Patent Grant
7454354
The present invention relates to a hierarchical lossless encoding and decoding technology for digital signals such as of music data, audio data, or the like.
One process for lossless encoding of audio signals such as music signals is a hierarchical lossless encoding process. Data encoded by the hierarchical lossless encoding process is characterized in that it includes low-bit-rate non-lossless encoded data hierarchically as part thereof. A summary of the hierarchical lossless encoding process is described in IEEE Signal Processing Magazine, pp. 21-32, July 2001. The conventional hierarchical lossless encoding and decoding technology will briefly be described below with reference to
In the conventional hierarchical lossless encoding apparatus shown in
In the conventional hierarchical lossless decoding apparatus shown in
The low-bit-rate encoding and decoding process may be a transform encoding and decoding process or a CELP (Code Excited Linear Prediction) encoding and decoding process. The transform encoding and decoding process will not be described in detail below as reference can be made to Proceedings on ICASSP, pp. 1093-1096, April 1990 for details of the transform encoding and decoding process. The CELP encoding and decoding process will not be described in detail below as reference can be made to Proceedings on ICASSP, pp. 937-940, March 1985 for details of the CELP encoding and decoding process. The lossless encoding and decoding process will not be described in detail below as reference can be made to IEEE Signal Processing Magazine, pp. 21-32, July 2001 for details of the lossless encoding and decoding process.
The conventional hierarchical lossless encoding and decoding process is problematic in that if the low-bit-rate decoded signal in the decoding apparatus is not identical to the low-bit-rate decoded signal in the encoding apparatus, then the lossless reproduced signal is not the same as the input signal. This problem occurs primarily when the processing accuracy in the low-bit-rate decoder in the encoding apparatus and the processing accuracy in the low-bit-rate decoder in the decoding apparatus are different from each other. For example, if international standards ISO/IEC 13818-7 (generally known as MPEG-2 AAC) for audio data encoding are used as the low-bit-rate encoding and decoding process, then the processing accuracy in the low-bit-rate decoder is allowed to be of such a level that the decoded signal produced thereby has a decoding error of ±1 at maximum with respect to a reference decoded signal. Therefore, each of the low-bit-rate decoded signals in the encoding apparatus and the decoding apparatus may possibly have a decoding error of ±1 at maximum with respect to the reference decoded signal, and hence the total error of the low-bit-rate decoded signals in the encoding apparatus and the decoding apparatus may possibly reach ±2 at maximum. As a result, it is possible that the lossless reproduced signal derived from the low-bit-rate decoded signal may not be identical to the input signal.
It is an object of the present invention to make a lossless reproduced signal identical to an input signal even when the processing accuracy in a low-bit-rate decoder in an encoding apparatus and the processing accuracy in a low-bit-rate decoder in a decoding apparatus are different from each other.
According to the present invention, there is provided a means for extracting corrective information from a low-bit-rate decoded signal in a hierarchical lossless encoding method of encoding, into lossless encoded data, a differential signal between a low-bit-rate decoded signal which has been produced by encoding an input signal into low-bit-rate encoded data and decoding the low-bit-rate encoded data, and an input signal. There is also provided a means for correcting a low-bit-rate decoded signal based on corrective information into a lossless reproduced signal in a hierarchical lossless decoding method of decoding hierarchical lossless encoded data which include low-bit-rate encoded data and lossless encoded data.
According to the present invention, a corrective information extractor in an encoding apparatus extracts corrective information from a low-bit-rate decoded signal in view of a maximum value that can occur in a low-bit-rate decoder, and a corrector in a decoding apparatus corrects a low-bit-rate decoded signal based on the corrective information. Since the low-bit-rate decoded signal in the encoding apparatus and the low-bit-rate decoded signal in the decoding apparatus are thus made identical to each other, the lossless reproduced signal is made identical to the input signal.
Embodiments of the present invention will be described in detail below with reference to the drawings.
As shown in
As shown in
Operation of the present embodiment will be described below.
