Entropy-constrained trellis coded quantization and jointly designed trellis coded quantization/modulation /
| Main Author: | |
|---|---|
| Other Authors: | , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1990.
|
| Subjects: | |
| Online Access: | ProQuest, Abstract Link to OAKTrust copy |
| Abstract: | The ideas of trellis coded quantization (TCQ) and entropy encoding are used to develop a new source coding technique called entropy-constrained trellis coded quantization (ECTCQ). Based on the entropy-constrained vector quantization and generalized Lloyd design algorithms, a design algorithm for ECTCQ is developed. ECTCQ systems are designed for encoding memoryless Gaussian and Laplacian sources. Using an 8-state Ungerboeck amplitude modulation trellis defined by a rate-1/2 convolutional code and a 4 subset-partitioned codebook, the ECTCQ signal to quantization noise ratio (SQNR) performance is within about 0.5 dB and 0.6 dB, respectively, (corresponding to 0.08 bits/sample and 0.1 bits/sample) of the best optimum encoding performance theoretically attainable. The implementation requirements are slightly higher than that of entropy-constrained scalar quantization (ECSQ), but the performance improves by 0.145 bits/sample and 0.125 bits/sample for memoryless Gaussian and Laplacian sources, respectively. ECTCQ and predictive coding are combined to form predictive entropy-constrained trellis coded quantization (PECTCQ) for encoding sources with memory. Encoding systems are designed for the first and second order Gauss-Markov sources, using an 8-state trellis. For the first order Gauss-Markov source, the SQNR performance gap between PECTCQ and the rate distortion function is about 0.85 dB and 0.7 dB for correlation coefficients of 0.9 and 0.8, respectively. For the second order Gauss-Markov source with McDonald's coefficients for speech (a1 = 1.515 and a2 = $-$0.752), the SQNR performance differs from the rate distortion function by about 0.8 dB. The ideas of entropy-constrained trellis coded quantization are extended to multi-dimensional source coding to form entropy-constrained trellis coded quantization (ECTCVQ) to achieve encoding rates of less than 1 bit/sample. Using an 8-state trellis and an 8 subset-partitioned codebook. ECTCVQ can achieve similar improvements over ECSQ for encoding rate less than 1 bit/sample. Trellis coded quantization of memoryless sources is developed for transmission over a binary symmetric channel... |
|---|---|
| Item Description: | Typescript (photocopy). Vita. "Major subject: Electrical engineering." |
| Physical Description: | xv, 137 leaves : illustrations ; 29 cm |
| Bibliography: | Includes bibliographical references. |