KR101001746B1 - Scalable audio data arithmetic decoding method and apparatus - Google Patents

Abstract

Disclosed are a scalable audio data arithmetic decoding method and apparatus and a scalable audio bitstream truncation method. The method of arithmetic decoding includes: (a) performing arithmetic decoding using a symbol to be decoded and a probability of generating the symbol; And (b) determining whether to continue decoding by checking an ambiguity indicating whether or not the symbol has been decoded.

According to the present invention, it is possible to efficiently decode when scalability is applied to arithmetic coding applied in MPEG-4 scalable lossless audio coding. Even when the bitstream is truncated, the end point of decoding can be known, and additional decoding can be performed on the cut-out part.

Description

Scalable audio data arithmetic decoding method and apparatus

1 is a flowchart illustrating a general arithmetic decoding method.

Figure 2 shows the case of bitstream truncation for general scalability.

 3 shows pseudo code for general binary arithmetic decoding.

4 illustrates a value input to a buffer near a truncation point when the bitstream is truncated.

5 is a block diagram showing the configuration of an apparatus for arithmetic decoding of scalable audio data according to the present invention.

6 is a block diagram illustrating a more detailed configuration of the ambiguity check unit 540.

7 is a flowchart illustrating a method for arithmetic decoding of scalable audio data according to the present invention.

8 shows an example of further decoding by applying the method of determining whether to continue restoring the scalable bitstream according to the present invention.

9 illustrates an example of processing an ambiguity when a sign bit occurs.

10 is a flowchart illustrating a method of cutting a bitstream of scalable audio data according to the present invention.

The present invention relates to scalable audio data decoding, and more particularly, to a scalable audio arithmetic decoding method and apparatus and a scalable audio bitstream truncation method.

Description of the art of audio lossless coding is in demand for audio broadcasting or archiving. In lossless audio coding, the main technique is to apply an entropy encoder using time / frequency conversion or linear prediction.

When applying scalability through bitstream re-parsing, the bitstream corresponding to one frame is truncated at an arbitrary position on the server side. To the decoder.

1 is a flowchart illustrating a general arithmetic decoding method. First, after initializing (step 100), a symbol to be decoded is found (step 110). Arithmetic decoding is performed by calculating a probability of a symbol using a context (step 120). If it is the end of the stream (step 140), if it is not the end, it finds the symbol to be decoded again, repeats the above process and ends the decoding. In general, when an arithmetic decoding process is performed, the entire symbol to be decoded is known or a specific termination code is inserted to inform the decoder when decoding should end. However, when the bitstream is truncated as shown in FIG. 2, since the information is truncated, the decoder cannot know when to end decoding. Since the exact end point is not known, unwanted data may be decrypted when decrypting.

SUMMARY OF THE INVENTION The present invention has been made in an effort to provide a scalable audio data arithmetic decoding method and apparatus for efficiently ending decoding without a decoding error in order to solve the above problems.

An object of the present invention is to provide a scalable audio data bitstream truncation method.

In the scalable audio data arithmetic decoding method according to the present invention for solving the above technical problem, in the method of decoding a scalable arithmetic encoded symbol, (a) arithmetic using a symbol to be decoded and the probability that the symbol occurs Decrypting; And (b) determining whether to continue decoding by checking an ambiguity indicating whether or not the symbol has been decoded.

In step (b), after decoding the valid bit stream remaining after truncation for scalable decoding, virtual data (dummy bit) is used to decode the bit stream cut out for scalability. If the symbol is decoded irrespective of the dummy data, the decoding is continued. If the symbol to be decoded depending on the virtual data is decoded, an ambiguity is determined. To complete the decryption.

The step (b) may include (b1) determining whether to continue decoding a symbol according to K by calculating K when the right side values of Equations 1 and 2 are K;

[Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates an upper limit and a lower limit of a range in which the probability value of the symbol exists. (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; (b3) if K is between 0 and 2 ^dummy −1, determining that an ambiguity has occurred and stopping decoding.

The scalable data arithmetic decoding method includes: finding a symbol to be decoded before the step (a); And calculating the probability of the symbol.

The calculating of the probability of the symbol may include finding a decoding mode from header information of a bitstream to be decoded; And if the decoding mode is a context-based arithmetic coding mode (cbac), obtaining a probability of the symbol by referring to the context of the symbol to be decoded.

Wherein the step (a) if the first non-zero sample among values on the bit-plane decoding, and arithmetic decoding a code (sign) bit corresponding thereto, between the (b3) is a step wherein K is 0 and 2 ^dummy -1 If it is, it is preferable to stop the decoding by determining that the ambiguity has occurred, and to end the decoding by setting the immediately decoded sample to zero.

