A 1200 b/s Voice Coder Based on Split Vector Quantization of Line Spectral Frequencies

This paper describes the current status of an on-going research and development effort whose objective is a 1200 b/s voice coder with a high degree of speech intelligibility and natural voice quality. The 1200 b/s voice coder is based on tenth-order linear prediction analysis, split vector quantization of line spectral frequencies, differential pitch and gain quantization, and adaptive postfiltering. An improved version of the 1200 b/s coder which employs harmonic deviations based excitation generation is also described, which operates at 1800 b/s during voiced frames and 1200 bts during unvoiced and silent frames. These voice coding technologies are critical components in the development of satellite or terrestrial based mobile and portable communications using miniature handheld transceivers.