1. Field of the Invention
Embodiments of the present invention relate to scalable audio data decoding, and more particularly, to a scalable audio data arithmetic decoding method, medium, and apparatus, and a method, medium, and apparatus truncating an audio data bitstream.
2. Description of the Related Art
Audio lossless encoding techniques have been required for audio broadcasting and/or archiving purposes. Major technologies for lossless audio encoding include application of an entropy encoder using time/frequency transformation or linear prediction, for example.
When scalability through bitstream re-parsing is applied, for example, a bitstream corresponding to a frame is truncated at an arbitrary position, at a server end, and transmitted to a decoding end.
First, initialization is performed, in operation 100, and a symbol desired to be decoded is detected, in operation 110. By using the corresponding context, a probability value for the symbol can be calculated, in operation 120, and arithmetic decoding can then be performed, in operation 130. Here, the probability value for a symbol corresponds to the probability that a symbol is a ‘1’ or ‘0’, for example where the symbol is a binary number. Whether the symbol is the end of the bitstream can then be checked, in operation 140, and if the symbol is not the end of the bitstream, a symbol to be decoded can again be determined and the above operations may be repeated. The decoding is finished when the symbol is determined to be the end of the bitstream.
Meanwhile, when an arithmetic decoding method is performed, all of the symbols to be decoded are known, or a predetermined termination code is inserted, and the decoder is informed of the time when the decoding should be finished. However, when a bitstream is truncated, as shown in
Embodiments of the present invention, as set forth herein, include a scalable audio data arithmetic decoding method, medium, and apparatus capable of efficiently terminating decoding without decoding errors.
Embodiments of the present invention also include a method, medium, and apparatus truncating a scalable audio data bitstream.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a scalable data arithmetic decoding method for decoding a scalable arithmetic coded symbol, including arithmetic decoding a desired symbol by using the symbol and a probability value for the symbol, and determining whether to continue a decoding of the symbol by checking for an ambiguity indicating whether the decoding of the symbol is complete, wherein, in the determining of whether to continue the decoding, when a valid bitstream remaining after truncation is decoded and then decoding is performed by using dummy bits in order to decode the bitstream, truncated for scalability, if the symbol is decoded regardless of the dummy bits, the decoding is continuously performed, and if the symbol is decoded relying on the dummy bits, and it is determined that the ambiguity occurs, then the decoding is correspondingly terminated.
The determining of whether to continue decoding may include calculating K, assuming that K is a right-hand side value of a following equation:
This may further include determining, according to a value of K, whether to continue the decoding, where in these equations, v1 denotes a value of the valid bitstream remaining after truncation, v2 denotes a value of the truncated bitstream after the truncation, dummy denotes a number of v2 bits, freq denotes the probability value for the symbol, high and low denote an upper limit and a lower limit, respectively, of a range in which the probability value exists, decoding the symbol as 1 if K is equal to or greater than 2dummy-1, and decoding the symbol as 0 if K is equal to or less than 0, and determining that the ambiguity occurs, if K is between 0 and 2dummy-1, and correspondingly terminating the decoding.
Before the arithmetic decoding of the symbol, the method may include finding the symbol, and calculating the probability value for the symbol.
The calculation of the probability value for the symbol may include finding a decoding mode from header information of a bitstream to be decoded, and obtaining the probability value for the symbol by referring to a context of the symbol if the decoding mode is a context-based arithmetic coding mode (cbac).
In the arithmetic decoding of the symbol, if a first non-zero sample on a bitplane is decoded, a sign bit corresponding to the sample may be arithmetic decoded, and in the determining that the ambiguity occurs, if K is between 0 and 2dummy-1, the ambiguity may have been determined to have occurred, and the decoding may be terminated by setting a sample, decoded immediately before the ambiguity, to 0.
The calculation of the probability value for the symbol may include finding a decoding mode from header information of a bitstream to be decoded, and if the decoding mode is a bitplane Golomb mode (bpgc), obtaining the probability value for the symbol, assuming that the data to be decoded has a Laplacian distribution.
