KR100947065B1 - Lossless audio decoding/encoding method and apparatus - Google Patents

Abstract

A method and apparatus for lossless audio encoding / decoding are disclosed. The encoding method includes a signal conversion step of converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; A mapping step of mapping an audio signal of a frequency domain to a bitplane signal according to frequency; And a lossless encoding step of lossless encoding on the binary sample constituting the bitplane through a probability model determined using a predetermined context. The decoding method includes demultiplexing an audio bit stream to extract a predetermined loss coded loss bit stream and an error bit stream of the error data; (bb) predetermined loss decoding the extracted loss bitstream; Lossless decoding the extracted error bitstream; Restoring the frequency spectral signal using the decoded lost bitstream and the error bitstream; And restoring an audio signal in a time domain by performing inverse integer time / frequency conversion on the frequency spectral signal.

According to the present invention, a model made through a statistical distribution is provided regardless of the input distribution, and a compression rate superior to BPGC is provided through a context-based encoding method.

Description

Lossless audio decoding / encoding method and apparatus

1 is a block diagram illustrating a configuration of an embodiment of a lossless audio encoding apparatus according to the present invention.

2 is a block diagram illustrating a configuration of a lossless encoding unit.

3 is a block diagram showing the configuration of another embodiment of a lossless audio encoding apparatus according to the present invention.

4 is a block diagram illustrating the configuration of the lossless encoding unit 340.

FIG. 5 is a flowchart illustrating an operation of an embodiment of a lossless audio encoding apparatus according to the present invention shown in FIG. 1.

6 is a flowchart illustrating an operation of the lossless encoding unit illustrated in FIG. 1.

FIG. 7 is a flowchart illustrating an operation of another embodiment of the lossless audio encoding apparatus according to the present invention shown in FIG. 3.

8 shows an example of mapping to bitplane data according to frequency.

9 is a block diagram illustrating a configuration of an embodiment of a lossless audio decoding apparatus according to the present invention.

FIG. 10 is a block diagram illustrating a configuration of the context calculator shown in FIG. 9.

11 is a block diagram illustrating a configuration of another embodiment of a lossless audio decoding apparatus according to the present invention.

12 is a block diagram showing the configuration of a lossless decoding unit.

FIG. 13 is a flowchart illustrating an operation of an embodiment of a lossless audio decoding apparatus according to the present invention shown in FIG. 9.

FIG. 14 is a flowchart illustrating an operation of another embodiment of the lossless audio decoding apparatus according to the present invention shown in FIG. 11.

The present invention relates to audio signal encoding / decoding, and more particularly, to a lossless audio encoding / decoding method and apparatus capable of adjusting a bit rate.

Lossless audio coding methods include MLP (Meridian Lossless Audio Compression), Monkey's Audio, Free Lossless Audio Coding, and MLP is applied to DVD-A. As the network bandwidth of the Internet increases, a large amount of multimedia content is provided, and audio requires a lossless audio method. In the EU, digital audio broadcasting has already begun through DAB (Digital Audio Broadcasting), and broadcasting stations and content providers are using audio lossless coding. In the MPEG group, standardization of lossless audio compression is underway under the name of ISO / IEC 14496-3: 2001 / AMD 5, Audio Scalable to Lossless Coding (SLS). It provides Fine Grain Scalablity (FGS) and allows lossless audio compression.

Compression rate, the most important factor in lossless audio compression technology, can be improved by eliminating redundant information between data. The duplicated information may be removed through prediction between adjacent data or may be removed through a context between adjacent data.

Integer MDCT coefficients show a Laplacian distribution, in which the best known compression method, called the Golomb code, is used. Bitplane coding is required to provide FGS. The combination of Golomb code and bitplane coding is called Bit Plane Golomb Coding (BPGC), which provides an optimal compression rate and FGS. However, the assumption that the integer MDCT coefficients have a Laplacian distribution may not be appropriate for the actual data distribution. Since the BPGC is an algorithm designed on the assumption that the integer MDCT coefficients have a Laplacian distribution, the BPGC cannot provide an optimal compression ratio when the integer MDCT coefficients do not have a Laplacian distribution. Therefore, there is a need for a lossless audio coding and decoding method that can provide an optimal compression ratio regardless of the assumption that the integer MDCT coefficients have a Laplacian distribution.

An object of the present invention is to provide a lossless audio encoding method and apparatus capable of providing an optimal compression ratio regardless of the assumption that the integer MDCT coefficients have a Laplacian distribution.

Another object of the present invention is to provide a lossless audio decoding method and apparatus capable of providing an optimal compression ratio regardless of the assumption that the integer MDCT coefficients have a Laplacian distribution.

According to an aspect of the present invention, there is provided a lossless audio encoding method comprising: a signal conversion step of converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; A mapping step of mapping the audio signal in the frequency domain into a bitplane signal according to frequency; And a lossless encoding step of lossless encoding on the binary sample constituting the bitplane through a probability model determined using a predetermined context. The lossless encoding may include mapping an audio signal in the frequency domain to bitplane data according to frequency; Obtaining a most significant bit and a Golomb parameter for each bit plane; Selecting a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from low frequency component to high frequency; For the selected binary sample, calculating a context of the selected binary sample using previous samples on the same bitplane; Selecting a probabilistic model using the obtained Golomb parameter and the calculated contexts; And lossless encoding the binary sample using the selected probability model.

