor to the components placed in one of the classes. The value of the second Gain factor is related to the value of the second threshold. If the second threshold is equal to or less than the first threshold, the 64 subband signal analyzer does not need to analyze the components that analyzer 61 put in the class for magnitudes greater than the first threshold.

The components of the subband signal block are quantified by quantifiers 67 and 68 in a way similar to the one considered above.

With reference to Fig. 5, the components in the block 124 of subband signals can be obtained by the successive application of profit stages, in which a subband signal analyzer 61 and gain element 62 apply a gain factor G = 2 to components that have a magnitude equal to or less than threshold 102, and analyzer 64 of Subband signals and gain element 65 apply a factor of gain G = 2 to the components modified by the gain whose magnitude that is still equal to or less than threshold 102. By For example, components 1 to 3 and 6 to 8 in block 121 are modified by a gain factor G = 2 in the first stage, what which produces a provisional result that has been illustrated in the block 122. Components 1, 3, 7 and 8 are modified by a gain factor G = 2 in the second stage, to obtain the result illustrated for block 124.

In embodiments in which stages of use are used cascade gain, control information must be provided suitable in the encoded signal, so that the decoder can carry out a complementary set of profit stages in waterfall.

3. Optimized bit allocation

There are several possible strategies to apply adaptive quantification by profit. According to a simple strategy the components of a block of signals are analyzed respective subband starting from a first threshold and from the first associated gain factor G = 2 and determine if the adaptive quantification by profit according to the first threshold and the first gain factor produces a reduction in bit allocation requirements. If you do not, the analysis and adaptive quantification is not performed by the gain. If it produces a reduction, the analysis continues with a second threshold and a second associated gain factor G = 4. If he use of the second threshold and the associated gain factor does not produce a reduction in bit allocation, the adaptive quantification by profit using the first threshold and the first gain factor. If the use of the second threshold and the second profit factor produces a reduction, the analysis with a third threshold and a third gain factor associated G = 8. This process is continued until either the use of a threshold and an associated gain factor does not produce a reduction in bit allocation, or until they have been considered all combinations of thresholds and gain factors Associates

According to another strategy, it seeks to optimize the choice of profit factor by calculating the cost and benefit provided by each possible threshold and profit factor associated, and using the threshold and the profit factor they produce The biggest net profit. For the example considered in what above, the net benefit for a threshold and a profit factor Particular associate is the gross profit minus the cost. The gross profit is the number of bits that are saved by reducing the bit allocation for small magnitude components that are modified by profit. The cost is the number of bits that are require to conduct control information for components of large magnitude that are not modified by the gain.

One way in which this can be implemented preferred strategy has been represented in the following fragment of Program. This program fragment has been expressed in pseudo code, using a syntax that includes some features syntax of the programming languages C, FRTRAN and BASIC. East program fragment and the other programs represented here not are intended to be appropriate source code segments for compilation, but have been provided to drive some few aspects of possible realizations.

one

two

The Gain function is provided with a matrix X of subband signal block components, the number N of block components, and the conventional bit allocation b for the component block. In the first expression of the function, a calculation is used according to the expression (5), specified above, to initialize the Th2 variable to represent the threshold that is associated with a gain factor G = 2 that has been obtained of a gf matrix. In this example, the gain factors gf [1], gf [2] and gf [3] are equal to G = 2, 4 and 8, respectively. The following expressions initialize the variables for thresholds that are related to gain factors G = 4 and 8. The counters are then initialized to zero, which will be used to determine the number of large magnitude components in several classes.

The expressions contained in the loop invoke the Abs function to obtain the magnitude for each component of the block of subband signals in matrix X and then compare the magnitude of the component with the thresholds, starting from the highest threshold. Yes the magnitude is greater than the Th2 threshold, for example, it increases in one the variable n2. When the loop is over, the variable n2 contains the number of components whose magnitude is greater than the Th2 threshold, variable n4 contains the number of components whose magnitude is greater than the Th4 threshold but equal to or less than the threshold Th2, and variable n8 contains the number of components whose magnitude is greater than the Th8 threshold but equal to or less than the threshold Th4.

The two expressions that follow immediately in the loop calculate the total number of components that are per above the respective thresholds. The number in variable n24 represents the number of components whose magnitude is greater than the Th4 threshold, and the number in variable n248 represents the number of components whose magnitude is greater than the Th8 threshold.