In the arithmetic decoding of the symbol, if a first non-zero sample on a bitplane is decoded, a sign bit corresponding to the sample may be arithmetically decoded, and, in the determining that the ambiguity occurs, if K is between 0 and 2dummy-1, the ambiguity may be determined to have occurred, and the decoding is terminated with setting a sample, decoded immediately before the ambiguity, to 0.
The calculation of the probability value for the symbol may further include finding a decoding mode from header information of a bitstream to be decoded, and if the decoding mode is a low energy mode, obtaining the probability value for the symbol by using probability model information of the bitstream header.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a scalable data arithmetic decoding apparatus to decode a scalable arithmetic coded symbol, including a symbol decoding unit to arithmetic decode a desired symbol by using the symbol and a probability value for the symbol, and an ambiguity checking unit to determine whether to continue a decoding by checking for an ambiguity, the ambiguity checking unit including a decoding continuation determination unit to calculate K, assuming that K is a right-hand side value of a following equation, and according to a value of K, determining whether to continue decoding:
Here, v1 denotes a value of a valid bitstream remaining after truncation, v2 denotes a value of a truncated bitstream after the truncation, dummy denotes a number of v2 bits, freq denotes the probability value for the symbol, high and low denote an upper limit and lower limit, respectively, of a range in which the probability value exists. The apparatus may further include an additional decoding unit to decode the symbol as 1 if K is equal to or greater than 2dummy-1, and to decode the symbol as 0 if K is equal to or less than 0, and a decoding termination unit to determine that the ambiguity occurs if K is between 0 and 2dummy-1, and to correspondingly terminate the decoding.
The apparatus may further include a symbol determination/probability prediction unit to find the symbol and to calculate the probability value for the symbol.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a method of truncating a scalable data bitstream including parsing a length of the bitstream, from a header of the bitstream, calculating bytes corresponding to a target bitrate by reading the bitstream, modifying the bitstream length with a smaller value between the calculated target bytes and an actual number of bits, and storing and transmitting a truncated bitstream based on the bitstream and the target length.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a scalable audio data arithmetic decoding method for decoding a scalable audio arithmetic coded symbol, including arithmetic decoding a desired symbol by using the symbol and a probability value for the symbol, and determining whether to continue a decoding of the symbol by checking for an ambiguity indicating whether the decoding of the symbol is complete, wherein the determining of whether to continue the decoding may include calculating K, assuming that K is a right-hand side value of following equation:
Here, the method may further include determining, according to a value of K, whether to continue decoding, where in these equations, v1 denotes a value of a valid bitstream remaining after truncation, v2 denotes a value of a truncated bitstream after the truncation, dummy denotes a number of v2 bits, freq denotes the probability value for the symbol, high and low denote an upper limit and a lower limit, respectively, of a range in which the probability value exists, decoding the symbol as 1 if K is equal to or greater than 2dummy-1, and decoding the symbol as 0 if K is equal to or less than 0, and determining that the ambiguity occurs, if K is between 0 and 2dummy-1, and correspondingly terminating the decoding.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a scalable audio data arithmetic decoding method for decoding a scalable audio arithmetic coded symbol, including arithmetic decoding a desired symbol by using the symbol and a probability value for the symbol wherein, in the calculation of the probability value for the symbol, a decoding mode is found from header information of a bitstream to be decoded and if the decoding mode is a context-based arithmetic coding mode (cbac), the probability value for the symbol is obtained by referring to a context of the symbol, and determining whether to continue the decoding of the symbol by checking for an ambiguity indicating whether decoding of a symbol is complete, wherein the determining of whether to continue decoding includes calculating K, assuming that K is a right-hand side value of a following equations:
Here, the method may further include the determining, according to a value of K, whether to continue decoding, where in the equations, v1 denotes a value of a valid bitstream remaining after truncation, v2 denotes a value of a truncated bitstream after the truncation, dummy denotes a number of v2 bits, freq denotes the probability value for the symbol, high and low denote an upper limit and a lower limit, respectively, of a range in which the probability value exists, decoding the symbol as 1 if K is equal to or greater than 2dummy-1, and decoding the symbol as 0 if K is equal to or less than 0, and determining that the ambiguity occurs, if K is between 0 and 2dummy-1, and correspondingly terminating the decoding.