According to an aspect of the present invention, there is provided a lossless audio encoding method comprising: (a) converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; (b) matching the audio spectral signal in the frequency domain with an input signal of a lossy encoder through scaling; (c) lossy compression coding the scaled frequency signal; (d) obtaining an error mapped signal corresponding to a difference between the lossy coded data and an audio spectral signal in the frequency domain having the integer value; (e) lossless encoding the error mapped signal using a context; And (f) multiplexing the lossless coded signal and the lossy coded signal into a bitstream. The step (e) may include (e1) mapping the error mapped signal of the step (d) to bitplane data according to frequency; (e2) obtaining the most significant bit and the Golomb parameter of the bitplane; (e3) selecting a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from low frequency component to high frequency; (e4) for the selected binary sample, calculating the context of the selected binary sample using previous samples on the same bitplane; (e5) selecting a probabilistic model using the obtained Golomb parameter and the calculated contexts; And (e6) lossless encoding the binary sample using the selected probability model.

In the step (e4), it is preferable to calculate the context value of the binary sample by making the encoded previous samples on the same bitplane into one scalar value with respect to the selected binary sample. Also, in the step (e4), a probability of '1' is obtained for predetermined samples existing in the same bit plane for the selected binary sample, and the integer value is multiplied by a predetermined integer value to represent an integer, and then the integer is represented. It is preferable to calculate the context value by using. Also, in the step (e4), it is preferable to calculate a context value with respect to the selected binary sample by using an already encoded upper bitplane value on the same frequency. In addition, in step (e4), a context value is calculated for the selected binary sample by using information on whether there is a value of an already encoded upper bit plane on the same frequency, but one '1' is included in the upper bit plane. Even if it is present, it is preferable to set the context value to '1'.

According to an aspect of the present invention, there is provided a lossless audio encoding apparatus comprising: an integer time / frequency conversion unit converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; And a lossless encoding unit for mapping the audio signal in the frequency domain into bitplane data according to frequency and losslessly encoding binary samples constituting the bitplane using a predetermined context. The lossless encoding unit includes: a bitplane mapping unit for mapping the audio signal in the frequency domain into bitplane data according to frequency; A parameter obtaining unit for obtaining a most significant bit and a Golomb parameter of the bit plane; A binary sample selection unit for selecting a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from low frequency component to high frequency; A context calculator for calculating the context of the selected binary sample for the selected binary sample using previous samples on the same bitplane; A probability model selecting unit which selects a probability model using the obtained Golomb parameter and the calculated contexts; And a binary sample encoder for lossless encoding the binary sample using the selected probability model. Preferably, the integer time / frequency converter is an integer MDCT.

According to an aspect of the present invention, there is provided a lossless audio encoding apparatus comprising: an integer time / frequency conversion unit converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; A scaling unit for matching an audio frequency signal of the integer time / frequency conversion unit with an input signal of a lossy coding unit; A lossy coding unit for lossy compression encoding the scaled frequency signal; An error mapping unit obtaining a difference between the lossy coded signal and the signal of the integer time / frequency converter; A lossless encoding unit for losslessly encoding the error mapped signal using a context; And a multiplexer for multiplexing the lossless coded signal and the lossy coded signal to generate a bitstream. The lossless encoding unit may include: a bitplane mapping unit configured to map the error mapped signal of the error mapping unit into bitplane data according to a frequency; A parameter obtaining unit for obtaining a most significant bit and a Golomb parameter of the bit plane; A binary sample selecting unit which selects a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from lowest frequency component to highest frequency; A context calculator for calculating the context of the selected binary sample for the selected binary sample using previous samples on the same bitplane; A probability model selecting unit which selects a probability model using the obtained Golomb parameter and the calculated contexts; And a binary sample encoder for lossless encoding the binary sample using the selected probability model.

According to another aspect of the present invention, there is provided a lossless audio decoding method comprising: obtaining a Golomb parameter from a bitstream of audio data; A sample selecting step of selecting a binary sample to be decoded in order from the most significant bit to the least significant bit, and then from a low frequency to a high frequency; A context calculation step of calculating a predetermined context using already decoded samples; A probability model selection step of selecting a probability model using the Golomb parameter and a context; An arithmetic decoding step of performing arithmetic decoding using the selected probability model; And repeating the sample selection to arithmetic decoding until all samples are decoded. The context calculating step includes calculating a first context using previously decoded samples existing on the same bitplane; And calculating a second context using samples of the upper bitplane already decoded on the same frequency.

In the lossless audio decoding method according to the present invention for achieving the above technical problem, when the difference between the loss coded audio data and the audio spectral signal in the frequency domain having an integer value is error data, (aa) an audio bitstream Demultiplexing and extracting a predetermined loss coded loss bit stream and an error bit stream of the error data; (bb) predetermined loss decoding the extracted loss bitstream; (cc) lossless decoding the extracted error bitstream; (dd) restoring a frequency spectral signal using the decoded lossy bitstream and error bitstream; And (ee) restoring the audio signal in the time domain by performing inverse integer time / frequency conversion on the frequency spectral signal. The (cc) step may include obtaining a Golomb parameter from a bitstream of (cc1) audio data; (cc2) selecting binary samples to be decoded from most significant bit to least significant bit, in order of low frequency to high frequency; (cc3) calculating a predetermined context using already decoded samples; (cc4) selecting a probabilistic model using the Golomb parameter and context; (cc5) performing arithmetic decoding using the selected probability model; And (cc6) repeating steps (cc2) to (cc5) until all samples are decoded. The step (cc3) may include calculating a first context using previously decoded samples existing on the same bitplane; And calculating a second context using samples of the upper bitplane already decoded on the same frequency.