The calculating of the probability of the symbol may include finding a decoding mode from header information of a bitstream to be decoded; And when the decoding mode is the bitplane gollum coding mode (bpgc), assuming that data to be decoded has a Laplacian distribution, obtaining a probability of a symbol to be decoded. Wherein the step (a) if the first non-zero sample among values on the bit-plane decoding, and arithmetic decoding a code (sign) bit corresponding thereto, between the (b3) is a step wherein K is 0 and 2 ^dummy -1 If it is, it is preferable to stop the decoding by determining that the ambiguity has occurred, and to end the decoding by setting the immediately decoded sample to zero. The calculating of the probability of the symbol may include finding a decoding mode from header information of a bitstream to be decoded; And when the decoding mode is a low energy mode, obtaining a probability of a symbol to be decoded using probability model information of a bitstream header.

According to an aspect of the present invention, there is provided an apparatus for decoding a scalable arithmetic coded symbol, comprising: a symbol decoding unit configured to perform arithmetic decoding on a symbol using a symbol to be decoded and a probability thereof; And an ambiguity check unit for determining whether to continue decoding by checking an ambiguity, wherein the ambiguity check unit calculates K when the right side value of Equations 1 and 2 is K; Decoding continued determination unit for determining whether to continue decoding the symbol according to;

[Equation 1]

[Equation 2]

Where v1 is the value of the valid bitstream remaining after truncation, v2 is the bitstream value truncated after truncation, dummy is the number of bits in v2, freq is the probability value of the symbol, and high and low are the probability values of the symbol. An additional decoding unit which decodes the symbol into 1 when the K is 2 ^dummy −1 or more, and decodes the symbol into 0 when the K is 0 or less; And a decoding stop part that determines that an ambiguity has occurred and stops decoding when K is between 0 and 2 ^dummy −1.

Preferably, the apparatus for decoding the scalable arithmetic coded symbol further includes a symbol determination / probability prediction unit that finds a symbol to be decoded and calculates a probability of the symbol.

According to another aspect of the present invention, there is provided a method of cutting a bitstream of scalable data, the method comprising: analyzing a bitstream length from a header; Reading a bitstream and calculating a byte corresponding to a target bit rate; Modifying a bitstream length by a smaller value of the calculated target byte and the actual number of bits; And storing and transmitting the bitstream by the target length.

In the scalable audio arithmetic decoding method according to the present invention for solving the above technical problem, in the method of decoding a scalable audio arithmetic encoded symbol, (a) arithmetic using a symbol to be decoded and the probability that the symbol occurs Decrypting; And (b) determining whether to continue decoding by checking an ambiguity indicating whether or not the symbol has been decoded, wherein step (b) includes (b1) a value of the right side of Equation 1 and Equation 2 as K; Calculating the K to determine whether to continue decoding the symbol according to K;

[Equation 1]

[Equation 2]

(Where v1 is the truncated bitstream for scalability, the remaining valid bitstream value, v2 is the truncated bitstream value after truncation, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low Represents an upper limit and a lower limit of a range in which a probability value of a symbol exists.) (B2) decoding the symbol as 1 when K is 2 ^dummy −1 or more, and decoding the symbol as 0 when K is 0 or less; (b3) if the K is between 0 and 2 ^dummy −1, determining that an ambiguity has occurred and completing decoding.

In the scalable audio arithmetic decoding method according to the present invention for solving the above technical problem, (a) arithmetic decoding using a symbol to be decoded and the probability that the symbol occurs, the probability calculation of the symbol is Finding a decoding mode from header information, and if the decoding mode is a context-based arithmetic encoding mode (cbac), obtaining a symbol probability by referring to a context of a symbol to be decoded; And (b) determining whether to continue decoding by checking an ambiguity indicating whether or not the symbol has been decoded, wherein step (b) includes (b1) a value of the right side of Equation 1 and Equation 2 as K; Calculating the K to determine whether to continue decoding the symbol according to K;

[Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates an upper limit and a lower limit of a range in which the probability value of the symbol exists. (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; (b3) if K is between 0 and 2 ^dummy -1, determining that an ambiguity has occurred and completing decoding.

Wherein the step (a) if the first non-zero sample among values on the bit-plane decoding, and arithmetic decoding a code (sign) bit corresponding thereto, between the (b3) is a step wherein K is 0 and 2 ^dummy -1 If it is, it is preferable to stop the decoding by determining that the ambiguity has occurred, and to end the decoding by setting the immediately decoded sample to zero.