In the arithmetic decoding of the symbol, if a first non-zero sample on a bitplane is decoded, a sign bit corresponding to the sample may be arithmetically decoded, and, in the determining that the ambiguity occurs, if K is between 0 and 2dummy-1, the ambiguity may be determined to have occurred, and the decoding is correspondingly terminated by setting a sample, decoded immediately before the ambiguity, to 0.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a scalable audio data arithmetic decoding method for decoding a scalable audio arithmetic coded symbol, including arithmetic decoding a desired symbol by using the symbol and a probability value for the symbol, wherein, in the calculation of the probability value for the symbol, a decoding mode is found from header information of a corresponding bitstream to be decoded and if the decoding mode is a bitplane Golomb mode (bpgc), the probability value for the symbol is obtained assuming that data to be decoded has a Laplacian distribution, and determining whether to continue decoding by checking for an ambiguity indicating whether the decoding of a symbol is complete, wherein the determining of whether to continue decoding includes calculating K, assuming that K is a right-hand side value of a following equation:
Here, the method may further include determining, according to a value of K, whether to continue decoding, where in these equations, v1 denotes a value of a valid bitstream remaining after truncation, v2 denotes a value of a truncated bitstream after the truncation, dummy denotes a number of v2 bits, freq denotes the probability value for the symbol, high and low denote an upper limit and a lower limit, respectively, of a range in which the probability value exists, decoding the symbol as 1 if K is equal to or greater than 2dummy-1, and decoding the symbol as 0 if K is equal to or less than 0, and determining that the ambiguity occurs, if K is between 0 and 2dummy-1, and correspondingly terminating the decoding.
In the arithmetic decoding of the symbol, if a first non-zero sample on a bitplane is decoded, a sign bit corresponding to the sample may be arithmetically decoded, and wherein the determining that the ambiguity occurs, if K is between 0 and 2dummy-1, the ambiguity may be determined to have occurred, and the decoding is correspondingly terminated with setting a sample, decoded immediately before the ambiguity, to 0.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a scalable audio data arithmetic decoding method for decoding a scalable audio arithmetic coded symbol, including arithmetic decoding a desired symbol by using the symbol and a probability value for the symbol, wherein, in the calculation of the probability value for the symbol, a decoding mode is found from header information of a corresponding bitstream to be decoded and if the decoding mode is a low energy mode, the probability value for the symbol is obtained by using probability model information of the bitstream header, and determining whether to continue decoding by checking for an ambiguity indicating whether decoding of the symbol is complete, wherein the determining of whether to continue decoding includes calculating K, assuming that K is a right-hand side value of a following equation:
Here, the method may further include determining, according to the K value, whether to continue decoding, where in the equations, v1 denotes a value of a valid bitstream remaining after truncation, v2 denotes a value of a truncated bitstream after the truncation, dummy denotes a number of v2 bits, freq denotes the probability value for the symbol, high and low denote an upper limit and a lower limit, respectively, of a range in which the probability value exists, decoding the symbol as 1 if K is equal to or greater than 2dummy-1, and decoding the symbol as 0 if K is equal to or less than 0, and determining that the ambiguity occurs, if K is between 0 and 2dummy-1, and correspondingly terminating the decoding.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a method of truncating a scalable data bitstream, including parsing a length of the bitstream from a header of the bitstream, calculating target bytes corresponding to a target bitrate by reading the bitstream, modifying the bitstream length with a smaller value between the calculated target bytes and the actual number of bits, storing and transmitting a truncated bitstream based on the bitstream and the target length, wherein the target bytes are obtained using a following equation:
target_bits=(int)(target_bitrate/2*1024.*osf/sampling_rate+0.5)−16; and target_bytes=(target_bits+7)/8.