According to another aspect of the present invention, there is provided a lossless audio decoding apparatus, including: a parameter obtaining unit configured to obtain a Golomb parameter from a bitstream of audio data; A sample selecting unit for selecting a binary sample to be decoded in order from the most significant bit to the least significant bit and then from a low frequency to a high frequency; A context calculator which calculates a predetermined context using samples that have already been decoded; A probabilistic model selection unit for selecting a probabilistic model using the Golomb parameter of the parameter obtaining unit and the context of the context calculating unit; And an arithmetic decoding unit for performing arithmetic decoding using the selected probabilistic model. The context calculator may include a first context calculator configured to calculate a first context using previously decoded samples existing on the same bit plane; And a second context calculator configured to calculate a second context using samples of the upper bitplane already decoded on the same frequency.

The lossless audio decoding apparatus according to the present invention for achieving the above another technical problem is to demultiplex an audio bitstream when a difference between lossy encoded audio data and an audio spectral signal in a frequency domain having an integer value is error data. A demultiplexer for extracting a predetermined loss coded loss bit stream and an error bit stream of the error data; A lossy decoding unit for predetermined lossy decoding of the extracted lossy bitstream; A lossless decoding unit for lossless decoding the extracted error bitstream; An audio signal synthesizing unit for synthesizing the decoded lossy bitstream and the error bitstream to restore a frequency spectral signal; And an inverse integer time / frequency converter for restoring an audio signal in a time domain by performing inverse integer time / frequency conversion on the restored frequency spectral signal.

The lossy decoding unit is preferably an AAC decoding unit. The lossless audio decoding apparatus may further include an inverse time / frequency conversion unit for restoring the audio signal in the frequency domain decoded by the lossy decoding unit to the audio signal in the time domain. The lossless decoding unit comprises: a parameter acquisition unit for obtaining a Golomb parameter from a bitstream of audio data; A sample selecting unit for selecting a binary sample to be decoded in order from the most significant bit to the least significant bit and then from a low frequency to a high frequency; A context calculator which calculates a predetermined context using samples that have already been decoded; A probability model selection unit for selecting a probability model using the Golomb parameter and a context; And an arithmetic decoding unit for performing arithmetic decoding using the selected probability model.

The context calculator may include a first context calculator configured to calculate a first context using previously decoded samples existing on the same bit plane; And a second context calculator for calculating a second context using samples of the upper bitplane already decoded on the same frequency.

A computer readable recording medium having recorded thereon a program for executing the invention described above is provided.

Hereinafter, a lossless audio encoding / decoding method and apparatus according to the present invention will be described in detail with reference to the accompanying drawings. In audio coding, Fine Grain Scalability (GFS) is provided, and an integer modified discrete cosine transform (MDCT) is used for lossless audio coding. In particular, according to the Laplacian distribution of the input sample distribution of the audio signal, the Bit Plane Golomb Coding (BPGC) method has an optimal compression result, which is equivalent to the Golomb code. The Golomb parameter can be obtained by the following procedure.

For (L = 0; (N << L + 1)) <= A; L ++);

The Golomb parameter L can be obtained by using. In the characteristics of the Golomb code, the probability of getting 0 or 1 in a bit plane smaller than L is 1/2. These results, however, are optimal for the Laplacian distribution, but cannot provide an optimal compression ratio. Therefore, the basic concept of the present invention is to provide the optimal compression ratio using the context through statistical analysis even if the data distribution does not follow the Laplacian distribution.

1 is a block diagram illustrating an embodiment of a lossless audio encoding apparatus according to the present invention, and includes an integer time / frequency conversion unit 100 and a lossless encoding unit 120. The integer time / frequency converter 100 converts an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value, and preferably uses an integer MDCT. The lossless encoding unit 120 maps an audio signal in the frequency domain to bitplane data according to frequency, and lossless encodes a binary sample constituting the bitplane using a predetermined context, and the bitplane mapping unit 200. , The parameter acquisition unit 210, the binary sample selecting unit 220, the context calculating unit 230, the probability model selecting unit 240, and the binary sample encoding unit 250.

The bitplane mapping unit 200 maps the audio signal in the frequency domain into bitplane data according to frequency. 8 shows an example of mapping to bitplane data according to frequency.

The parameter acquisition unit 210 obtains the most significant bit (MSB) and the Golomb parameter of the bit plane. The binary sample selector 220 selects a binary sample on a bitplane to be encoded in order from the most significant bit to the least significant bit (LSB) and the lowest frequency component to the highest frequency.

The context calculator 230 calculates a context of the selected binary sample using the previous samples on the same bitplane, for the selected binary sample. The probability model selector 240 selects a probability model using the obtained Golomb parameter and the calculated contexts. The binary sample encoder 250 losslessly encodes the binary sample using the selected probability model.

3 is a block diagram showing the configuration of another embodiment of a lossless audio encoding apparatus according to the present invention. The integer time / frequency conversion unit 300, the scaling unit 310, the loss encoding unit 320, and the error mapping unit ( 330, a lossless encoding unit 340, and a multiplexer 350.

The integer time / frequency converter 300 converts an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value, and preferably uses an integer MDCT. The scaling unit 310 matches the audio frequency signal of the integer time / frequency converter 300 with the input signal of the lossy encoder 320. Since the output signal of the integer time / frequency converter 300 is represented by an integer, it cannot be directly used as an input of the lossy encoder 320. Accordingly, the audio frequency signal of the integer time / frequency converter 300 is matched to be used as an input signal of the loss encoder 320 through the scaling unit 310.

The lossy coding unit 320 performs lossy compression coding on the scaled frequency signal, and preferably uses an AAC core encoder. The error mapping unit 330 obtains an error mapped signal corresponding to a difference between the loss coded signal and the signal of the integer time / frequency conversion unit 300. The lossless encoding unit 340 losslessly encodes the error mapped signal using a context. The multiplexer 350 multiplexes the lossless coded signal of the lossless encoder 340 and the lossy coded signal of the lossy encoder 320 into a bitstream.