In the scalable audio arithmetic decoding method according to the present invention for solving the another technical problem, (a) arithmetic decoding using a symbol to be decoded and the probability that the symbol occurs, but the probability calculation of the symbol is a bit to decode Finding a decoding mode from the header information of the stream, and if the decoding mode is a bitplane gollum coding mode (bpgc), assuming that the data to be decoded has a Laplacian distribution, obtaining a probability of a symbol to be decoded; And (b) determining whether to continue decoding by checking an ambiguity indicating whether or not the symbol has been decoded, wherein step (b) includes (b1) a value of the right side of Equation 1 and Equation 2 as K; Calculating the K to determine whether to continue decoding the symbol according to K;

[Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates an upper limit and a lower limit of a range in which the probability value of the symbol exists. (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; (b3) if K is between 0 and 2 ^dummy -1, determining that an ambiguity has occurred and completing decoding.

Wherein the step (a) if the first non-zero sample among values on the bit-plane decoding, and arithmetic decoding a code (sign) bit corresponding thereto, between the (b3) is a step wherein K is 0 and 2 ^dummy -1 If it is determined that the ambiguity has occurred, the decoding is stopped, and the decoding is terminated by setting the immediately decoded sample to 0.

In the scalable audio arithmetic decoding method according to the present invention for solving the another technical problem, (a) arithmetic decoding using a symbol to be decoded and the probability that the symbol occurs, but the probability calculation of the symbol is a bit to decode Finding a decoding mode from the header information of the stream, and when the decoding mode is a low energy mode, obtaining a probability of a symbol to be decoded using probability model information of a bitstream header; And (b) determining whether to continue decoding by checking an ambiguity indicating whether or not the symbol has been decoded, wherein step (b) includes (b1) a value of the right side of Equation 1 and Equation 2 as K; Calculating the K to determine whether to continue decoding the symbol according to K;

[Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates an upper limit and a lower limit of a range in which the probability value of the symbol exists. (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; (b3) if K is between 0 and 2 ^dummy -1, determining that an ambiguity has occurred and completing decoding.

According to another aspect of the present invention, there is provided a method of cutting a bitstream of scalable audio data, the method comprising: analyzing a bitstream length from a header; Reading a bitstream and calculating a byte corresponding to a target bit rate; Modifying a bitstream length by a smaller value of the calculated target byte and the actual number of bits; And storing and transmitting a bitstream by a target length, wherein the target byte is

&Quot; (3) "

target_bits = target_bitrate / 2 * 1024 * osf / sampling_rate + 0.5) -16

&Quot; (4) "

target_bytes = (target_bits + 7) / 8

(Equation 3, 4, target_bitrate is the desired transmission bit rate (bit / sec), sampling_rate is the sampling frequency of the input signal (Hz), osf has one of oversampling factor (1, 2, 4)) It is characterized by the following formula (3, 4).

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

Hereinafter, a scalable audio data arithmetic decoding method and apparatus and a bitstream truncation method according to the present invention will be described in detail with reference to the accompanying drawings.

3 shows pseudo code for general binary arithmetic decoding. This is an arithmetic decoding algorithm used in the entropy coder of MPEG-4 scalable lossless audio coding.

According to the pseudo code shown in FIG. 3, a decoded symbol is first determined by a current frequency / low / high / value value, and then rescaling and updating of a low / high / value value are performed.

4 illustrates a value input to a buffer near a truncation point when the bitstream is truncated. The bitstream after the truncated buffer index no longer has any meaningful information, so it is meaningless. In this case, this value is called v2 and the value in the rest of the buffer is called v1. In this example, since the v1 bit (dummy bit) is 3 bits, the range of v2 has a value of 0-7.

FIG. 5 is a block diagram illustrating a structure of an arithmetic decoding apparatus of scalable audio data according to the present invention, and includes a symbol decoding unit 520 and an ambiguity check unit 540. The arithmetic decoding device may further include a symbol determination / probability prediction unit 500.

The symbol determination / probability prediction unit 500 determines a symbol to be decoded in the bitstream to predict the probability of the symbol.

Probability prediction of the symbol is performed as follows. First, a decoding mode is found from header information of a bitstream to be decoded. If the decoding mode is a context-based arithmetic encoding mode (cbac), the probability of the symbol is obtained by referring to the context of the symbol to be decoded. When the decoding mode is the bitplane gollum coding mode (bpgc), a probability of a symbol to be decoded is obtained assuming that data to be decoded has a Laplacian distribution. In addition, when the decoding mode is a low energy mode, the probability of a symbol to be decoded is obtained using the probability model information of the bitstream header.

