Key Techniques Research on a New Vehicular Speech Recognition System

摘要

This paper proposes a new vehicular speech recognition system, which uses the multipLication of the frame energy and frame zero-crossing rate as index quantity to test the speech activity, meanwhile, takes MFCC as speech signal feature vector, and recognizes speech based on HMM speech recognition model. In addition, a new method of anti-noise Speech Recognition is discussed in this article. Through improved iterative Wiener filters together with PUM model, the new method can be more effective to suppress noise interference and improve speech recognition rate.