The following three expressions calculate the benefit per component of small magnitude by the use of each profit factor. This benefit can be up to 1, 2 or 3 bits per component, as illustrated above in Table I, but the benefit is also limited to no more than b-1 per component, since the Assignment to each component is limited to a minimum of one bit. For example, the number in the benefit2 variable represents the number of bits per component of small magnitude that are saved by the use of a gain factor G = 2. As illustrated in Table I, that benefit can be up to one bit; however, the benefit is also limited to not being greater than the conventional bit allocation b minus one. The calculation of this benefit is obtained by using the Min function to return to the minimum of the two values b-1 and 1.

The net benefits are then calculated and assigned to the elements of the net matrix. The net element [0] represents the net benefit of not using adaptive quantification by profit, which is zero. The net benefit of using a gain factor G = 2 is assigned to the net [1] by multiplying the appropriate benefit by components of small magnitude benefit2 by the appropriate number of components of small magnitude (N-n2) and then subtracting the cost, what is the number of components n2 of large magnitude multiplied by the length of the unused value of the quantizer used for the control information. This length is the bit length of the small magnitude components, which can be obtained from the reduced conventional bit allocation b in the bits saved by small magnitude components. For example, the bit length of small magnitude components when the gain factor G = 2 is the quantity (b-benefit2). Similar calculations are made to assign net profit by using gain factors G = 4 and 8 to net [2] and net [3] variables, respectively.

The IndexMax function is invoked to obtain the index j of the matrix for the maximum net benefit in the net matrix. This index is used to obtain the appropriate gain factor of the gf matrix, which is returned by the Gain function.

4. Improved performance using the threshold function simplified

It was mentioned in the foregoing that you can incorporate several features of the present invention in a process of perceptual bit allocation, as illustrated in Fig. 3. In particular, these features can be activated in step 53. Step 53 is performed within a loop that determines repeatedly a proposed bit allocation to quantify components in each block of subband signals to be encoded. Due to this, the performance of the operations carried out in the Step 53 is very important.

The process considered above for the Gain function, which determines the optimum gain factor for each block, is relatively inefficient, because it must count the number of components of the subband signal block that are placed in the various classes . Component counts should be calculated during each iteration, because the thresholds obtained according to expression (5) cannot be calculated until the proposed bit allocation b for each iteration is known.

In contrast to the thresholds obtained according to the expression (5), the thresholds obtained according to the expression (1) are less accurate, but can be calculated before the proposed bit allocation b is known. This allows thresholds and component counts to be calculated out of repetition. With reference to the method illustrated in Fig. 3, thresholds Th1, Th2 and Th3 and component counts n2, n24 and n248 can be calculated in step 52, for example.

In the program fragment that follows, an alternative version of the Gain function considered above has been represented, which can be used in this embodiment.

3

The expressions in the Gain 2 function are identical to the corresponding expressions in the Gain function considered above, which calculates the net benefits for each profit factor and then selects the optimal profit factor.

5. Quantification functions a) Split interval functions

The accuracy of the quantification of large magnitude components using for it a function of quantifying split interval that quantify the input values within two intervals not contiguous

Line 105 in Fig. 7 is an illustration graph of a function that represents the end-to-end effect of a 3-bit symmetric half finger sign quantifier and a complementary decoder. The values along the x axis represent input values to the quantifier, and the values at axis length qx represent the corresponding output values obtained from the quantifier. Lines 100 and 109 represent the limits of positive and negative overload values, respectively, for this quantifier. Lines 102 and 108 represent the positive and negative thresholds, respectively, for the gain factor G = 2, according to the expression (1) and how illustrated in Fig. 4. Lines 104 and 107 represent the positive and negative thresholds, respectively, for the factor of gain G = 4.

With reference to Fig. 1, if the analyzer 14 of subband signals classifies the components of the block of Subband signals according to threshold 102, it is then known that the magnitudes of the components provided to the Quantifier 18 are all greater than threshold 102. In others words, quantifier 18 would not be used to quantify any value that falls between thresholds 108 and 102. This vacuum represents an underutilization of the quantifier.

