Speech Recognition Algorithm Simulation And Software Design Based On DTW And HMM

In underwater voice communication,using speech recognition and speech synthesis to perform source coding and decoding to voice signal is able to reduce the data volume and transmission rate greatly.Speech recognition technology is the key to this method of low byte rate underwater voice communication.In order to provide a flexible and efficient speech recognition solution for the method of low byte rate underwater voice communication,this thesis systemically studies various common techniques involving in a small vocabulary speech recognition system.Then a software of small vocabulary isolated words speech recognition system is compiled.The work included in this thesis can be divided into the following 5 parts:1. Introduce the basic conceptions and principles involving in the speech recognition systemjncluding the pretreatment of speech signal,the technology of the characteristic parameter extracting,template matching and technology of training of model .2. Create a digital voice database for training and testing.3. Study the end-point detection algorithm,three characteristic extraction algorithms and two pattern recognition algorithms including dynamic time warping and hidden markov model.Then these algorithms are implemented in MATLAB language.4. Compile a software of a small vocabulary isolated words speech recognition system which is based on the Windows operating system.All the algorithms which are used in the software is implemented in C program language.5. Do Substantial experiment about different values of the key parameter,and record the statistical recognition rate.Then this thesis discusses the influence on recognition rate while the key parameter is set to different value.