The symbol decoding unit 520 generates a symbol by arithmetically decoding the symbol by using the predicted probability. Here, the code bit decoding on the bitplane is performed as follows. In decoding of the MPEG-4 scalable lossless bitstream, a non-zero sample of a value on a bitplane is first decoded and then a corresponding sign is decoded. However, if the decoding is terminated immediately because an ambiguous error has occurred in the code value, the code of the non-zero sample decoded immediately before is not known. For this reason, when decoding is terminated at the sign value, the sample immediately decoded is set to 0 and terminated.

The ambiguity checker 540 calculates K when the right side value of Equations 1 and 2 is K, and determines whether to continue decoding the symbol according to K.

Where v1 is the value of the valid bitstream remaining after truncation, v2 is the bitstream value truncated after truncation, dummy is the number of bits in v2, freq is the probability value of the symbol, and high and low is the range of the probability value of the symbol. The upper limit and lower limit of. (Study on 14496-3: 2001 / PDAM 5, Scalable Lossless Coding (SLS), ISO / IEC JTC 1 / SC 29 / WG 11 N6792] (where v1 is the value of the valid bitstream remaining after truncation, v2 is Bitstream values truncated after truncation, dummy represents the number of bits in v2, freq represents the probability value of the symbol, and high and low represent the upper and lower limits of the range of the probability value of the symbol.)

Equation 1 and Equation 2 will be described in more detail as follows. The decoding equation of the pseudo code shown in FIG. 3 is divided into v1 and v2 to develop.

If (v1 + v2-low + 1) · 2 ¹⁴ <(high-low +1) · freq, a symbol is generated with one. If this is summarized centering on v2, it becomes like Formula (1).

Also, if (v1 + v2-low + 1) · 2 ¹⁴ ≥ (high-low +1) · freq, the symbol is generated with zero. If this is summarized centering on v2, it becomes like Equation 2.

In the case of Equation 1, if the expression value on the right side is larger than 7, it is decoded to 1 regardless of v2. In Equation 2, if the equation value on the right side is smaller than 0, it is decoded to 0 regardless of v2. . In other cases, the decoding ambience occurs and the decoding ends.

6 shows a more detailed configuration of the ambiguity checker 540 as a block diagram, and includes a decoding continued determination unit 600, an additional decoding unit 620, and a decoding stop unit 640.

The decoding continued determination unit 600 calculates K when the right side value of Equations 1 and 2 is K, and determines whether to continue decoding the symbol according to K. The additional decoding unit 620 decodes the symbol as 1 when the K is 2 ^dummy −1 or more, and decodes the symbol as 0 when the K is 0 or less. If the K stops between 0 and 2 ^dummy −1, the decoding stop unit 640 determines that an ambiguity has occurred and stops decoding.

FIG. 7 is a flowchart illustrating a method of performing arithmetic decoding of scalable audio data according to the present invention. A method of determining whether to continue restoring a scalable bitstream according to the present invention will be described with reference to the flowchart.

First, a symbol to be decoded is determined in an arithmetic coded scalable bitstream (step 700), and a probability of the determined symbol is predicted (step 710).

Probability prediction of the symbol is performed as follows. First, a decoding mode is found from header information of a bitstream to be decoded. If the decoding mode is a context-based arithmetic encoding mode (cbac), the probability of the symbol is obtained by referring to the context of the symbol to be decoded. When the decoding mode is the bitplane gollum coding mode (bpgc), a probability of a symbol to be decoded is obtained assuming that data to be decoded has a Laplacian distribution. In addition, when the decoding mode is a low energy mode, the probability of a symbol to be decoded is obtained using the probability model information of the bitstream header.

The symbol is generated by arithmetic decoding using the predicted probability (step 720).

When the right side value of Equations 1 and 2 is K, if the K is calculated and K is between 0 and 2 ^dummy -1 (step 730), it is determined that an ambiguity has occurred and the arithmetic decoding is stopped. (Step 740)

If K is equal to or less than 0 (step 750), the symbol is decoded to 0 (step 760). If K is equal to or greater than 2 ^dummy −1, the symbol is decoded to 1 (step 770).

8 shows an example of further decoding by applying the method of determining whether to continue restoring the scalable bitstream according to the present invention. In FIG. 8, five samples are additionally decoded.

In MPEG-4 scalable lossless decoding, a first non-zero sample of a value on a bitplane is decoded and then a corresponding sign is decoded. However, if it is determined that the ambiguity has occurred in the code value and the decoding ends, the code of the immediately decoded non-zero sample cannot be known. For this reason, when decoding is terminated at the sign value, the sample immediately decoded is set to 0 and terminated.

FIG. 9 illustrates an example of processing an ambiguity in a sign bit.

