For various continuous digital speech signals, a continuous digital speech recognition system based on TMS320C5x evaluation module (EVM) and fixed-point digital signal processor ADSP2181 is implemented. Based on the analysis of continuous probability density hidden Markov model (CDHMM), LPC cepstral coefficients, LPC difference cepstral coefficients, energy normalization coefficients and their difference coefficients are used as speech feature vectors. Viterbi algorithm and Baum-Welch re-estimation method are used for training and recognition, and ADSP2181 is used to implement the speech recognition algorithm. The recognition rate of the system is effectively improved. The time required for implementing each stage is given, and the influence of different speech feature parameters on the recognition rate is compared. In the specific implementation, the issues of noise resistance, fixed-point real-time implementation and identification of people by continuous digital strings are focused on. The experimental results show that the system achieves satisfactory results in ordinary environments, with a correct recognition rate of 93.2%, which provides an important technical approach for its practical application.
You Might Like
Recommended ContentMore
Open source project More
Popular Components
Searched by Users
Just Take a LookMore
Trending Downloads
Trending ArticlesMore