Institute for Systems Research

In order to conceal errors occurring under bad channel conditions, a novel slicing method and a joint source channel coding scenario that combines RCPC with CRC and utilizes the distortion information to allocate convolutional coding rates are proposed. A new performance index based on JND is proposed and used to evaluate the overall performance at different signal-to-noise ratios (SNR). Our system uses OQPSK modulation scheme.

We propose a new classification scheme, dubbed spectral classification, which uses the spectral characteristics of the image blocks to classify them into one of a finite number of classes. The spectral classifier is used in adaptive image coding based on the discrete wavelet transform and shown to outperform gain-based classifiers while requiring a lower computational complexity. The resulting image coding system provides one of the best available rate-distortion performances in the literature. Also, we introduce a family of multiresolution image coding systems with different constraints on the complexity. For the class of rate-scalable image coding systems, we address the problem of progressive transmission and propose a method for fast reconstruction of a subband-decomposed progressively transmitted image.

Another important problem studied in this dissertation is the transmission of images over noisy channels, especially for the wireless channels in which the characteristics of the channel is time-varying. We propose an adaptive rate allocation scheme to optimally choose the rates of the source coder and channel coder pair in a tandem source-channel coding framework. Also, we suggest two adaptive coding systems for quantization and transmission over a finite-state channel using a combined source and channel coding scheme. Finally, we develop simple table- lookup encoders to reduce the complexity of channel-optimized quantizers while providing a slightly inferior performance. We propose the use of lookup tables for transcoding in heterogeneous networks

We develop DCT-based subpixel motion estimation algorithms without the need of image interpolation as required by conventional methods, resulting in significant reduction in computations and data flow and flexible fully DCT-based coder design because the same hardware can support different levels of required accuracy. Simulation demonstrates comparable and even better performance of the DCT-based approach than BKM-ME in terms of MSE and BPS.

We devise the DCT-based motion compensation schemes via the bilinear and cubic interpolation without any increase in computations to achieve more accurate approximation and better visual quality for motion- compensated pictures. Less computations are required in DCT domain because of sparseness and block alignment of DCT blocks and use of fast DCT to reduce computations.

We begin by quantifying the amount of ﲲesidual redundancy inherent in the LSP's of Federal Standard 1016 CELP. This is done by modeling the LSP's as first and second-order Markov chains. Two models for LSP generation are proposed; the first model characterizes the intra-frame correlation exhibited by the LSP's, while the second model captures both intra-frame and inter-frame correlation. By comparing the entropy rates of the models thus constructed with the CELP rates, it is shown that as many as one-third of the LSP bits in every frame of speech are redundant.

We next consider methods by which this residual redundancy can be exploited by an appropriately designed channel decoder. Before transmission, the LSP's are encoded with a forward error control (FEC) code; we consider both block (Reed- Solomon) codes and convolutional codes. Soft-decision decoders that exploit the residual redundancy in the LSP's are implemented assuming additive white Gaussian noise (AWGN) and independent Rayleigh fading environments. Simulation results employing binary phaseshift keying (BPSK) indicate coding gains of 2 to 5 dB over soft-decision decoders that do not exploit the residual redundancy.

In this dissertation, new algorithmic- level techniques for compensating the increased delays based on the multirate approach are proposed.

Given the digital signal processing (DSP) problems, we apply the multirate approach to reformulate the algorithms so that the desired outputs can be obtained from the decimated input sequences. Since the data rate in the resulting multirate architectures is M- times slower than the original data rate while maintaining the same throughput rate, the speed penalty caused by the low supply voltage is compensated at the algorithmic/architectural level.

This new low-power design technique is applied to several important DSP applications. The first one is a design methodology for the low- power design of FIR/IIR systems. By following the proposed design procedures, users can convert a speed-demanding system function into its equivalent multirate transfer function. This methodology provides a systematic way for VLSI designers to design low- power/high-speed filtering architectures at the algorithmic/architectural level.

The multirate approach is also applied to the low-power transform coding architecture design. The resulting time-recursive multirate transform architectures inherit all advantages of the existing time-recursive transform architectures such as local communication, regularity, modularity, and linear hardware complexity, but the speed for updating the transform coefficients becomes M-times slower.

The last application is a programmable video co-processor system architecture that is capable of performing FIR/IIR filtering, subband filtering, discrete orthogonal transforms (DT) and adaptive filtering for the host processor in video applications. The system can be easily reconfigurated to perform multirate FIR/IIR/DT operations. Hence, we can either double the processing speed on-the-fly, based on the same processing elements, or apply this feature to the low- power implementation of this co- processor.

The methodology and the applications presented in this dissertation constitute a design framework for achieving low-power consumption at the algorithmic/architectural level for DSP applications.

By combining trellis codes (that achieve a significant granular gain) with SVQ, ECSQ, and AECQ, we obtain Trellis-Based Scalar- Vector Quantizer (TB-SVQ), Entropy-Constrained Trellis- Coded Quantizer (ECTCQ), and Pathwise-Adaptive ECTCQ (PA-ECTCQ), respectively. With an 8-state underlying trellis code, these trellis-coded quantization schemes perform about 1.0 dB better than their naive counterparts. There are two approaches that can extend the quantizers (TB-SVQ, ECTCQ, and PA-ECTCQ) for quantizing sources with memory. The first is to combine the predictive coding operation of the Differential Pulse Code Modulation scheme with various quantizers, yielding Predictive TB-SVQ, Predictive ECTCQ, and Predictive PA-ECTCQ, respectively. There is a duality between quantizing sources with memory and transmitting data over channels with memory. Laroia, Tretter, and Farvardin have recently introduced a precoding idea that helps transmitting data efficiently over channels with memory. By exploiting this duality, the second approach combines the precoder with TB-SVQ and ECTCQ to arrive at Precoded TB-SVQ and Precoded ECTCQ, respectively. Simulation results indicate that the porformance of these quantizers are also close to the rate- distortion limit.

The PA-ECTCQ performance has been shown to be robust, in the presence of source scale and, to a lesser extent, shape mismatch conditions. We also considered adjusting the underlying entropy encoder based on the quantized output (which provide some approximate information on the source statistics). The performance of the resulting Shape-Adjusting PA-ECTCQ has been shown to be robust to a rather wide range of source shape mismatch conditions. are also close to the rate-distortion limit.