IIR AnalysisjSynthesis Systems
Mark J. T. Smith · 1991
A subband coding system may be convenientlY viewed as having two con stituent components: the analysis/synthesis section pair and the co ding sec tion pair. The term analysis describes the process of splitting the input into critically-sampled frequency related subband signals while the term synthesis refers to the dual operation of interpolating and merging the signals to recon struct the input. The coding section pair, which consists of an encoder and decoder, appears between the analysis and synthesis operations and enables the input to be represented at a reduced bit rate. Due to the dissociate nature of the constituent section pairs, these sub systems have traditionally been treated separately in the literat ure. In both one-dimensional (I-D) and two-dimensional (2-D) signal applications a plethora of articles may be found on theoretical issues related to band splitting and merging [5], [51], [26], [35], [7], [53], [30]. Similarly, many papers concerned with basic coding can by cited which later have been applied to coding sub bands [18], [10], [15]. Unquestionably, both section pairs are critical since dis tortions introduced by either will contribute to the overall system error. Each of these section pairs contains a set of unique issues and problems that must be addressed effectively in order to assure high quality operation. Discussion of quantization, bit allocation and other related coding issues may be found in various sources, including [28], [27], [8], and Chapters 5 through 80fthis book. However, this chapter is devoted exclusively to the analysis/synthesis sections, with specific emphasis on systems which employ recursive filter banks.