作者: C. Tian , S.S. Hemami
关键词:
摘要: This paper introduces a new high-rate analysis of the multiple description scalar quantizer (MDSQ) with balanced descriptions. The provides insight into structure MDSQ, suggesting nonoptimality uniform central cell lengths, as well method to approximate optimal lengths. For both level-constrained and entropy-constrained upper bounds on granular distortion for sources smooth probability density functions (pdfs) are derived under mean-squared error measure, which 0.4 dB lower than previous results. Based insights, universal (UMDSQ) is proposed which, at high rate, can achieve nearly same performance fully optimized MDSQ (ECMDSQ), without requiring extensive training. UMDSQ has only two control parameters, continuum tradeoff points between side distortions be achieved parameters varied.