4 is a block diagram illustrating the configuration of the lossless encoding unit 340, and includes a bitplane mapping unit 400, a parameter acquisition unit 410, a binary sample selection unit 420, a context calculation unit 430, and a probability. A model selector 440 and a binary sample coder 450 are provided.

The bitplane mapping unit 400 maps the error mapped signal of the error mapping unit to bitplane data according to frequency. The parameter acquisition unit 410 obtains the most significant bit and the Golomb parameter of the bitplane. The binary sample selector 420 selects a binary sample on a bitplane to be encoded in order from the most significant bit to the least significant bit, and then the lowest frequency component to the highest frequency. The context calculator 430 calculates the context of the selected binary sample using the previous samples on the same bitplane for the selected binary sample. The probability model selection unit 440 selects a probability model using the obtained Golomb parameter and the calculated contexts. The binary sample encoder 450 losslessly encodes the binary sample using the selected probability model.

The context calculators 230 and 430 illustrated in FIGS. 2 and 4 may calculate the context value of the binary sample by making the encoded previous samples on the same bit plane into one scalar value with respect to the selected binary sample. In addition, the context calculators 230 and 430 obtain a probability that '1' appears for predetermined samples existing in the same bit plane for the selected binary sample, and multiply the probability value by a predetermined integer value to represent an integer. After that, the context value can be calculated using the integer. The context calculating units 230 and 430 may calculate a context value with respect to the selected binary sample by using an already encoded upper bitplane value on the same frequency, and with respect to the selected binary sample, the contextual value is already encoded on the same frequency. Calculate the context value using information on whether there is a value of the upper bitplane, but if there is at least one '1' in the upper bitplane, set the context value to '1' and not at least one '1'. Otherwise, the context can be calculated with the context value '0'.

FIG. 5 is a flowchart illustrating an operation of one embodiment of the lossless audio encoding apparatus according to the present invention shown in FIG. 1, and the configuration of the lossless audio encoding apparatus will be described with reference to the flowchart. First, when a PCM signal corresponding to an audio signal in a time domain is input to the integer time / frequency converter 100, the PCM signal is converted into an audio spectral signal in a frequency domain having an integer value (step 500). It is preferable to use int MDCT here. Then, the audio signal in the frequency domain is mapped to a bitplane signal according to frequency as shown in FIG. 8 (step 520). A probability model determined by using a predetermined context for binary samples constituting the bitplane. Lossless coding through (step 540).

FIG. 6 is a flowchart illustrating an operation of the lossless encoding unit 120 illustrated in FIG. 1, and the operation thereof will be described with reference to FIG. 6.

First, when the audio signal of the frequency domain is input to the bitplane mapping unit 200, the audio signal of the frequency domain is mapped to bitplane data according to the frequency (step 600). The most significant bit and the Golomb parameter are obtained for each bitplane (step 610). The binary sample on the bitplane to be encoded in order from the most significant bit to the least significant bit, and then the lowest frequency component through the binary sample selector 220. In operation 610, the context of the binary sample selected by the binary sample selector 220 is calculated using the previous samples on the same bitplane. A probabilistic model is selected using the Golomb parameter obtained from the unit 210 and the contexts calculated by the context calculator 230. (Step 630) by using the probability model selected in the probability model selection unit 240 and a lossless coding the binary sample (step 640)

FIG. 7 is a flowchart illustrating an operation of another embodiment of the lossless audio encoding apparatus according to the present invention shown in FIG. 3, and the operation of another embodiment of the lossless audio encoding apparatus will be described with reference to the flowchart. . First, the audio signal in the time domain is converted into an audio spectral signal in the frequency domain having an integer value through the integer time / frequency converter 300 (step 710).

Then, the audio spectral signal in the frequency domain is scaled by the scaling unit 310 and matched so that it can be used as an input signal of the lossy coding unit 320 (step 720). The signal is subjected to lossy compression encoding by the lossy compression encoder 320 (step 730). The lossy compression encoding is preferably performed by an AAC Core encoder.

An error mapped signal corresponding to a difference between the data loss coded by the loss encoder 320 and the audio spectral signal in the frequency domain having the integer value is obtained by the error mapper 330 (step 740). The error mapped signal is losslessly encoded using the context in the lossless encoding unit 340 (step 750).

The lossless encoding signal in the lossless coding unit 340 and the lossless coding signal in the lossy coding unit 320 are multiplexed by the multiplexer 350 to generate a bitstream. (Step 760) The lossless coding (step 750). Maps the error-mapped signal to bitplane data according to frequency. Since the process of obtaining the most significant bit (MSB) and the Golomb parameter is the same as in FIG. 6, description thereof is omitted.

8 illustrates a range of samples in which a sample to be encoded selects a context on a bitplane. The probability distribution of the current sample is calculated from the already coded samples using the portion indicated by the dotted line.

In general, due to spectral leakage by MDCT, there is a correlation between neighboring samples on the frequency axis. That is, if the value of the adjacent sample is X, the probability that the value of the current sample is the value of X is very high. In this case, if the neighboring adjacent sample is selected as the context, the compression ratio can be improved by using the correlation.

* Also, statistical analysis showed that the upper bitplane value has a high correlation in the probability distribution of lower samples. If the neighboring adjacent samples are selected as the context, the correlation can be improved by using correlation.

The method of calculating the context is as follows. First, the context is calculated using previously coded samples that exist on the same bitplane. There may be a number of methods in the above method, and representative methods are as follows.

