Memory vector quantization by power series expansion [in speech coding]
Thomas Eriksson, Fredrik Norden · 2003
Memory quantization is studied in detail. We propose a framework, power series vector quantization (PSVQ), for analysis and development of memory quantizers. Furthermore, we present an optimization algorithm for the framework, and new memory quantization methods are derived from the framework. The new methods are applied to spectrum quantization, and are shown to outperform previous methods.