Here, target_bitrate denotes a desired target bitrate in bits/sec, sampling_rate denotes a sampling frequency of an input audio signal in Hz, and osf denotes an oversampling factor having any one value of 1, 2, and 4.
To achieve the above and/or other aspects and advantages, embodiments of the present invention include a medium including computer readable code to implement an embodiment of the present invention.
Additional aspects and/or advantages of the invention will be set forth in part in the description which follows and, in part, will be apparent from the description, or may be learned by practice of the invention.
These and/or other aspects and advantages of the invention will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
Reference will be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout. Embodiments are described below to explain the present invention by referring to the figures.
Accordingly, a scalable audio data arithmetic decoding method, medium, and apparatus, and a method, medium, and apparatus truncating an audio data bitstream according to embodiments of the present invention will now be described in greater detail.
According to the pseudo code shown in
According to an embodiment of the present invention, there are 3 bits (dummy bits) in v2, e.g., the value v2 accordingly ranging from 0 to 7.
The arithmetic decoding apparatus may include a symbol decoding unit 520 and an ambiguity checking unit 540. The arithmetic decoding apparatus may further, for example, include a symbol determination/probability prediction unit 500.
The symbol determination/probability prediction unit 500 identifies a symbol to be decoded in a bitstream and predicts the probability value for the symbol.
Performing the probability prediction for the symbol will now be explained. First, from the header information of the bitstream to be decoded, a decoding mode may be detected. If the decoding mode is a context-based arithmetic coding (cbac) mode, as referred to by the context of the symbol to be decoded, the probability value of the symbol may be obtained. If the decoding mode is a bitplane Golomb coding mode, the probability value for the symbol to be decoded may be obtained by assuming that the data to be decoded has a Laplacian distribution. Also, if the decoding mode is a low energy mode, the probability value for of the symbol to be decoded is obtained by using the probability model information of the bitstream header.
The symbol decoding unit 520 may perform arithmetic decoding of the symbol by using the predicted probability and may then generate the symbol. Decoding of the sign bit on a bitplane will now be explained. In decoding of an MPEG-4 scalable lossless bitstream, a first non-zero sample among values on the bitplane may be decoded, and then, the sign corresponding to the sample may be decoded. However, if an ambiguity error occurs in the sign value and the decoding is immediately terminated because of the occurrence of the ambiguity error, the sign of the non-zero sample that is decoded immediately before cannot be known. For this reason, when the decoding is terminated in the sign bit, the sample decoded immediately before is set to 0 and the decoding is terminated.
Assuming that the right-hand side value of the below Equations 1 and 2 is K, the ambiguity checking unit 540 may calculate K, and according to the value of K, determine whether to continue to decode a symbol:
Here, v1 denotes the value of the valid bitstream remaining after truncation, v2 denotes the value of the truncated bitstream after the truncation, dummy denotes the number of v2 bits, freq denotes the probability value for the symbol, high and low denote the upper limit and lower limit of a range in which the probability value for the symbol exists. This is more fully explained in “Study on ISO/IEC 14496-3: 2001/PDAM 5, (Scalable Lossless Coding)”, ISO/IEC JTC 1/SC 29/WG 11 N6792.
Equations 1 and 2 will now be explained in greater detail. A decoding expression of the pseudo code shown in
If (v1+v2−low+1)·214<(high−low+1)·freq, the symbol (sym) may be generated as having the value of 1. Here, if this is rearranged in relation to v2,Equation 1 is obtained.
Also, if (v1+v2−low+1)·214≧(high−low+1)·freq, the symbol (sym) may be generated as having the value of 0. Here, if this is rearranged in relation to v2,Equation 2 is obtained.
In equation 1, if the value of the right-hand side expression is greater than 7, the symbol may be decoded as 1, regardless of v2. In Equation 2, if the value of the right-hand side expression is less than 0, the symbol may be decoded as 0, regardless of v2. In other cases, a decoding ambiguity occurs and the decoding is finished.