The first method is to make binary scalar values earlier than a predetermined length on the same bitplane and use them as context. For example, if four previous binary samples are to be used as the context, if the previous samples are 0100, this is regarded as binary and 4 is set as 4 with 0100 (2) as 4. In general, if 0001 (2), the probability that the current sample is '1' is the highest. In some cases, the scope of the context is limited in consideration of the size of the model. Usually the context is in the range of 8-16.

The second method is a method of calculating a probability that the sample is '1' for the encoded samples before the sample to be counted by counting '1' existing on the same bitplane. In order to express this in context, the probability value of '1' is multiplied by an integer N and expressed as an integer. If this value is 0, then there is no '1' in the previous sample on the same bitplane, so the probability that the current sample is '1' is very high. Since there are many ',' the probability of '0' is high. In some cases, the scope of the context is limited in consideration of the size of the model. Usually the context is in the range of 8-16.

The next way to calculate the context is to compute the context using samples of the upper bitplane on the same frequency. There may be various methods in the above method, and representative methods are as follows.

The first method uses, as context, the value of a higher bitplane that has already been encoded. For example, if the upper bitplane samples are 0110, it is calculated as binary and expressed in the context of 0110 (2), that is, 6. In some cases, the scope of the context is limited in consideration of the size of the model. Usually the context is in the range of 8-16.

The second method uses information about whether there is a value of a higher bitplane that is already encoded. If there is any '1' in the upper bitplane on the same frequency, the context value is '1'. If there is no '1', the context value is '0'. This means that if the most significant bit (MSB) has not been encoded yet, the probability that the current sample is '1' is very high.

As an example of the actual encoding, assume that the fourth sample of the third bitplane is encoded. The current sample to be encoded is '0'. If the golomb parameter of the data to be encoded is 4, first, the context of the samples on the same bitplane is calculated. If the context is obtained according to the first method among the methods of obtaining the context on the same bitplane, the context value is 1 since the sample value is 001 (2). If we calculate the context of the samples on the upper bitplane of the same frequency, 10 (2), the context value is 2 according to the first method. Therefore, the probabilistic model selects the probabilistic model using the above three parameters, that is, the Golomb parameter is '4', the context 1 is '1', and the context 2 is '2'. The probabilistic model may vary depending on the implementation, but one of them can be represented as Prob [Golomb] [Context1] [Context2] with a three-dimensional array.

Lossless coding is performed using the stochastic model. Arithmetic coding may be used as a representative of the lossless coding.

Meanwhile, a lossless audio decoding apparatus and method according to the present invention will be described. 9 is a block diagram illustrating a configuration of an embodiment of a lossless audio decoding apparatus according to the present invention. The parameter obtaining unit 900, the sample selecting unit 910, the context calculating unit 920, and the probability model selecting unit are shown. 930, and includes an arithmetic decoding unit 940.

When the bitstream of the audio data is input, the parameter obtaining unit 900 obtains the most significant bit (MSB) and Golomb parameters from the bitstream. The sample selector 910 selects a binary sample to be decoded from the most significant bit (MSB) to the least significant bit (LSB) in order of low frequency to high frequency.

The context calculator 920 calculates a predetermined context by using the already decoded samples, and includes a first context calculator 1000 and a second context calculator 1020 as shown in FIG. 10. . The first context calculator 1000 calculates a first context using previously decoded samples existing on the same bitplane. The second context calculator 1020 calculates a second context by using samples of the upper bitplane already decoded on the same frequency.

The probability model selecting unit 930 selects a probability model using the Golomb parameter of the parameter obtaining unit 900 and the context calculated by the context calculating unit 920. The arithmetic decoding unit 940 performs arithmetic decoding using a probability model selected by the probability model selecting unit 930.

11 is a block diagram illustrating a configuration of another embodiment of a lossless audio decoding apparatus according to the present invention. The demultiplexer 1100, the lossy decoding unit 1110, the lossless decoding unit 1120, and the audio signal synthesis unit are shown. 1130, including an inverse integer time / frequency conversion unit 1140, and preferably further including an inverse time / frequency conversion unit 1150.

When the audio bitstream is input, the demultiplexer 1100 demultiplexes the audio bitstream, and the loss bitstream and the error data formed by a predetermined loss coding method used when the bitstream is encoded. Extract the error bitstream of.

The lossy decoder 1110 loss-decodes the lost bitstream extracted by the demultiplexer 1100 by a predetermined loss decoding scheme corresponding to the predetermined lossy coding scheme used when the bitstream is encoded. The lossless decoding unit 1120 losslessly decodes the error bitstream extracted by the demultiplexer 1100 by using a lossless decoding method corresponding to lossless encoding.

The audio signal synthesis unit 1130 synthesizes the decoded lossy bitstream and the error bitstream and restores the frequency spectral signal. The inverse integer time / frequency converter 1140 converts the frequency spectral signal restored by the audio signal synthesizer 1130 to inverse integer time / frequency to restore the audio signal in the time domain.

The inverse time / frequency converter 1150 restores the audio signal of the frequency domain decoded by the lossy decoder 1110 to an audio signal of the time domain, and the restored signal is a lossy decoded signal.

12 is a block diagram showing the configuration of the lossless decoding unit 1120. The parameter obtaining unit 1200, the sample selecting unit 1210, the context calculating unit 1220, the probability model selecting unit 1230, and the arithmetic decoding It includes a portion 1240.

The parameter acquisition unit 1200 obtains the most significant bit (MSB) and Golomb parameters from the bitstream of the audio data. The sample selector 1210 selects a binary sample to be decoded in order from the most significant bit to the least significant bit and from the lowest frequency to the highest frequency.