Corrective information extractor 107 in hierarchical lossless encoding apparatus 1 extracts corrective information rx in view of a maximum error of low-bit-rate decoder 103 from a low-bit-rate decoded signal that has been calculated by low-bit-rate decoder 103 in the same manner as heretofore, and outputs corrective information rx to output terminal 108. Corrective information rx may employ a value represented by low-order M bits of a sampled value of a low-bit-rate decoded signal. If the absolute value of maximum errors of a low-bit-rate decoded signal in the encoding apparatus and a low-bit-rate decoded signal in the decoding apparatus is represented by E, then M satisfies the following formula:
2E+1≦2M (1)
If the bit width of the sampled value of the low-bit-rate decoded signal is N bits, then 0<M<N. Though the absolute value E of maximum errors is preset in the present embodiment, the absolute value E may be variable depending on the amplitude of the input signal. If the absolute value E is variable, then the setting information of the absolute value needs to be indicated to the decoding apparatus through a transmitting means or a recording medium. The number M of bits of corrective information rx can be smaller as E is smaller as can be understood from the formula (1). Therefore, the number of transmitted bits can be reduced by making E variable depending on the amplitude of the input signal and transmitting corrective information rx having the number of bits depending on E to the decoding apparatus. Low-bit-rate encoded data, lossless encoded data, and corrective information rx that are output from respective output terminals 105, 106, 108 of hierarchical lossless encoding apparatus 1 are inputted respectively to input terminals 200, 202, 206 of hierarchical lossless decoding apparatus 2.
Corrector 207 of hierarchical lossless decoding apparatus 2 corrects a low-bit-rate decoded signal calculated by low-bit-rate decoder 201 in the same manner as heretofore, based on corrective information rx inputted from input terminal 206, and outputs the corrected low-bit-rate decoded signal. The corrected low-bit-rate decoded signal and a differential signal calculated by lossless decoder in the same manner as heretofore are added to each other by adder ADD2, and the sum is output as a lossless reproduced signal to output terminal 205. Corrector 207 operates per sample of the low-bit-rate decoded signal. An example of operation of corrector 207 will be described in detail below with reference to a flowchart shown in
As an initial setting, mt is calculated using the low-order bit number M according to the following equation (2) (step A1):
mt=2M (2)
Value xhigh represented by high-order (N−M) bits of input value x and value xlow represented by low-order M bits of input value x are calculated according to the following equations (3), (4) (step A2):
xhigh=x & mhigh (3)
xlow=x & mlow (4)
where the symbol “&” represents a logical product for bits. mhigh and mlow are mask patterns, and calculated according to the following equations (5), (6):
mhigh=−mt (5)
mlow=mt−1 (6)
Value xlow of low-order M bits of the low-bit-rate decoded signal output from low-bit-rate decoder 201 and corrective information rx inputted from input terminal 206 are compared with each other to check if the inequality (7) shown below is satisfied or not (step A3). Specifically, in step A3, it is checked whether or not a carry occurs on the low-order (M+1)th bit of the low-bit-rate decoded signal due to an error.
xlow+E<rx (7)
If the inequality (7) is satisfied (Yes in step A3), then it is judged that a carry occurs on the low-order (M+1)th bit due to an error, and output value y is calculated according to the following equation (8) (step A5):
y=(xhigh−mt)|rx (8)
where the symbol “|” represents a logical sum for bits. According to the equation (8), the effect of the carry occurring on the low-order (M+1)th bit is eliminated by xhigh−mt, and the low-order M bits are set to rx, making output value y identical to the low-bit-rate decoded signal in the encoding apparatus.
If the inequality (7) is not satisfied (No in step A3), then it is checked whether the inequality (9) shown below is satisfied or not (step A4). Specifically, in step A4, it is checked whether or not a borrow occurs on the low-order (M+1)th bit due to an error.
rx<xlow−E (9)
If the inequality (9) is satisfied (Yes in step A4), then it is judged that a borrow occurs on the low-order (M+1)th bit due to an error, and output value y is calculated according to the following equation (10) (step A7):
y=(xhigh+mt)|rx (10)
According to the equation (10), the effect of the borrow occurring on the low-order (M+1)th bit is eliminated by xhigh+mt, and the low-order M bits are set to rx, making output value y identical to the low-bit-rate decoded signal in the encoding apparatus.
