Implementing Speech-Recognition Algorithms on the TMS320C2xx Platform
This application report presents the hidden-Markov-model-based speech-recognition system using a 16-bit fixed-point digital signal processor (DSP) from Texas Instruments (TI). It addresses the problems that have been encountered and the compromises that have been made to implement the hidden-Markov-model-based isolated-word recognizer using the TI TMS320C2xx DSP. This report addresses fixed-point representation issues and techniques used to prevent overflows. All recognizer functions, including signal processing, model evaluation, and control, are performed by a single TMS320C2xx DSP.
|
Samples & Purchase Cart
