Linear prediction model of speech production with applications to deaf speech training

The linear prediction model of speech production is reviewed and its various formulations are related. Algorithms are presented for efficient fixed-point analysis and synthesis, together with their execution times, on a small computer. The properties of the model are discussed, and the acoustic-tube analogue is developed. This forms the basis of a system currently being used for deaf speech training. The system is described, and results and experience gained from the initial evaluation period are discussed.