Statistical adaptive transform coding of speech signals

The authors consider the adaptive transform coding (ATC) of signals and propose a statistical approach to ATC, called statistical adaptive transform coding (SATC). This scheme uses short-time statistical features of the transform coefficients for the purpose of adaptive bit allocation and adaptive quantization. This set of features is sent to the receiver after quantization. The receiver, in turn, uses this side information in adaptively reconstructing the signal. Vector quantization of the side information is used. It is concluded that the subblock approach makes SATC technique much simpler than previously proposed ATC methods. It is further concluded that the test results for SATC with speech signals are encouraging for low-bit-rate coding applications.