On the other hand, the arithmetic code of each arithmetic decoding for BPGC, CBAC and loww energy modes according to the present invention is as follows. ambiguity_check (f) is an ambiguity detection function in arithmetic decoding. argument f represents the probability of 1 terminate_decoding () is a function that terminates decoding of LLE data when an ambiguity occurs. smart_decoding_cbac_bpgc () is a function for decoding additional symbols when there is no input bit in cbac / bpgc mode. The decoding according to the present invention continues up to the point where there is no ambiguity. This includes the ambiguity_check (f) and terminate_decoding () functions. smart_decoding_low_energy () is a function that decodes additional symbols when there is no input bit in low energy mode. This also includes the ambiguity_check (f) and terminate_decoding () functions.

while ((max_bp [g] [sfb] cur_bp [g] [sfb] <LAZY_BP) && (cur_bp [g] [sfb]> = 0)) {

 for (g = 0; g <num_windows_group; g ++) {

   for (sfb = 0; sfb <num_sfb; sfb ++) {

     if ((cur_bp [g] [sfb]> = 0) && (lazy_bp [g] [sfb]> 0)) {

width = swb_offset [g] [sfb + 1] swb_offset [g] [sfb];

for (win = 0; win <window_group_len [g]; win ++) {

 for (bin = 0; bin <width; bin ++) {

if (! is_lle_ics_eof ()) {

if (interval [g] [win] [sfb] [bin]> res [g] [win] [sfb] [bin] + (1 << cur_bp [g] [sfb])

{

freq = determine_frequency ();

res [g] [win] [sfb] [bin] + = decode (freq) << cur_bp [g] [sfb];

/ * decode bit-plane cur_bp * /

if ((! is_sig [g] [win] [sfb] [bin]) && (res [g] [win] [sfb] [bin])) {

/ * decode sign bit of res if necessary * /

res [g] [win] [sfb] [bin] * = (decode (freq_sign))? 1: -1;

is_sig [g] [win] [sfb] [bin] = 1;

}

}

}

else {

smart_decoding_cbac_bpgc ();

}

cur_bp [g] [sfb]-; / * progress to next bit-plane * / removed ??

}

}

cur_bp [g] [sfb]-; / * progress to next bit-plane * /

}

}

}

}

/ * low energy mode decoding * /

for (g = 0; g <num_windows_group; g ++) {

  for (sfb = 0; sfb <num_sfb + num_osf_sfb; sfb ++) {

    if ((cur_bp [g] [sfb]> = 0) && (lazy_bp [g] [sfb] <= 0))

      {

        width = swb_offset [g] [sfb + 1] swb_offset [g] [sfb];

for (win = 0; win <window_group_len [g]; win ++) {

res [g] [sfb] [win] [bin] = 0;

pos = 0;

* for (bin = 0; bin <width; bin ++) {

if (! is_lle_ics_eof ()) {

/ * decoding of binary string and reconstructing res * /

while (decode (freq_silence [pos]) == 1) {

res [g] [sfb] [win] [bin] ++;

pos ++;

if (pos> 2) pos = 2;

if (res [g] [sfb] [win] [bin] == (1 << (max_bp [g] [sfb] +1))-1) break;

}

/ * decoding of sign of res * /

if (! is_sig [g] [win] [sfb] [bin]) && res [g] [sfb] [win] [bin]) {

res [g] [sfb] [win] [bin] * = (decode (freq_sign))? -1: 1;

is_sig [g] [win] [sfb] [bin] = 1;

}

}

else smart_decoding_low_energy ();

}

     }

   }

 }

}

Arithmetic decoding of the truncated SLS bitstream according to the present invention provides an efficient method of decoding an intermediate layer corresponding to a given target bit rate. This is based on the fact that even if there are no bits in the decryption buffer, the decryption buffer still contains meaningful information. The decoding process continues up to the point where there is no ambiguity in the symbol. The following pseudo code represents an algorithm for detecting an ambiguity in an arithmetic decoding module. The variable num_dummy_bits represents the number of bits that were not put in the value buffer due to truncation.

int ambiguity_check (int freq)

{

/ * if there is no ambiguity, returns 1 * /

/ * otherwise, returns 0 * /

upper = 1 << num_dummy_bits;

decisionVal = ((high-low) * freq >> PRE_SHT) -value + low-1;

if (decisionVal> upper || decisionVal <0) return 0;

else return 1;

}

smart_decoding_cbac_bpgc () or smart_decoding_low_energy () is performed when num_dummy_bits is greater than zero. To prevent sign bit errors, the spectral value of the current spectral line is set to zero when an ambiguity occurs while decoding the sign bit. In the arithmetic decoding process according to the present invention, all index variables are performed from a previous arithmetic decoding process.