The context calculator 1220 calculates a predetermined context by using the already decoded samples, and includes a first context calculator and a second context calculator. The first context calculator calculates a first context using previously decoded samples existing on the same bit plane. The second context calculator calculates a second context by using samples of the upper bitplane already decoded on the same frequency.

The probability model selector 1230 selects a probability model using the Golomb parameter and context. The arithmetic decoding unit 1240 performs arithmetic decoding using the selected probability model.

FIG. 13 is a flowchart illustrating an operation of an embodiment of a lossless audio decoding apparatus according to the present invention shown in FIG. 9, and a configuration of the lossless audio decoding apparatus will be described with reference to the flowchart.

First, when a bitstream of audio data is input to the parameter acquisition unit 900, a Golomb parameter is obtained from the bitstream of the audio data (step 1300). Then, the sample selector 910 moves from the most significant bit to the least significant bit. In operation 1310, a binary sample to be decoded is selected in order from a low frequency to a high frequency.

When the sample to be decoded is selected by the sample selector 910, the context calculator 920 calculates a predetermined context using the samples that have already been decoded (step 1320). It is composed of the second context, and as shown in FIG. 10, the first context is calculated using the already decoded samples existing on the same bit plane through the first context calculator 1000. The second context calculator 1020 calculates a second context by using samples of the upper bitplane already decoded on the same frequency.

Then, a probability model is selected using the Golomb parameter, the first context, and the second context through the probability model selecting unit 930. In operation 1330, a probability model is selected in the probability model selecting unit 930. When selected, the arithmetic decoding is performed using the selected probability model. (Step 1340) Steps 1310 to 1340 are repeated until all samples are decoded.

FIG. 14 is a flowchart illustrating an operation of another embodiment of the lossless audio decoding apparatus according to the present invention shown in FIG. 11, and the operation of the lossless audio decoding apparatus will be described with reference to the flowchart.

The difference between the lossy coded audio data and the audio spectral signal in the frequency domain having an integer value is defined as error data. First, when an audio bitstream is input to the demultiplexer 1100, the bitstream is demultiplexed to extract a lost bitstream generated through a predetermined loss encoding and an error bitstream of the error data. step)

The extracted loss bitstream is inputted to a lossy decoding unit 1110 and lossy decoded by a predetermined lossy decoding method corresponding to the lossy encoding at the time of encoding. (Step 1410) The extracted error bitstream is losslessly decoded. It is inputted to the unit 1120 to perform lossless decoding. (Step 1420) The detailed process of the lossless decoding (step 1420) is the same as that shown in FIG.

The lossy bitstream lost by the loss decoding unit 1110 and the error bitstream losslessly decoded by the lossless decoding unit 1120 are input to the audio signal synthesis unit 1130 to restore a frequency spectral signal. (Step 1430) The frequency spectral signal is input to the inverse integer time / frequency converter 1140 to restore the audio signal of the time domain (step 1440).

The present invention can be embodied as code that can be read by a computer (including all devices having an information processing function) in a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording devices include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like.

Although the present invention has been described with reference to the embodiments shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

According to the method and apparatus for lossless audio encoding / decoding according to the present invention, an optimum performance is provided through a model made through a statistical distribution regardless of an input distribution. In other words, the optimal compression ratio is provided regardless of the assumption that integer MDCT coefficients have a Laplacian distribution. Therefore, it provides superior compression rate than BPGC through context-based coding.

Claims (40)

