Speech coding is a well researched area, and researchers are making new proposals for im provements to current algorithms. The family of code-excitated linear prediction coders rep resents a recent breakthrough, which almost caters to the present need for low bit rates. In mobile communications, low bit-rate speech coders play a crucial role in spectrally efficient transmission. Theoretically, a rate of a few hundred bits per second is enough to code speech efficiently. Many coders have been de veloped using computationally intensive tech niques that can achieve rates of 4 to 6 kilobits per second. Theoretically, there is room for improvement. The trend is to use analytical models of speech production and perception. This paper briefly reviews existing speech cod ers, identifies future trends in this research area and describes results obtained from simu lating vowels and consonant waveforms and from a multi-pulse excitation/linear predictive speech coders algorithm.