This underutilization can be counteracted using a quantifier that performs a quantification function of split interval. A variety of functions of split interval. Fig. 8 is a graphic illustration of a function that represents the end-to-end effect of a 3-bit signed quantizer of split interval and a complementary quantifier. Line 101 represents the function for positive quantities, and line 106 represents the function for negative amounts

The function illustrated in Fig. 8 has eight quantification levels, in contrast to the function illustrated in Fig. 7, which has only seven levels of quantification. The level of additional quantification is obtained using the level previously considered that, for a footprint quantification function average, corresponds to 1.

b) Non-overload quantizers

A 3-bit quantifier and a complementary quantifier that perform the function illustrated in Fig. 8 to quantify values within an interval match, from -1.0 to about -0.5 and from approximately +0.5 to +1.0, because the quantifier does not It can be overloaded. As explained above, a overload value to a quantifier if the quantization error of This value exceeds half the size of the quantization step. In the example illustrated in Fig. 8, the outputs of the Quantifier are defined for values equal to -0.9375, -0.8125, -0.6875, -0.5625, +0.5625, +0.6875, + 0.8125 and +0.9375, and The size of the quantization step is equal to 0.125. Magnitude of the quantification error for all values within the match interval mentioned above is not greater than 0.0625, which is equal to half the size of the quantification step. Such quantifier is referred to here as a "quantifier no of overload ", since it is immune to overload.

Non-single overload and split-interval quantifiers can be performed for essentially any size of the quantization step, thereby activating a quantization function that has quantizer outputs that are limited by quantifier "decision points" appropriately spaced within the intervals. of values to be quantified. Generally speaking, the decision points are spaced from each other at a certain distance d and the decision points that are closest to a respective end of a range of input values are spaced from the respective end by the amount d . This spacing provides a quantifier that, when used with a complementary quantifier, provides quantified output values, uniformly spaced apart from each other by a particular quantization step size and having a maximum quantization error that is equal to half of that size of the particular quantification step.

c) Planning functions

An interval quantifier can be performed match in a variety of ways. It is not critical any particular realization An embodiment, illustrated in Fig. 9A, comprises planning transformation 72 cascade with the quantifier 74. Planning transformation 72 receives input values on path 71, plan those values for input at an appropriate interval, and pass the planned values to along path 73, to quantifier 74.

If quantifier 74 is a quantifier with half asymmetric footprint sign, then the function of planning represented by lines 80 and 81 illustrated in the Fig. 9B would be suitable for planning function 72. From according to this planning function, the values that are within the range from -1.0 to -0.5 are planned linearly in an interval ranging from -1.0 - 1/2 \ DeltaQ to -1/2 \ DeltaQ, where \ DeltaQ is the step size of quantification of quantifier 74, and the values within the range from +0.5 to +1.0 are planned linearly in a range from -1/2 \ DeltaQ to + 1,0-½ \ DeltaQ. In this For example, no component of large magnitude can have a value that is exactly equal to -0.5 or +0.5, because the components that have those values are classified as Small magnitude components. Because of this, the transformation of planning 72 will not plan any input value to -½? Exactly; however, you can plan values of entry arbitrarily close to, and on either side of -½ \ DeltaQ.

The effect of this planning can be seen if refers to Figs. 9B and 9C. With reference to Fig. 9B, it can be seen in it that the planning transformation 72 transforms entry points 82 and 84 into planned points 86 and 88, respectively. With reference to Fig. 9C, which illustrates a function that represents the end-to-end effects of a quantifier with sign of half asymmetric footprint of 3 bits and a complementary quantifier, it can be seen that the points planned 86 and 88 are on either side of point 87 of quantifier decision, which has the value ½ \ DeltaQ.

You can perform a quantifier of complementary match interval using a quantifier with sign of half asymmetric footprint, which is complementary to the quantifier 74 followed by a planning transformation that it is the inverse of the planning transformation 72.

d) Composite functions

In an example considered above, adaptive quantification by gain with a gain factor G = 2 is used to quantify components of a subband signal for which the conventional bit allocation b is equal to three bits. As explained above in relation to Table I, 3 bits are used to quantify the large magnitude components, and 2 = (b-1) bits are used to quantify the modified components by the small magnitude gain. Preferably, a quantifier is used that performs the quantification function of Fig. 8 to quantify the components of large magnitude.