A signal conversion step of converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; A mapping step of mapping the audio signal in the frequency domain into a bitplane signal according to frequency; And And a lossless encoding step of lossless encoding the binary sample constituting the bitplane through a probability model determined using a predetermined context. The method of claim 1, wherein the lossless encoding step is Mapping the audio signal in the frequency domain to bitplane data according to frequency; Obtaining a most significant bit and a Golomb parameter for each bit plane; Selecting a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from low frequency component to high frequency; For the selected binary sample, calculating a context of the selected binary sample using previous samples on the same bitplane; Selecting a probabilistic model using the obtained Golomb parameter and the calculated contexts; And And lossless encoding the binary sample using the selected probability model. (a) converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; (b) matching the audio spectral signal in the frequency domain with an input signal of a lossy encoder through scaling; (c) lossy compression coding the scaled frequency signal; (d) obtaining an error mapped signal corresponding to a difference between the lossy coded data and an audio spectral signal in the frequency domain having the integer value; (e) lossless encoding the error mapped signal using a context; And and (f) multiplexing the lossless coded signal and the lossy coded signal into a bitstream. The method of claim 3, wherein step (e) (e1) mapping the error mapped signal of step (d) to bitplane data according to frequency; (e2) obtaining the most significant bit and the Golomb parameter of the bitplane; (e3) selecting a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from low frequency component to high frequency; (e4) for the selected binary sample, calculating the context of the selected binary sample using previous samples on the same bitplane; (e5) selecting a probabilistic model using the obtained Golomb parameter and the calculated contexts; And and (e6) lossless encoding the binary sample using the selected probability model. The method of claim 4, wherein step (e4) And for the selected binary sample, calculate the context value of the binary sample by making the encoded previous samples on the same bitplane into a scalar value. The method of claim 4, wherein step (e4) With respect to the selected binary sample, a probability of '1' is obtained for predetermined samples existing in the same bitplane, and the probability value is multiplied by a predetermined integer value, expressed as an integer, and the context value is calculated using the integer. Lossless audio coding method characterized in that. The method of claim 4, wherein step (e4) And a context value is calculated for the selected binary sample using an already encoded upper bitplane value on the same frequency. The method of claim 4, wherein step (e4) For the selected binary sample, a context value is calculated using information on whether there is a value of an already encoded higher bitplane on the same frequency, but if at least one '1' exists in the higher bit plane, the context value is' Lossy audio encoding method, wherein a context value is set to '0' if at least one '1' does not exist. An integer time / frequency converter for converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; And And a lossless encoding unit for mapping the audio signal in the frequency domain into bitplane data according to frequency and losslessly encoding a binary sample constituting the bitplane using a predetermined context. The method of claim 9, wherein the lossless coding unit A bitplane mapping unit for mapping the audio signal in the frequency domain into bitplane data according to frequency; A parameter obtaining unit for obtaining a most significant bit and a Golomb parameter of the bit plane; A binary sample selecting unit which selects a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from lowest frequency component to highest frequency; A context calculator for calculating the context of the selected binary sample for the selected binary sample using previous samples on the same bitplane; A probability model selecting unit which selects a probability model using the obtained Golomb parameter and the calculated contexts; And And a binary sample encoder for lossless encoding the binary sample using the selected probability model. 10. The method of claim 9, wherein the integer time / frequency converter Lossless audio coding apparatus characterized in that the integer MDCT. An integer time / frequency converter for converting an audio signal in a time domain into an audio spectral signal in a frequency domain having an integer value; A scaling unit for matching an audio frequency signal of the integer time / frequency conversion unit with an input signal of a lossy coding unit; A lossy coding unit for lossy compression encoding the scaled frequency signal; An error mapping unit obtaining a difference between the lossy coded signal and the signal of the integer time / frequency converter; A lossless encoding unit for losslessly encoding the error mapped signal using a context; And a multiplexer for multiplexing the lossless coded signal and the lossy coded signal into a bitstream. The method of claim 12, wherein the lossless coding unit A bitplane mapping unit for mapping the error mapped signal of the error mapping unit to bitplane data according to a frequency; A parameter obtaining unit for obtaining a most significant bit and a Golomb parameter of the bit plane; A binary sample selection unit for selecting a binary sample on a bitplane to be encoded in order from most significant bit to least significant bit, and then from low frequency component to high frequency; A context calculator for calculating the context of the selected binary sample for the selected binary sample using previous samples on the same bitplane; A probability model selecting unit which selects a probability model using the obtained Golomb parameter and the calculated contexts; And And a binary sample encoder for lossless encoding the binary sample using the selected probability model. The method of claim 13, wherein the context calculation unit And for the selected binary sample, calculates a context value of the binary sample by making encoded previous samples on the same bitplane into one scalar value. The method of claim 13, wherein the context calculation unit With respect to the selected binary sample, a probability of '1' is obtained for predetermined samples existing in the same bitplane, and the probability value is multiplied by a predetermined integer value, expressed as an integer, and the context value is calculated using the integer. Lossless audio encoding device characterized in that. The method of claim 13, wherein the context calculation unit And a context value is calculated for the selected binary sample by using an already encoded upper bitplane value on the same frequency. The method of claim 13, wherein the context calculation unit For the selected binary sample, a context value is calculated using information on whether there is a value of an already encoded higher bitplane on the same frequency, but if at least one '1' exists in the higher bit plane, the context value is' Lossy audio encoding apparatus, characterized in that the context value is '0' if at least one is present. Obtaining a Golomb parameter from a bitstream of audio data; A sample selecting step of selecting a binary sample to be decoded in order from the most significant bit to the least significant bit, and then from a low frequency to a high frequency; A context calculation step of calculating a predetermined context using already decoded samples; A probability model selection step of selecting a probability model using the Golomb parameter and a context; An arithmetic decoding step of performing arithmetic decoding using the selected probability model; And And repeating the sample selection step and the arithmetic decoding step until all samples are decoded. 