Research And Implementation Of Speech Recognition Algorithm

With the rapid development of mobile Internet and the popularity of mobile terminal increasingly close to people's daily lives,more natural human-computer interaction becomes very important.The speech recognition technology provides the possibility for the better human-computer interaction,make it more free and more convenient.Although speech recognition technology has now developed more mature,but there is still much research space,to improve the speed recognition system,robustness,and recognition rate.In this paper,the main speech recognition algorithm,using HTK tools to build a small vocabulary continuous non-specific vocal voice recognition system,simulated scenarios restaurant meal.Through the front and rear ends of the process and the system has been improved and optimized,to the extent applicable to the actual situations.This paper introduces the basic principles of the model of speech recognition,a detailed discussion of the process principles and the characteristic parameters of the speech signal extraction process to determine the MFCC as identification parameters.And in-depth study of the HMM of three basic algorithms:forward-backward algorithm,Viterbi algorithm,Baum-Welch algorithm.Next,based on the design of HTK speech recognition system,including a platform and the working environment,as well as specific system implementation,and by filtering to reduce noise and endpoint detection system is optimized.Use HTK toolkit for training speech samples,and trained to use speech recognition engine template identification.