A quantifier with a sign of 2-bit symmetric half fingerprint and a quantifier complementary to perform the function 111 illustrated in Fig. 10, for the components modified by the small magnitude gain. Function 111, as illustrated, takes into account the effects of scaling and scaling of the gain factor G = 2 used together with the quantifier and the quantifier, respectively. The output values for the quantifier are -0.3333 ..., 0.0 and +0.3333 ..., and the decision points of the Quantifier are at -0.1666 ... and +0.1666 ....

In Fig. 11 a compound of the functions for the magnitude components of large and magnitude little.

e) Alternative split interval functions

The use of a split interval quantifier with a gain factor G = 2 and a threshold at, or about, 0.500, provides an improvement in the resolution of the quantification of approximately one bit. This enhanced resolution can be used to preserve the quantification resolution of the components of large magnitude, while reducing bit allocation to those components in a bit. In the example considered above, they could 2-bit quantifiers used to quantify the components, both those of large magnitude and those of small magnitude. In the Fig. 12 a compound of the functions of quantification performed by the two quantifiers. Could use quantifiers that perform quantification functions 112 and 113 to quantify the components of large magnitude that have positive and negative amplitudes, respectively, and could use a quantifier that performs the quantification function 111 to quantify the components of small magnitude.

The use of split interval quantification functions with higher gain factors and lower thresholds does not provide a full bit of improved quantization resolution; therefore, the bit allocation cannot be reduced without sacrificing the resolution of the quantization. In preferred embodiments, the allocation b of bits for mantissa of large magnitude is reduced by one bit for blocks that are quantitatively quantified by gain, using a gain factor G = 2.

The decuantification function provided in the decoder must be complementary to the function of quantification used in the encoder.

6. Intra-frame coding

The denomination of "block of signals encoded "is used here to refer to the information encoded representing all subband signal blocks for frequency subbands across bandwidth Useful of the input signal. Some coding systems assemble multiple blocks of coded signals into units older, what we call here as "a framing" of the coded signal A framing structure is useful in many applications to share information through blocks of coded signals, thereby reducing "expenses general information ", or to facilitate signals from synchronization such as audio and video signals. They have considered a diversity of issues that involve coding of audio information in frames for applications audio / video in U.S. patent application Serial number PCT / US 98 20751, filed on October 17, 1998, which It is here incorporated by reference.

The characteristics of quantification adaptable by profit, previously considered, can be applied to groups of subband signal blocks that are in different coded signal blocks. You can use this aspect advantageously in applications that, for example, group in frames coded signal blocks. This technique groups essentially within a frame the components of multiple subband signal blocks, and then classify the components and apply a gain factor to that group of components, as has been described above. This so-called coding technique intra framing can share control information between blocks within a frame. For the practice of this technique no no particular grouping of signal blocks is critical coded

D. Realization

The present invention can be performed in a great diversity of forms, including software in a system of general purpose computer or some other device that includes more specialized components, such as processor circuits of digital signals (DSP) coupled to components, similar to which are in a general purpose computer system. Fig. 13 is a block diagram of device 90 that can be used to implement various aspects of the present invention. The DSP 92 provides calculation resources. RAM 93 is a memory Direct access (RAM) system. ROM 94 represents a form of persistent storage, such as a memory read-only (ROM) to store the programs that are needed to operate the device 90 and to implement several Aspects of the present invention. The 95 I / O control (Input / Output) represents interface circuits to receive and transmit signals audio via communication channel 96. In control 95 I / O analog-to-digital converters can be included and digital to analog converters, as desired, to receive and / or transmit analog audio signals. In the realization illustrated, all major system components connect with bus 91, which can represent more than a physical bus; without However, a bus architecture is not required to perform the present invention

In embodiments implemented in a general purpose computer system, can be included additional components for interface with devices such as a keyboard or mouse, and a presentation, and to control a storage device that has a storage medium such as a magnetic tape or a magnetic disk, or a medium optical. The storage medium can be used to record instruction programs for operating systems, services and applications, and may include realizations of programs that carry to practice several aspects of the present invention.

The functions required for commissioning practice of various aspects of the present invention can be carried out through components that are materialized in a great diversity in ways, including those of discrete logical components, some or more ASIC (Integrated Circuit for Specific Applications) and / or program controlled processors. The way it is materialize those components is not important for the present invention.

