David H. Crawford and Emmanuel Roy
Motorola

Over the last few years, a great deal of effort has been spent in developing methods of coding digital speech signals for applications such as mobile communications, transmission of speech over the Internet, and mass storage by call centres of recorded telephone conversations. This involves the removal of redundant and irrelevant information in the speech signal, and can lead to reasonably good-quality speech at bit rates lower than 8kbits/s. However, in order to achieve lower and lower bit rates while maintaining an acceptable quality of speech, speech coding algorithms are becoming increasingly sophisticated and more demanding in terms of the number of calculations required per second. Therefore, there is a continuous need for efficient implementation of speech coding algorithms on digital signal processors (DSPs). This paper describes techniques and approaches commonly used to realize such implementations.