If the inequality (9) is not satisfied (No in step A4), then it is judged that neither carry nor borrow occurs on the low-order (M+1)th bit, and output value y is calculated according to the following equation (11) (step A6):
y=xhigh|rx (11)
According to the equation (11), the low-order M bits of the low-bit-rate decoded signal are set to corrective information rx, making output value y identical to the low-bit-rate decoded signal in the encoding apparatus. Output value y thus calculated is inputted to one of the input terminals of adder ADD2. Operation of corrector 207 has been described above with reference to
Using numerical examples of the above correcting process will be described below in relation to corrections according to the equations (8) and (10). If N=16, M=3, E=2, for example, then xlow ranges from 0 to 7, and rx can take 5 (=2E+1) types of values shown in the following table:
For any of combinations of xlow and rx in the above table satisfy the inequality (7) or (9), e.g., for xlow=0, rx=6 or xlow=7, rx=1, the high-order (N−M) bit value xhigh of the low-bit-rate decoded signal in the decoding apparatus is not identical to the high-order (N−M) bit value of the low-bit-rate decoded signal in the encoding apparatus due to the effect of an error ±E. In order to correct the error, the high-order (N−M) bit value xhigh needs to be corrected (xhigh−mt, xhigh+m) according to the equation (8) or the equation (10). Since corrective information rx represents the low-order M bit value of the low-bit-rate decoded signal in the encoding apparatus and the corrected high-order (N−M) bit value is the same as the high-order (N−M) bit value of the low-bit-rate decoded signal in the encoding apparatus, a signal which is the same as the low-bit-rate decoded signal in the encoding apparatus can be reproduced in the decoding apparatus, using corrective information and the corrected high-order (N−M) bit value. Accordingly, even when the processing accuracy in the low-bit-rate decoder in the encoding apparatus and the processing accuracy in the low-bit-rate decoder in the decoding apparatus are different from each other, the present correcting process makes the low-bit-rate decoded signals in the encoding apparatus and the decoding apparatus identical to each other, thus making the lossless reproduced signal identical to the input signal.
Another embodiment of the present invention will be described below with reference to
Hierarchical lossless encoding apparatus 1a shown in
Hierarchical lossless decoding apparatus 2a shown in
In hierarchical lossless encoding apparatus 1a, corrective information extractor 107 has its input supplied as a delayed input signal calculated by delay compensator 102. Corrective information extractor 107 extracts corrective information (e.g., a value represented by the low-order M bits of the delayed input signal) from the delayed input signal in view of a maximum error that can occur in the low-bit-rate decoded signal calculated by low-bit-rate decoder 103, and outputs the corrective information to output terminal 108.
In hierarchical lossless decoding apparatus 2a, corrector 207 corrects a lossless reproduced signal, which is produced by adding the low-bit-rate decoded signal calculated by low-bit-rate decoder 201 and the differential signal calculated by lossless decoder 203, using corrective information supplied through input terminal 206, in the same manner as with the first embodiment, and outputs the corrected lossless reproduced signal. Since the error in the low-bit-rate decoded signal is the same as the error in the lossless reproduced signal, the lossless reproduced signal becomes identical to the input signal when the lossless reproduced signal is corrected. The second embodiment has been described above.
A third embodiment of the present invention will be described in detail below.
Hierarchical lossless encoding apparatus 1b shown in
Hierarchical lossless decoding apparatus 2b shown in
In hierarchical lossless encoding apparatus 1b, corrective information extractor 109 extracts corrective information expressed by a single bit in view of a maximum error from the low-bit-rate decoded signal calculated by the low-bit-rate decoder 103 in the same manner as with the above embodiments, and outputs single-bit corrective information rb to output terminal 108. Single-bit corrective information rb may be, for example, a value represented by the single Mth bit counted from the low-order bit of a sampled value of the low-bit-rate decoded signal. The low-order bit comprises the first bit. M satisfies the above formula (1).
Corrector 110 corrects the low-order M bits of the low-bit-rate decoded signal output from low-bit-rate decoder 103 by applying a 0 mask thereto, and outputs the corrected low-bit-rate decoded signal.
Adder ADD1b outputs a differential signal representing the difference between the delayed input signal and the corrected low-bit-rate decoded signal. The output differential signal is encoded by lossless encoder 104, which outputs lossless encoded data to output terminal 106.
In hierarchical lossless decoding apparatus 2b, corrector 208 corrects the low-bit-rate decoded signal calculated by the low-bit-rate decoder 201 in the same manner as with the above embodiments, based on single-bit corrective information rb, and outputs the corrected low-bit-rate decoded signal. The corrected low-bit-rate decoded signal is added to a differential signal calculated by lossless decoder 203 in the same manner as with the above embodiments, and outputs the sum signal as a lossless reproduced signal to output terminal 205.
An example of operation of corrector 208 will be described in detail below with reference to a flowchart shown in
As initial settings,
mt=2M,
t1=E−1, and
th=mt−E
are calculated using the low-order bit number M and the maximum error E (step B1).