19. The method of claim 18, wherein the context calculation step Calculating a first context using previously decoded samples existing on the same bitplane; And And calculating a second context using samples of a higher bitplane already decoded on the same frequency. When the difference between the lossy coded audio data and the audio spectral signal in the frequency domain having an integer value is error data, (aa) demultiplexing an audio bit stream to extract a predetermined loss coded loss bit stream and an error bit stream of the error data; (bb) predetermined loss decoding the extracted loss bitstream; (cc) lossless decoding the extracted error bitstream; (dd) restoring a frequency spectral signal using the decoded lossy bitstream and error bitstream; And (ee) restoring an audio signal in a time domain by inverse integer time / frequency transforming the frequency spectral signal. The method of claim 20, wherein the (cc) step (cc1) obtaining the Golomb parameter from the bitstream of the audio data; (cc2) selecting binary samples to be decoded from most significant bit to least significant bit, in order of low frequency to high frequency; (cc3) calculating a predetermined context using already decoded samples; (cc4) selecting a probabilistic model using the Golomb parameter and context; (cc5) performing arithmetic decoding using the selected probability model; And and (cc6) repeating steps (cc2) to (cc5) until all samples are decoded. The method of claim 21, wherein (cc3) is A method for lossless audio decoding, characterized in that the first context is calculated using previously decoded samples existing on the same bitplane. The method of claim 21, wherein (cc3) is And a second context is calculated using samples of a higher bitplane already decoded on the same frequency. The method of claim 21, wherein (cc3) is Calculating a first context using previously decoded samples existing on the same bitplane; And And calculating a second context using samples of a higher bitplane already decoded on the same frequency. A parameter obtaining unit which obtains a Golomb parameter from a bitstream of audio data; A sample selecting unit for selecting a binary sample to be decoded in order from the most significant bit to the least significant bit and then from a low frequency to a high frequency; A context calculator which calculates a predetermined context using samples that have already been decoded; A probabilistic model selection unit for selecting a probabilistic model using the Golomb parameter of the parameter obtaining unit and the context of the context calculating unit; And And an arithmetic decoding unit configured to perform arithmetic decoding by using the selected probabilistic model. The method of claim 25, wherein the context calculation unit A first context calculator configured to calculate a first context using previously decoded samples existing on the same bitplane; And And a second context calculator configured to calculate a second context using samples of an upper bitplane already decoded on the same frequency. When the difference between the lossy coded audio data and the audio spectral signal in the frequency domain having an integer value is error data, A demultiplexer for demultiplexing an audio bitstream to extract a predetermined loss coded loss bitstream and an error bitstream of the error data; A lossy decoding unit for predetermined lossy decoding of the extracted lossy bitstream; A lossless decoding unit for lossless decoding the extracted error bitstream; An audio signal synthesizing unit for synthesizing the decoded lossy bitstream and the error bitstream to restore a frequency spectral signal; And And an inverse integer time / frequency converter for restoring an audio signal in a time domain by performing inverse integer time / frequency conversion on the restored frequency spectral signal. The method of claim 27, wherein the loss decoding unit Lossless audio decoding apparatus characterized in that the AAC decoder. The method of claim 27, And an inverse time / frequency converting unit for restoring the audio signal in the frequency domain decoded by the lossy decoding unit to the audio signal in the time domain. The method of claim 27, wherein the lossless decoding unit A parameter obtaining unit for obtaining a Golomb parameter from a bitstream of audio data; A sample selecting unit for selecting a binary sample to be decoded in order from the most significant bit to the least significant bit and then from a low frequency to a high frequency; A context calculator which calculates a predetermined context using samples that have already been decoded; A probability model selection unit for selecting a probability model using the Golomb parameter and a context; And And an arithmetic decoding unit configured to perform arithmetic decoding by using the selected probabilistic model. The method of claim 30, wherein the context calculation unit A first context calculator configured to calculate a first context using already decoded samples existing in front of the same bitplane; And And a second context calculator configured to calculate a second context using samples of an upper bitplane already decoded on the same frequency. A computer-readable recording medium having recorded thereon a program for executing the invention according to any one of claims 1 to 8 and 18 to 24 on a computer. Obtaining a Golomb parameter from a bitstream of audio data; Selecting a bit plane symbol to be decoded in order from a low frequency to a high frequency from a most significant bit plane to a least significant bit plane for each predetermined frequency band; Determining a context using the already decoded bit plane symbols, and selecting a probability model of the bit plane symbol to be currently decoded using the determined context; And And performing arithmetic decoding using the selected probabilistic model. Obtaining a Golomb parameter from a bitstream of audio data; Selecting binary samples to be decoded in order from least significant bit plane to least significant bit plane for each predetermined frequency band; Determining a context using the already decoded binary samples and selecting a probabilistic model of a binary sample to be currently decoded using the determined context; And And performing arithmetic decoding using the selected probabilistic model. Obtaining a Golomb parameter from a bitstream of audio data; Selecting a bit plane symbol to be decoded in order from a low frequency to a high frequency from a most significant bit plane to a least significant bit plane for each predetermined frequency band; Determining a context using the already decoded bit plane symbols, and selecting a probability model of the bit plane symbol to be currently decoded using the determined context; Performing arithmetic decoding using the selected probability model, Selecting the symbol, selecting the probabilistic model, and performing the arithmetic decoding process are repeatedly performed until all the bit plane symbols are decoded. Obtaining a Golomb parameter from a bitstream of audio data; Selecting binary samples to be decoded in order from least significant bit plane to least significant bit plane for each predetermined frequency band; Determining a context using the already decoded binary samples and selecting a probabilistic model of a binary sample to be currently decoded using the determined context; And Performing arithmetic decoding using the selected probability model, Selecting the sample, selecting the probabilistic model, and performing the arithmetic decoding, which are performed repeatedly until all samples are decoded. Obtaining a Golomb parameter from a bitstream of audio data; Selecting bit plane symbols to be decoded in order of least significant bit plane from most significant bit plane for each predetermined frequency band; Determining a context using the already decoded bit plane symbols, and selecting a probability model of the bit plane symbol to be currently decoded using the determined context; And And a computer-readable recording medium having recorded thereon a program for executing the arithmetic decoding using the selected probability model. Obtaining a Golomb parameter from a bitstream of audio data; Selecting binary samples to be decoded in order from least significant bit plane to least significant bit plane for each predetermined frequency band; Determining a context using the already decoded binary samples and selecting a probabilistic model of a binary sample to be currently decoded using the determined context; And And a computer-readable recording medium having recorded thereon a program for executing the arithmetic decoding using the selected probability model. Obtaining a Golomb parameter from a bitstream of audio data; Selecting bit plane symbols to be decoded in order of least significant bit plane from most significant bit plane for each predetermined frequency band; Determining a context using the already decoded bit plane symbols, and selecting a probability model of the bit plane symbol to be currently decoded using the determined context; Performing arithmetic decoding using the selected probability model, The step of selecting the symbol, the step of selecting the probabilistic model and the arithmetic decoding step are computer readable recordings of programs for executing the invention repeatedly performed until all the bit plane symbols are decoded. media. Obtaining a Golomb parameter from a bitstream of audio data; Selecting binary samples to be decoded in order from least significant bit plane to least significant bit plane for each predetermined frequency band; Determining a context using the already decoded binary samples and selecting a probabilistic model of a binary sample to be currently decoded using the determined context; And Performing arithmetic decoding using the selected probability model, Selecting the sample, selecting the probabilistic model, and performing the arithmetic decoding process, the computer-readable recording medium having recorded thereon a program for executing the invention repeatedly performed until all samples are decoded.

Images