Assuming that the right-hand side value of Equations 1 and 2 is K, the decoding continuation determination unit 600 may calculate the value of K, and according to value of K, determine whether or not to continue to decode a symbol. The additional decoding unit 620 may decode the symbol as 1 if K is equal to or greater than 2dummy-1, and if K is equal to or less than 0, decode the symbol as 0. If K is between 0 and 2dummy-1, the decoding termination unit 640 may determine that an ambiguity has occurred, and terminate the decoding.
A symbol to be decoded in an arithmetic coded scalable bitstream may be determined, in operation 700, and the probability value for the determined symbol may be predicted, in operation 710.
Performing the probability prediction of the symbol will now be further explained.
From the header information of the bitstream to be decoded, a decoding mode may be determined. If the decoding mode is a context-based arithmetic coding (cbac) mode, e.g., by referring to the context of the symbol to be decoded, the probability value for the symbol may be obtained. If the decoding mode is a bitplane Golomb coding mode, the probability value for the symbol to be decoded may be obtained by assuming that the data to be decoded has a Laplacian distribution. Also, if the decoding mode is a low energy mode, the probability value for the symbol to be decoded may be obtained by using the probability model information of the bitstream header.
By using the predicted probability, the symbol may be arithmetically decoded and generated, in operation 720.
Assuming that the right-hand side value of equations 1 and 2 is K, when K is calculated, if K found to be between 0 and 2dummy-1, in operation 730, it may be determined that an ambiguity has occurred, and the arithmetic decoding may be determined, in operation 740.
If K is found to be equal to or less than 0, in operation 750, the symbol may be decoded as 0, in operation 760, and if K found to be is equal to or greater than 2dummy-1, the symbol may be decoded as 1, in operation 770.
In the MPEG-4 scalable lossless decoding, a first non-zero sample among values on the bitplane is decoded, then the sign corresponding to the sample is decoded. However, if an ambiguity error occurs in the sign value, and the decoding is immediately terminated because of the occurrence of the ambiguity error, the sign of the non-zero sample that is decoded immediately before cannot be known. For this reason, when the decoding is terminated in the sign bit, the sample decoded immediately before is set to 0 and the decoding is terminated.
First, the pseudo code for arithmetic decoding for each of BPGC, CBAC and low energy modes will now be explained in greater detail. Here, ambiguity_check(f) is a function to detect ambiguity for the arithmetic decoding, with the argument indicating a probability value of 1. The function terminate_decoding() is a function to terminate decoding of LLE data when an ambiguity occurs. The function smart_decoding_cbac bpgc() is a function to decode additional symbols in the absence of incoming bits in cbac/bpgc mode decoding. A scalable audio data arithmetic decoding, according to an embodiment of the present invention, continues up to the point where no ambiguity exists. This code (the pseudo code) includes the above functions, ambiguity_check(f) and terminate_decoding(). In addition, the function smart_decoding low energy() is a function to decode additional symbols in the absence of incoming bits in the low energy mode. This also includes the functions, ambiguity_check(f) and terminate_decoding(), see below:
An arithmetic decoding of the truncated SLS bitstream, according to an embodiment of the present invention, provides an efficient method for decoding an intermediate layer corresponding to a given target bitrate, such that, even when there are no bits input to the decoding buffer, meaningful information is still included in the decoding buffer. The decoding process is performed up to the point where no ambiguity exists in the symbol. The following pseudo code shows an algorithm for detecting an ambiguity in an arithmetic decoding module, according to an embodiment of the present invention. A variable num_dummy bits indicates the number of bits not input to a value buffer because of truncation.
Below, smart_decoding_cbac bpgc or smart_decoding_low energy() may be performed when num_dummy bits is greater than 0. In order to prevent sign bit errors, the spectral value of the current spectral line is set to be zero when an ambiguity can occur while decoding a sign bit. All index variables in the arithmetic decoding process according to an embodiment of the present invention are carried over from the previous arithmetic decoding process.
Below, the process of re-parsing and bitstream truncation, when the size of a bitstream is transmitted in the header, according to a method of generating a truncated bitstream by re-parsing will now be explained.