smart_decoding_cbac_bpgc ()

{

  / * BPGC / CBAC normal decoding with ambiguity detection * /

  while ((max_bp [g] [sfb]-cur_bp [g] [sfb] <LAZY_BP) && (cur_bp [g] [sfb]> = 0)) {

    for (; g <num_windows_group; g ++) {

      for (; sfb <num_sfb; sfb ++) {

        if ((cur_bp [g] [sfb]> = 0) && (lazy_bp [g] [sfb]> 0)) {

           width = swb_offset [g] [sfb + 1]-swb_offset [g] [sfb];

           for (; win <window_group_len [g]; win ++) {

for (; bin <width; bin ++) {

if (interval [g] [win] [sfb] [bin]> res [g] [win] [sfb] [bin] + (1 << cur_bp [g] [sfb])

{

freq = determine_frequency ();

if (ambiguity_check (freq)) {

/ * no ambiguity for arithmetic decoding * /

res [g] [win] [sfb] [bin] + = decode (freq) << cur_bp [g] [sfb];

/ * decode bit-plane cur_bp * /

if ((! is_sig [g] [win] [sfb] [bin]) && (res [g] [win] [sfb] [bin])) {

/ * decode sign bit of res if necessary * /

if (ambiguity_check (freq)) {

res [g] [win] [sfb] [bin] * = (decode (freq_sign))? 1: -1;

is_sig [g] [win] [sfb] [bin] = 1;

}

else {

/ * discard the decoded symbol prior to sign symbol * /

res [g] [win] [sfb] [bin] = 0;

terminate_decoding ();

}

}

}

else terminate_decoding ();

}

}

}

cur_bp [g] [sfb]-; / * progress to next bit-plane * /

        }

      }

    }

  }

}

smart_decoding_low_energy ()

{

  / * low energy mode decoding * /

  for (; g <num_windows_group; g ++) {

     for (; sfb <num_sfb + num_osf_sfb; sfb ++) {

       if ((cur_bp [g] [sfb]> = 0) && (lazy_bp [g] [sfb] <= 0))

       {

          width = swb_offset [g] [sfb + 1] swb_offset [g] [sfb];

          for (; win <window_group_len [g]; win ++) {

             res [g] [sfb] [win] [bin] = 0;

               pos = 0;

for (; bin <width; bin ++) {

while (1) {

/ * if ambiguity check is false, discard the spectrum is set to be 0 * /

if (! ambiguity_check (freq)) res [g] [sfb] [win] [bin] = 0, terminate_decoding ();

tmp = decode (freq_silence [pos]);

if (tmp == 0) break;

res [g] [sfb] [win] [bin] ++;

pos ++;

if (pos> 2) pos = 2;

if (res [g] [sfb] [win] [bin] == (1 << (max_bp [g] [sfb] +1))-1) break;

}

/ * decoding of sign of res * /

if (! is_sig [g] [win] [sfb] [bin]) && res [g] [sfb] [win] [bin]) {

/ * if ambiguity check is false, the current spectrum value is set to be 0 * /

if (! ambiguity_check (freq)) res [g] [sfb] [win] [bin] = 0, terminate_decoding ();

res [g] [sfb] [win] [bin] * = (decode (freq_sign))? -1: 1;

is_sig [g] [win] [sfb] [bin] = 1;

}

}

             }

          }

       }

    }

}

On the other hand, in the truncated bitstream generation method by re-parsing, when the size of the bitstream is transmitted to the header, the truncated bitstream is reparsed as follows to perform bitstream truncation.

10 is a flowchart illustrating a method of cutting a bitstream of scalable audio data according to the present invention. A method of cutting a bitstream of the scalable audio data will be described with reference to FIG. 10.

First, the bitstream length is interpreted from the bitstream header information. (Step 1000) The bitstream is read. (Step 1010) By using Equation 3 and Equation 4, the byte corresponding to the target bitrate is calculated. (Step 1020) The target bit rate may be given externally such as a server or a user.

target_bits = (int) (target_bitrate / 2. * 1024. * osf / sampling_rate + 0.5) -16

target_bytes = (target_bits + 7) / 8;

Modify the bitstream length with the obtained target byte. In other words, the bitstream length is the bitstream length, which is the smaller of the actual number of bits and target_bytes. (Step 1030) The bitstream is stored and transmitted as much as the target length.