Software realizations of the present invention can be conducted by a variety of readable means with machine, such as baseband communications paths or modulated across the spectrum, including from frequencies supersonic to ultraviolet, or storage media including those who conduct information using essentially any magnetic or optical recording technology, including magnetic tapes, magnetic discs and optical discs. Too various aspects can be implemented in various components of computer system 90 by means of processing circuits such such as ASICs, integrated circuits for general purposes, microprocessors controlled by programs made in various forms of read-only memory (ROM) or RAM, and others techniques

Claims (29)

1. A method to encode an input signal which includes:

receive the input signal and generate a block of subband signals from signal components of subband representing a frequency subband of the signal of entry;

compare the magnitudes of the components of the subband signal block with a threshold, put each component in one of two or more classes, according to the magnitude of the component, and obtain a profit factor;

apply the gain factor to the components placed in one of the classes to modify the magnitudes of some of the components in the signal block of subband;

quantify the components in the subband block; Y

assemble into a signal encoded the control information that drives the classification of components and symbols of non-uniform length that represent the components of the subband signal quantified

2. A method according to claim 1, by which the control information is assembled in the signal encoded indicating those components of the subband signal quantified that have magnitudes that are not modified from according to the gain factor, in which the control information is driven by one or more reserved symbols that are not used to represent components of the subband signal quantified 3. A method according to claim 1 or 2, which comprises obtaining the threshold of a function that is dependent of the gain factor but independent of the step size of quantification of the quantified components. 4. A method according to claim 1 or 2, which comprises obtaining the threshold of a function that is dependent of the gain factor and the size of the quantification step of the quantified components. 5. A method according to any one of the claims 1 to 4, comprising:

change for adaptation a size of the respective quantification step for each component of the subband signal block according to the class in which the component is set, assigning for this by adaptation bits to the component, and

get the gain factor such that the number of bits assigned to the components with modified quantities be reduced, while preserving the respective size of the quantification step.

6. A method according to any one of the claims 1 to 5, comprising quantifying the components placed in one of the classes according to a function of quantization of split interval. 7. A method according to any one of the claims 1 to 6, which puts each component in one of three or more classes, according to the magnitude of the component, and which includes:

get one or more factors of additional gain, each of them associated with a class respective; Y

apply each of the additional gain factors to the components placed in the respective associated class.

8. A method according to any one of the claims 1 to 7, comprising:

compare the magnitudes of al minus some of the subband signal block components with a second threshold, put each component in one of two or more second classes according to the magnitude of the component, and get a second profit factor; Y

apply the second factor of gain to the components placed in one of the second classes to modify the magnitudes of some of the components of the subband signal block;

in which the symbols of non-uniform length they represent the quantified components, as modified by the profit factor and by the second factor of gain. 9. A method according to any one of the claims 1 to 8, which quantifies at least some of the components using one or more non-overload quantifiers. 10. A method to decode a signal coded, comprising:

receive the encoded signal and get the same control information and length symbols not uniform, and get from the symbols of non-uniform length quantified subband signal components that represent a frequency subband of an input signal;

unquantify the components of the subband signal to obtain unquantified components of the subband signal;

apply a gain factor to modify the magnitudes of some of the components unquantified, according to the control information; Y

generate an output signal in response to the unquantified components of the signal Subband

11. A method according to claim 10, by which signal control information is obtained encoded indicating those components of the subband signal quantified that have magnitudes that are not to be modified from according to the gain factor, in which the control information It is driven by one or more reserved symbols, which are not used to represent components of the subband signal quantified 12. A method according to claim 10 or 11, which includes quantifying some of the components quantified in the subband signal block according to a decuantification function that is complementary to a function of quantization of split interval. 13. A method according to any one of claims 10 to 12, which comprises applying a second gain factor to modify the magnitudes of some of the unquantified components according to the information in control. 14. A method according to any one of claims 10 to 13, wherein they are at least quantified some of the components quantified using one or more quantifiers that are complementary to a quantifier not of respective overload. 15. An apparatus for encoding a signal of entry, comprising:

an analysis filter (12) that it has an input (11) that receives the input signal and that has an outlet (13, 23) through which a block of subband signals of components of the subband signal that represent a subband of the signal's frequency entry;

an analyzer (12, 24; 61) of the block of subband signals coupled to the analysis filter (12), which compares the magnitudes of the components in the block of Subband signals with a threshold, puts each component in one of between two or more classes, according to the magnitude of the component, and get a profit factor;

a processor (15, 25; 62) of components of the subband signal coupled to the analyzer (14, 24; 61) of the subband signal block that applies the gain to the components placed in one of the classes, to modify the magnitudes of some of the components of the block subband signals;

a first quantifier (17, 27; 67) coupled to the subband signal processor that quantifies the components of the subband signal block that have the magnitudes modified according to the gain factor; Y

a formatter (19) coupled to the first quantifier (17, 27; 67) that assembles symbols of non-uniform length representing the signal components of quantified subband, and control information that drives the classification of the components in an encoded signal.

