Data processing: speech signal processing – linguistics – language – Speech signal processing – For storage or transmission
Reexamination Certificate
2001-12-12
2003-12-09
Chawan, Vijay (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
For storage or transmission
C704S503000, C375S341000, C327S018000
Reexamination Certificate
active
06662154
ABSTRACT:
FIELD OF THE INVENTION
This invention generally relates to encoding and decoding an information signal. In particular, the invention relates to a method for information signal encoding and decoding using variable-length combinatorial and Huffman codes.
BACKGROUND OF THE INVENTION
Methods for coding an information signal form the basis for a significant amount of speech, audio, image and video transmissions through communication channels or from storage devices. Digital speech encoding standards for code division multiple access (CDMA) mobile phone systems, for example, are described in the approved specification (ANSI/TIA/EIA-95-B-1999) entitled “Mobile Station—Base Station Compatibility Standard for Wideband Spread Spectrum Cellular Systems” (1999), published by the Electronic Industries Association (EIA), 2500 Wilson Blvd., Arlington, Va., 22201. A variable rate speech codec, and specifically a Code Excited Linear Prediction (CELP) codec for use in communication systems compatible with IS-95, is defined in the document (TIA/EIA/IS-127) entitled “Enhanced Variable Rate Codec, Speech Service Option 3 for Wideband Spread Spectrum Digital Systems” (1997), published by the Electronics Industries Association (EIA), 2500 Wilson Blvd., Arlington, Va., 22201.
A method for encoding a speech signal using factorial packing (FP) is described in “Factorial Packing Method and Apparatus for Information Coding”, U.S. Pat. No. 6,236,960 by Peng et al., issued May 22, 2001. This speech coder utilizes four characteristics to uniquely describe any given pulse combination: number of degenerate pulses, signs of pulses, positions of non-zero pulses, and pulse magnitudes. A four-stage iterative classification of the pulse combination is performed, starting with the number of non-zero pulses and then determining the remaining parameters. The most significant bits in FP have most of the information about the number of non-zero pulses, while the least significant bits have primarily sign information showing a partial segregation of constituent information in FP. There is no complete segregation in this method, and therefore an error in the central bits does not always ensure that the number of degenerate pulses can be decoded correctly. Factorial packing is an example of an enumeration method, where indices are assigned to the pulse configurations. The codeword that is generated with this method, however, has no well-defined bit fields that represent the pulse characteristics.
The pyramid vector quantization (PVQ) method, as described by Fischer, “A Pyramid Vector Quantizer”, IEEE Transactions on Information Theory, Vol. IT-32, July 1986, pp. 568-583, is another example of an enumeration method. The enumeration technique may be very sensitive to bit errors. Hung et al., in the article “Error-Resilient Pyramid Vector Quantization for Image Compression”, IEEE Transactions on Image Processing, Vol. 7, October 1998, pp. 1373-1386, proposed some PVQ enumeration methods that are less sensitive to bit errors than those proposed by Fischer. Two of their proposed enumeration methods, Conditional Product Code (CPC) and Conditional Product-Product Code (CPPC), were found to be robust to bit errors when used for representing Discrete Cosine Transform (DCT) coefficients in image compression. FP may be considered a variant of CPC. FP and CPC achieve robust performance by “partial segregation” of information present in a pulse configuration. The CPPC method has more segregation of constituent information, although it also does not ensure complete segregation. The comparisons between CPPC and CPC indicate that the CPPC method may be more robust to bit errors, suggesting that increasing information segregation may increase robustness. With complete information segregation, none of the bits in the codeword is affected by more than one of the constituents. It would be preferred that addition and multiplication functions would not be used to combine constituent codewords into a complete codeword, as these operations may add to the complexity of the coding process, and also tend to decrease segregation. The codeword should be formed by a concatenation of the constituent codewords, thus maintaining complete information segregation for higher error resilience.
A method for coding an information signal based on parameters inherent to the information signal is described in “Method and Apparatus for Coding an Information Signal”, U.S. Pat. No. 6,141,638, W. Peng and J. Ashley, issued Oct. 31, 2000. The method selects one of a plurality of configurations based on predetermined parameters related to the information signal, each of the plurality of configurations having a codebook; and searches the codebook over the length of an excitation vector which is shorter than a sub-frame length, to determine a codebook index from the codebook corresponding to the selected configuration; and transmits the predetermined parameters and the codebook index to a destination.
A code-excited linear prediction (CELP) technique is described in the paper by James P. Ashley, Edgardo M. Cruz-Zeno, Udar Mittal and Weimen Peng, “Wideband Coding of Speech Using a Scalable Pulse Codebook”, Proceedings IEEE Workshop on Speech Coding 2000, Lake Delavan, Wis., September, 2000. The technique is scalable to a wide range of bit rates. The method improves the coding efficiency of multiple pulse ACELP tracks in wideband speech coding. The method was also shown to be extendable beyond Algebraic Code-Excited Linear Predictive (ACELP) speech coders, such that the standard track constraints are eliminated while still achieving high quality speech.
Three principal figures of merit for speech encoding methods include: the bit rate, the quality of the reconstructed speech, and the amount of computation necessary to encode and decode the speech or similar information signal. Improvements may be made for a fixed bit rate by using improved codes to increase the computational efficiency of the encoder, and incorporating techniques for improved error resiliency of the transmitted codewords to achieve higher quality speech transmission.
It is an object of this invention, therefore, to improve upon the computational complexity of indexed coding and decoding, to provide higher error resiliency associated with factorial packing methods, and to overcome the deficiencies and obstacles described above.
SUMMARY OF THE INVENTION
One aspect of the invention provides a method of coding an information signal. An information signal may be represented by a sequence of pulses. A plurality of pulse parameters based on the sequence of pulses may be determined, with one of the pulse parameters corresponding to the number of non-zero pulse positions in the sequence of pulses. The non-zero pulse parameter may be coded using a variable-length codeword.
Parameters may represent the number of non-zero pulses in the sequence of pulses, the location or positions of the non-zero pulses, the magnitudes of the non-zero pulses, and the phase or signs of the non-zero pulses.
The coded pulse parameters may be assembled into a sequence of codewords. The first codeword may represent the number of non-zero pulses in the sequence of pulses. The non-zero pulse parameter may be coded using a variable-length Huffman code. The pulse position parameter ma be coded using a combinatorial expression related to the total number of available positions and the number of non-zero positions in the sequence of pulses. The combinatorial function may be accessed using a medium memo storage routine (MMSR). The combinatorial function may be accessed using a smallest memory storage routine (SMSR). The pulse magnitude parameter may coded using a combinatorial function related to the sum of the pulse magnitude and the number of non-zero pulse positions in the sequence of pulses. The combinatorial function may be accessed using a medium memory storage routine. The combinatorial function may be accessed using a smallest memory storage routine. The pulse magnitude parameter may contain a predetermined sum of pulse magnitudes in the
Ashley James Patrick
Cruz-Zeno Edgardo Manuel
Mittal Udar
Chawan Vijay
Haas Kenneth A.
Nolan Daniel A.
LandOfFree
Method and system for information signal coding using... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for information signal coding using..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for information signal coding using... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3152458