A method of truncating the bitstream by re-parsing the bitstream will be described in more detail as follows. The SLS bitstream can be truncated at a given target bit rate in a simple way. Correction to the value of lle_ics_length does not affect the LLE decoding result in the bitstream before the truncation point. lle_ics_length is independent from the LLE decoding procedure. Bitstream truncation may be performed as follows. Read lle_ics_length from the bitstream. Read the LLE bitstream. Compute the useful frame length at a given target bit rate. The simplest way to calculate the useful frame length is as follows.

target_bits = (int) (target_bitrate / 2. * 1024. * osf / sampling_rate + 0.5) -16

target_bytes = (target_bits + 7) / 8

Here, the variable target_bitrate is a target bit rate having a bit / sec unit. The variable osf represents the oversampling factor. The variable sampling_rate represents the sampling frequency in Hz of the input audio signal.

Update lle_ics_length by taking the smaller of the useful frame length and the current frame length.

lle_ics_length = min (lle_ics_length, target_bytes)

Create a truncated bitstream with the updated lle_ics_length .

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 disk, optical data storage, 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 scalable audio data arithmetic decoding method and apparatus according to the present invention, scalability is applied in arithmetic coding applied in MPEG-4 scalable lossless audio coding. When decoding efficiently.

Even when the bitstream is truncated, the end point of decoding can be known, and additional decoding can be performed on the cut-out part.

Claims (23)

A method of decoding scalable audio arithmetic encoded symbols, the method comprising: (a) performing arithmetic decoding using a symbol to be decoded and a probability of generating the symbol; And (b) checking whether or not to continue decoding by checking an ambiguity indicating whether the symbol has been decoded; Step (b) is After decoding the valid bit stream remaining after truncation for scalable decoding and decoding using a bit bit for decoding the bit stream cut out for scalability, If the symbol is decoded irrespective of the virtual data, decoding continues, and if the symbol to be decoded depending on the virtual data is decoded, determining that an ambiguity has occurred and completing decoding. A scalable audio data arithmetic decoding method comprising a. The method of claim 1, wherein step (b)  (b1) when the right side values of Equations 1 and 2 are K, determining whether to continue decoding the symbol according to K by calculating K; [Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates the upper and lower limits of the range in which the probability value of the symbol exists.) (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; and (b3) determining that an ambiguity has occurred if K is between 0 and 2 ^dummy −1, and stopping decoding. The method of claim 2, wherein before step (a) Finding a symbol to decode; And And calculating a probability of the symbol. 4. The method of claim 3, wherein calculating the probability of the symbol Finding a decoding mode from header information of a bitstream to be decoded; And And if the decoding mode is a context-based arithmetic encoding mode (cbac), obtaining a probability of the symbol by referring to the context of the symbol to be decoded. The method of claim 4, wherein step (a) And a first non-zero sample of a value on the bitplane is decoded, and arithmically decodes a corresponding sign bit. 4. The method of claim 3, wherein calculating the probability of the symbol Finding a decoding mode from header information of a bitstream to be decoded; And And obtaining a probability of a symbol to be decoded on the assumption that the data to be decoded has a Laplacian distribution when the decoding mode is the bitplane gollum coding mode (bpgc). The method of claim 6, wherein step (a) When the first non-zero sample of the value on the bitplane is decoded, the corresponding sign bit is arithmetic decoded. Step (b3) is When K is between 0 and 2 ^dummy -1, it determines that an ambiguity has occurred, stops decoding, and sets the immediately decoded sample to 0 to terminate decoding. Way. The method of claim 2, wherein calculating the probability of the symbol Finding a decoding mode from header information of a bitstream to be decoded; And And if the decoding mode is a low energy mode, obtaining a probability of a symbol to be decoded using the probability model information of a bitstream header. An apparatus for decoding scalable audio arithmetic encoded symbols, the apparatus comprising: A symbol decoding unit for arithmetically decoding a symbol using a symbol to be decoded and a probability thereof; And It characterized in that it comprises an ambiguity check unit for determining whether to continue decoding by checking the ambiguity, The ambiguity check unit Decoding continuity determining unit for determining whether to continue decoding the symbol according to K by calculating the K when the right side value of the equation (1) and (2) K; [Equation 1]

[Equation 2]

