Code-excited linear predictive speech processing for digital transmission and storage

The authors present an implementation of code excited linear prediction (CELP) in the development of a high-quality speech processing system for digital speech transmission and storage. This system is designed for speech communication systems that utilize low-bit-rate channels or networks, such as packet radio channels. A speech compression scheme is necessary for real-time transmission on such channels. The CELP scheme is chosen for its ability to compress the speech data down to 4.8 kb/s while maintaining the quality of speech. The proposed US Federal Standard of CELP coding FS-1016 has been implemented to maintain the interoperability between the system and the other FS-1016 systems. This version of CELP also provides a forward error protection scheme that is suitable for real-time application. The system is implemented on a low-cost personal computer equipped with a digital signal processor module. The system can be connected to any channels or networks that use RS-2323C as the physical layer link and operate at rates of 4.8 kb/s and up.<
>