16. An apparatus according to claim 15, which comprises a second quantifier (18, 28; 68) coupled to the analyzer (14, 24; 61) of the subband signal block, which quantify the components placed in one of the classes, according with a quantification function of split interval, in which the formatter (19) is also coupled to the second quantizer (18, 28; 68). 17. An apparatus according to claim 15 or 16, in which the formatter (19) assembles information from control in the encoded signal, which indicates those components of the quantified subband signal of magnitudes that are not modified according to the profit factor, in which the control information is driven by one or more symbols reserved, which are not used to represent components of the quantified subband signal. 18. An apparatus according to any one of claims 15 to 17, which obtains the threshold of a function which is dependent on the profit factor but independent of size of the quantification step of the components quantified 19. An apparatus according to any one of claims 15 to 17, which obtains the threshold of a function which is dependent on the gain factor and step size of quantification of the quantified components. 20. An apparatus according to any one of claims 15 to 19, which changes by adaptation a size of the respective quantification step for each component of the subband signal block, according to the class in which the component is put through the allocation for adaptation of bits to the component, and get the gain factor such that the number of bits assigned to the modified magnitude components be reduced, while preserving the respective size of the quantification step 21. An apparatus according to any one of claims 15 to 20, which puts each component in one of between three or more classes, according to the magnitude of the component, get one or more additional profit factors, each one associated to a respective class, and applies each of the additional gain factors to the components placed in the respective associated class. 22. An apparatus according to any one of claims 15 to 1, wherein:

the analyzer (14, 24; 64) of the subband signal block compares the magnitudes of at least some of the components of the subband signal block with a second threshold, puts each component in one of two or more second classes, according to the magnitude of the component, and get a second profit factor; Y

the processor (15, 25; 65) of components of the subband signal apply the second factor of gain to the components placed in one of the second classes, to modify the magnitudes of some of the components of the subband signal block;

in which the symbols of non-uniform length they represent the quantified components, as modified by the profit factor and by the second factor of gain. 23. An apparatus according to any one of claims 15 to 22, which quantifies at least some of the components using one or more non-overload quantifiers. 24. An apparatus for decoding a signal coded, comprising:

a deformer (32) that receive the encoded signal and get the same information from control and symbols of non-uniform length, and obtains from non-uniform length symbols subband signal components quantified;

a first quantifier (33, 43) coupled to the deformer (32), which quantifies some of the subband signal components of the block, according with the control information, to get first components unquantified;

a processor (35, 45) of block of subband signals coupled to the first quantifier (33, 43) that applies a profit factor to modify the magnitudes of some of the first unquantified components of the subband signal block, according to the information in control; Y

a synthesis filter (39) that it has an input (38, 48) coupled to the signal processor of subband and that has an exit (40) through which it Provides an output signal.

25. An apparatus according to claim 24, comprising a second quantifier (34, 44) coupled to the deformer (32) that decrypts other components of the signal block subband, according to a function of quantification that is complementary to a function of quantization of split interval, to obtain second unquantified components, and in which the synthesis filter (39) it has an input (38, 48) coupled to the second quantifier (34, 44). 26. An apparatus according to claim 24 or 25, in which the deformer (32) obtains information from control of the coded signal indicating those components of the quantified subband signal that have magnitudes that are not to be modified according to the profit factor, in which the control information is driven by one or more symbols reserved, which are not used to represent components of the quantified subband signal. 27. An apparatus according to any one of claims 24 to 26, wherein the processor (35, 45) of the Subband signal block applies a second gain factor to modify the magnitudes of some of the components Unquantified, according to the control information. 28. An apparatus according to any one of claims 24 to 27, which quantifies at least some of quantified components using one more quantifiers than are complementary to a non-overload quantifier respective. 29. A computer program product made in a machine-readable medium, said said Computer program product program instructions executable by said machine to implement such a method as defined in any one of claims 1 to 14.