Value xhigh represented by high-order (N−M) bits of input value x and value xlow represented by low-order M bits of input value x are calculated in the same manner as in step A2 in the above embodiments (step B2).
Then, value xlow represented by low-order M bits is compared with threshold value t1 to check if the inequality (12) shown below is satisfied or not (step B3). Specifically, it is checked whether or not a carry occurs on the low-order (M+1)th bit due to an error.
t1<xlow (12)
If the inequality (12) is not satisfied, then since there is a possibility that a carry occurs on the low-order (M+1)th bit, it is checked whether single-bit corrective information rb is an upper limit (“1”) or not (step B5).
If single-bit corrective information rb is “1” (Yes in step B5), then it is judged that a carry occurs on the low-order (M+1)th bit, and output value y is calculated according to the equation (13) shown below (step B7). Specifically, in step B7, the effect of the carry due to an error is eliminated.
y=xhigh−mt (13)
If single-bit corrective information rb is “0” (No in step B5), then it is judged that neither carry nor borrow occurs on the low-order (M+1)th bit, and output value y is calculated according to the equation (14) shown below (step B8). Specifically, in step B8, a signal produced by setting all the low-order M bits of the low-bit-rate decoded signal output from low-bit-rate. decoder 201 to “0” is output.
y=xhigh (14)
If the answer to step B3 is Yes, then it is checked whether the inequality (15) shown below is satisfied or not (step B4). That is, it is checked whether or not there is a possibility that a borrow occurs on the low-order (M+1)th bit due to an error.
xlow<th (15)
If the inequality (15) is not satisfied, then since there is a possibility that a borrow occurs on the low-order (M+1)th bit, it is checked whether single-bit corrective information rb is a lower limit (“0”) or not (step B6).
If single-bit corrective information rb is “0”), then it is judged that a borrow occurs, and output value y is calculated according to the equation (16) shown below (step B9). Specifically, in step B9, the effect of the borrow is eliminated.
y=xhigh+mt (16)
If single-bit corrective information rb is “1” (No in step B6), then it is judged that neither carry nor borrow occurs on the low-order (M+1)th bit, and output value y is calculated according to the equation (14) (step B8). Operation of corrector 208 has been described above with. reference to
In steps B7, B9, xhigh is corrected for the reasons that as with the first and second embodiments, the high-order (N−M) bit value xhigh of the corrected low-bit-rate decoded signal in the decoding apparatus is not identical to the high-order (N−M) bit value xhigh of the corrected low-bit-rate decoded signal in the encoding-apparatus due to the effect of an error ±E. According to the present correcting process, the corrected low-bit-rate decoded signals in the encoding apparatus and the decoding apparatus are made identical to each other, thus making the lossless reproduced signal identical to the input signal.
The number of bits of corrective information rx in the first and second embodiments, varies depending on maximum error E. According to the third embodiment, the number of bits of corrective information rx is 1 at all times independently of maximum error E, and hence the number of bits required for transmitting the corrective information is reduced.
According to the present invention, as described in detail above, the corrective information extractor in the encoding apparatus extracts corrective information in view of a maximum error that can occur in the low-bit-rate decoder from the low-bit-rate decoded signal, and the corrector in the decoding apparatus corrects the low-bit-rate decoded signal based on the corrective information. Therefore, the low-bit-rate decoded signal in the encoding apparatus and the low-bit-rate decoded signal in the decoding apparatus are made identical to each other, thus making the lossless reproduced signal identical to the input signal.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2002-085705 | Mar 2002 | JP | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/JP03/03513 | 3/24/2003 | WO | 00 | 11/24/2003 |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO03/081196 | 10/2/2003 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 3569934 | Parr, Jr. | Mar 1971 | A |
| 4716565 | Suma | Dec 1987 | A |
| 4803684 | Kozuki et al. | Feb 1989 | A |
| 4888780 | Imazu | Dec 1989 | A |
| 5313472 | Hara | May 1994 | A |
| 5781564 | Peterson | Jul 1998 | A |
| 6125348 | Levine | Sep 2000 | A |
| 6598197 | Peterson et al. | Jul 2003 | B1 |
| 20010050959 | Nishio et al. | Dec 2001 | A1 |
| Number | Date | Country |
|---|---|---|
| H07-015413 | Jan 1995 | JP |
| 8-46517 | Feb 1996 | JP |
| 9-93132 | Apr 1997 | JP |
| 11-251917 | Sep 1999 | JP |
| WO 9953677 | Oct 1999 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20040161043 A1 | Aug 2004 | US |
