A 2 kb/s enhanced waveform interpolation speech coder

In this paper, a 2 kb/s waveform interpolation speech coding scheme is described based on the EWI speech coding model proposed by Oded Gottesman. The algorithm produces high quality speech with the improved quantization methods for encoder parameters. In this 2 kb/s enhanced waveform interpolation speech coding, the rapidly evolving waveform (REW) magnitude is represented by the 2-D parametric surface and quantized with the analysis-by-synthesis (A-b-S) approach, the slowly evolving waveform (SEW) is optimized and quantized with A-b-S method as well, and the power is quantized with switched-predictive A-b-S method. The computer simulation results and subjective tests show that the coded speech quality exceeds that of federal standard 2.4 kb/s MELP coder greatly.