Multiresolution joint source and channel coding for wireless communications

Abstract

of thesis entitled Multiresolution Joint Source and Channel Coding for Wireless Communications submitted by Xufang Wang for the degree of Master of Philosophy at the University of Hong Kong in May 2001 Today,transmission over mobile radio channels is still a great challenge due to the severe impairment caused by shadowing, Rayleigh fading,etc. Many Joint Source and Channel Coding (JSCC) schemes have achieved great performance gains and optimal broadcast scenarios were proved to be Multiresolution (MR) in character. Most of MR-JSCC applications, however, considered only the additive white Gaussian noise (AWGN) channel or the system without feedback does not effectively exploit the time vaiying nature of the channels. In this thesis, a novel Multiresolution Joint Source and Channel Coding (MRJSCC) scheme with feedback is presented for mobile radio channels (both shadowing and Rayleigh fading are considered). The transmitter and receiver can match the resolution of each other synchronously, according to the channel state information (CSI). For Gaussian source model, simulation results show that the proposed MR-JSCC scheme with feedback has much better performances than baseline-I ( E E P E q u a l Error Protection, no feedback) and lmseline-II (UEP~Unequal Error Protection, no feedback), as the MR-JSCC system with feedback is able to adjust the optimal resolution according to the variation of SNR. Therefore, a balance between channel errors and quantization noise is always achieved,resulting in a better distortion performance compared with the baseline systems. The resolution threshold determination algorithm is detailed. We then successfully applied the scheme to image transmission, one of the main system bottlenecks of wireless multimedia services. Tree Structured Vector Quantizer (TSVQ) is chosen as the multiresolution image source while the Adaptive Bit-Interleaved Coded Modulation (ABICM) is adopted for channel protection. Simulation results show the substantial gain of the MR-JSCC image over the single resolution image. The multiresolution image is able to avoid catastrophically large errors by choosing lower resolution mode and better protection during the harsh channel condition and get improved by selecting higher resolution mode when channel condition gets better. The image resolution thresholds could be adjusted to meet different bandwidth requirement according to the threshold determination algorithm.