Cycle extraction for perfect reconstruction and rate scalability

A cycle extractor is presented to be used in a speech coder independently from the coding stage. It samples cycle waveforms (CyWs) of the original prediction residual signal at their natural nonuniform rate. It is shown that perfect reconstruction is possible due to the interplay of these properties for two cycle length normalization and denormalization techniques. The coding stage is coupled to the cycle extractor in the analysis stage by an evolving waveform interpolator that may handle several interpolation methods and sampling rates for a variety of fixed and variable rate coders. The description of extraction, evolution interpolation and synthesis stages is cast in discrete time. The upper performance bound is perfect reconstruction while the lower bound is equivalent to conventional waveform interpolation (WI) speech coding.