Where v1 is the value of the valid bitstream remaining after truncation, v2 is the bitstream value truncated after truncation, dummy is the number of bits in v2, freq is the probability value of the symbol, and high and low are the probability values of the symbol. Indicates the upper and lower limits of the range.) An additional decoding unit to decode a symbol to 1 when K is 2 ^dummy -1 or more, and to decode a symbol to 0 when K is 0 or less; And And a decoding stop part which determines that an ambiguity has occurred and stops decoding when the K is between 0 and 2 ^dummy -1. 10. The method of claim 9, And a symbol determination / probability prediction unit for finding a symbol to be decoded and calculating a probability of the symbol. delete A method of decoding scalable audio arithmetic encoded symbols, the method comprising: (a) performing arithmetic decoding using a symbol to be decoded and a probability of generating the symbol; And (b) checking whether or not to continue decoding by checking an ambiguity indicating whether the symbol has been decoded; Step (b) is (b1) when the right side values of Equations 1 and 2 are K, determining whether to continue decoding the symbol according to K by calculating K; [Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates the upper and lower limits of the range in which the probability value of the symbol exists.) (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; and (b3) determining that an ambiguity has occurred if K is between 0 and 2 ^dummy -1, and completing decoding. (a) Arithmetic decoding using a symbol to be decoded and the probability of occurrence of the symbol, wherein the probability calculation of the symbol finds a decoding mode from header information of a bitstream to decode, and the decoding mode is a context-based arithmetic encoding mode ( cbac), obtaining a probability of the symbol by referring to the context of the symbol to be decoded; And (b) checking whether or not to continue decoding by checking an ambiguity indicating whether the symbol has been decoded; Step (b) is (b1) when the right side values of Equations 1 and 2 are K, determining whether to continue decoding the symbol according to K by calculating K; [Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates the upper and lower limits of the range in which the probability value of the symbol exists.) (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; and (b3) determining that an ambiguity has occurred if K is between 0 and 2 ^dummy −1, and completing decoding. The method of claim 13, wherein step (a) And a first non-zero sample of a value on the bitplane is decoded, and arithmically decodes a corresponding sign bit. (a) Arithmetic decoding using a symbol to be decoded and a probability of generating the symbol, wherein the probability calculation of the symbol finds a decoding mode from header information of a bitstream to be decoded, and the decoding mode is a bitplane gollum encoding mode. (bpgc), assuming that data to be decoded has a Laplacian distribution, obtaining a probability of a symbol to be decoded; And (b) checking whether or not to continue decoding by checking an ambiguity indicating whether the symbol has been decoded; Step (b) is (b1) when the right side values of Equations 1 and 2 are K, determining whether to continue decoding the symbol according to K by calculating K; [Equation 1]

[Equation 2]

(Where v1 is the truncated bitstream for scalability, the remaining valid bitstream value, v2 is the truncated bitstream value after truncation, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low Represents the upper and lower limits of the range in which the probability value of the symbol exists.) (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; and (b3) determining that an ambiguity has occurred if K is between 0 and 2 ^dummy −1, and completing decoding. The method of claim 15, wherein step (a) And if the first non-zero sample of the value on the bitplane is decoded, arithmetic decoding of a corresponding sign bit is performed. (a) Arithmetic decoding using a symbol to be decoded and a probability of generating the symbol, wherein the probability calculation of the symbol finds a decoding mode from header information of a bitstream to be decoded, and the decoding mode is a low energy mode. In the case of, obtaining a probability of a symbol to be decoded using the probability model information of the bitstream header; And (b) checking whether or not to continue decoding by checking an ambiguity indicating whether the symbol has been decoded; Step (b) is (b1) when the right side values of Equations 1 and 2 are K, determining whether to continue decoding the symbol according to K by calculating K; [Equation 1]

[Equation 2]

(Where v1 is the truncation of the bitstream for scalability, the value of the remaining valid bitstream, v2 is the truncated bitstream value, dummy is the number of bits in v2, freq is the probability value of the symbol, high, low indicates the upper and lower limits of the range in which the probability value of the symbol exists.) (b2) decoding the symbol as 1 when the K is 2 ^dummy −1 or more, and decoding the symbol as 0 when the K is 0 or less; and (b3) determining that an ambiguity has occurred if K is between 0 and 2 ^dummy −1, and completing decoding. delete A computer-readable recording medium having recorded thereon a program for executing the invention according to any one of claims 1 to 8 and 12 to 17 on a computer. The method of claim 4, wherein step (b3) When K is between 0 and 2 ^dummy -1, it is determined that an ambiguity has occurred and the decoding is stopped, and the decoding is terminated by setting the immediately decoded sample to 0 to terminate decoding. Way. The method of claim 13, wherein step (b3) When K is between 0 and 2 ^dummy -1, it is determined that an ambiguity has occurred and the decoding is stopped, and the decoding is terminated by setting the immediately decoded sample to 0 to terminate decoding. Way. The method of claim 15, wherein step (b3) When K is between 0 and 2 ^dummy -1, it is determined that an ambiguity has occurred and the decoding is stopped, and the decoding is terminated by setting the immediately decoded sample to 0 to terminate decoding. Way. A computer-readable recording medium having recorded thereon a program for executing the invention according to any one of claims 20 to 22.

Images