From the bitstream header information, the length of the bitstream may be parsed, in operation 1000. By using the following equations 3 and 4, bytes corresponding to a target bitrate may be calculated, in operation 1020. The target bitrate may be provided from the outside, for example, by a server or a user.
target_bits=(int)(target_bitrate/2*1024.*osf/sampling_rate+0.5)−16 (3)
target_bytes=(target_bits+7)/8 (4)
With the obtained target byte, the bitstream length can be modified. That is, a smaller value between the actual number of bits and the target_bytes is determined as the length of the bitstream, in operation 1030. A bitstream of the target length may also be stored and transmitted, in operation 1040.
The method of re-parsing and truncating the bitstream will now be explained in more detail. The SLS bitstream can be truncated in a given target bitrate in a simple way. The modification of the values of lle_ics_length does not affect LLE decoding results before the truncation point. The lle_ics_length is independent from an LLE decoding procedure. The bitstream truncation will now be explained. The LLE bitstream is read from the bitstream. The available frame length at a given target bitrate is calculated. The simplest way to calculate the available frame length is by using the above Equations 3 and 4.
Here, in Equations 3 and 4, the variable target_bitrate represents the target bitrate in bits/sec, the variable osf represents an oversampling factor, and the variable sampling_rate represents the sampling frequency of the input audio signal in Hz. By taking a smaller value of the available frame length and the current frame length, lle_ics_length may be updated as follows:
lle_ics_length=min(lle_ics_length, target_bytes).
The truncated bitstream with the updated lle_ics_length can be generated.
Embodiments of the present invention can also be embodied as computer readable code in/on a medium, e.g., on a computer readable recording medium. The medium may be any data storage device that can store/transmit data which can be thereafter be read by a computer system. Examples of the media may include read-only memory (ROM), random-access memory (RAM), CD-ROMs, magnetic tapes, floppy disks, and optical data storage devices, noting that these are only examples.
Thus, according to a scalable audio data arithmetic decoding method, medium, and apparatus of the above described embodiments of the present invention, data to which scalability is applied when arithmetic coding is performed in MPEG-4 scalable lossless audio coding can be efficiently decoded. Even when a bitstream is truncated, a decoding termination point can be known such that additional decoding of the truncated part can be performed.
Although a few embodiments of the present invention have been shown and described, it would be appreciated by those skilled in the art that changes may be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the claims and their equivalents.
Number | Date | Country | Kind |
---|---|---|---|
10-2005-0110878 | Nov 2005 | KR | national |
This application claims the benefit of U.S. Provisional Patent Application Nos. 60/643,118, filed on Jan. 12, 2005, 60/670,643, filed on Apr. 13, 2005, and 60/673,363, filed on Apr. 21, 2005, in the U.S. Patent and Trademark Office, and Korean Patent Application No. 10-2005-0110878, filed on Nov. 18, 2005, in the Korean Intellectual Property Office, the disclosures of which are incorporated herein in their entirety by reference.
Number | Name | Date | Kind |
---|---|---|---|
5592163 | Kimura et al. | Jan 1997 | A |
6108622 | Xue et al. | Aug 2000 | A |
6122618 | Park | Sep 2000 | A |
6229463 | Van Der Vleuten et al. | May 2001 | B1 |
6275176 | Bruekers et al. | Aug 2001 | B1 |
6349284 | Park | Feb 2002 | B1 |
6765510 | Koyama et al. | Jul 2004 | B2 |
7006702 | Kondo | Feb 2006 | B2 |
20020006225 | Kimura et al. | Jan 2002 | A1 |
20030187634 | Li | Oct 2003 | A1 |
20070016427 | Thumpudi et al. | Jan 2007 | A1 |
20070115153 | Choo et al. | May 2007 | A1 |
Number | Date | Country |
---|---|---|
2280816 | Feb 1995 | GB |
Number | Date | Country | |
---|---|---|---|
20060284748 A1 | Dec 2006 | US |
Number | Date | Country | |
---|---|---|---|
60673363 | Apr 2005 | US | |
60670643 | Apr 2005 | US | |
60643118 | Jan 2005 | US |