Pulse or digital communications – Systems using alternating or pulsating current – Plural channels for transmission of a single pulse train
Reexamination Certificate
1999-12-14
2003-01-07
Le, Amanda T. (Department: 2634)
Pulse or digital communications
Systems using alternating or pulsating current
Plural channels for transmission of a single pulse train
C375S298000, C341S200000, C704S222000, C704S230000
Reexamination Certificate
active
06504877
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to fixed-rate, successively refinable quantizers with improved rate-distortion performance. More specifically, the invention relates to successively refinable Trellis-Based Scalar-Vector quantizers (TB-SVQ) which achieve improved rate-distortion performance and which allow robust transmission of audio and video signals.
2. Description of the Related Art
Modem data communication systems rely on structured vector quantization schemes wherein a set of data points in a transmitted message word is modeled as a constellation of points. The constellation is a subset of coded bits enclosed within a bounded region wherein each encoded bit is assigned an energy level according to its position in the constellation. A trellis-based scalar-vector quantizer (TB-SVQ) is a type of structured vector quantizer scheme that takes advantage of the fact that for a certain class of memoryless sources, the TB-SVQ can approach a rate-distortion limit, as for example a Gaussian or Laplacian distribution. The TB-SVQ technique is therefore quite useful for transmitting data over additive white Gaussian noise (AWGN) channels which are typical data channels for the Internet or, indeed, any transmission system using modems or other direct data lines such as digital subscriber lines (DSL), T
1
lines or other high-speed data links.
Prior encoding techniques have tended to be codebook-based in that they require a memory of codes that can be compared against current incoming data to reconstruct the data word after transmission. Since a memory of codes is used in these systems they are inherently accurate in reproducing the data word, but much slower and less robust than systems that utilize memoryless sources. However, other prior art systems have extended the TB-SVQ scheme to effectively solve the excitation codebook search problem embedded in code excited linear prediction (CELP) speech coders in an effort improve the speed of such systems while maintaining the reliability achieved with the use of a codebook. See C. C. Lee and R. Laroia, “Trellis Code Excited Linear Prediction (TCELP) Speech Coding.” Bell Labs Technical Memorandum 11332-981030-26TM.
Successively refinable source coders have been designed to output bit streams so that rate-scalability can be achieved. In this type of encoding scheme, partial reconstruction of the data is done with the core bit stream and additional approximations to the original signal are obtained by each additional refinement layer. Successive refineability can be achieved by using a hierarchical (multi-stage) coding structure whereby at each stage the residual between the original and the reproduction from the previous stage is quantized using a Trellis Coded Quantizer (TCQ). Use of the TCQ achieves a granular or shaping gain of about 1.53 dB, which is the theoretical upper limit. Unfortunately, the performance of each quantization stage is in general notably inferior to that of a TCQ of an equivalent bit rate.
To improve rate-distortion performance, an alternative similar to the successively refinable scalar quantizers has heretofore been employed. Utilizing this approach, at each stage of the fixed-rate successively refinable trellis coded quantizers (SR-TCQ), each reproduction symbol of the current stage is assigned an embedded alphabet that is confined for use in the subsequent refinement TCQ. This idea has also been applied to design entropy-constrained embedded trellis coded quantizers which perform very close to the rate-distortion boundary. However, the variable-rate nature of these quantizers sometimes causes other practical problems such as buffering control and error propagation.
Apart from the granular or shaping grain mentioned above, vector quantizers are also superior to scalar quantizers in that they achieve a “boundary gain”, which is realized by selecting a codebook boundary which ensures that most of the code-vectors are placed in a high-probability region of the m-space, and a “non-uniform” density gain which results from having the code-vectors closely spaced in higher probability density regions and farther apart in lower probability density regions of the m-space. Although the TCQ can realize a significant granular gain, it makes no attempt to exploit the boundary gain and realizes only some non-uniform density gain by allowing the underlying reproduction alphabet to have non-uniformly spaced levels. It would therefore be useful if a method and apparatus were developed in which vector quantizers could achieve an acceptable granular gain as well as high levels of boundary and non-uniform gain. Such needs have not heretofore been achieved in the art.
Yet other approaches have been proposed to shape the constellation and achieve an optimal m-sphere codebook boundary in an m-dimensional space. See U.S. Pat. No. 5,388,124 to Laroia et al., titled Precoding Scheme For Transmitting Data Using Optimally-Shaped Constellations Over Intersymbol-Interference Channels, the teachings of which are expressly incorporated herein by reference. Laroia et al. introduced the TB-SVQ for memoryless sources. The TB-SVQ achieves a large boundary gain while the underlying trellis code enables it to realize a significant granular gain. Since the TB-SVQ can be derived from a non-uniform scalar quantizer, it can also achieve non-uniform density gain. It would therefore be desirable to exploit the advantages inherent in TB-SVQs to achieve high levels of granular gain as well as high levels of boundary and non-uniform density gain. This has not heretofore been achieved in the art.
SUMMARY OF THE INVENTION
The aforementioned problems are solved, and long felt needs met, by methods of the present invention for designing a successively refinable TB-SVQ for a memoryless source which outputs a signal that can be characterized as a set of digital data for transmission in a communication system. The inventive methods preferably comprise at least a two-stage process for creating a codebook for the source so that the reproduction symbols can be reproduced in a robust manner. In a first stage, a TB-SVQ is applied to the data so that the codebook boundary can be obtained and acceptable boundary and non-uniform density gains can be achieved. In at least one more successive stage, a TCQ is applied so that a high granular or shaping gain is achievable, preferably about 1.53 dB.
The methods of designing successively refinable TB-SVQs for memoryless sources provided in accordance with the present invention reduce the complexity of the resulting TB-SVQs and ensure that highly robust data transmission is achieved. By employing the multiple stage approach recursively on the signal data, rate-distortion of the encoded signals is minimized and multimedia signals such as audio and video signals can be transmitted robustly through the communication system. This tends to reduce computational complexity in the systems employing the inventive methods and greatly improves the efficiency of data transmission in and through the optical communication system. Moreover, by employing the inventive methods, a shaping gain of close to or approaching or substantially 1.53 dB can be achieved, which is the theoretical upper limit of the shaping gain. Such results have not heretofore been achieved in the art.
These and other features of the present invention will become apparent from the following detailed description considered in conjunction with the accompanying drawings. It is to be understood, however, that the drawings are designed solely for purposes of illustration and not as a definition of the limits of the invention, for which reference should be made to the appended claims.
REFERENCES:
patent: 5388124 (1995-02-01), Laroia et al.
Agere Systems Inc.
Le Amanda T.
LandOfFree
Successively refinable Trellis-Based Scalar Vector quantizers does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Successively refinable Trellis-Based Scalar Vector quantizers, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Successively refinable Trellis-Based Scalar Vector quantizers will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3059765