Research And Application On The Android Platform Of Robust Continuous Speech Recognition

Humans always have a dream that one day they can communication with machine smoothly. The working efficiency of the human will improve greatly, if humans can communication with machine smoothly. The study of speech recognition technology can help humans to realize their dream. Computer is becoming more and more intelligent and miniature with the development of computer technology. And with the advent of mobile Internet, the computer will complete a lot of work of calculation in the cloud, and then how to improve the performance of interaction between human and computer is a particularly significant and important thing. Voice communication will play an increasingly important role in the interaction between the human and machine. The implementation of all is based on the study of speech recognition technology. The work of this thesis is based on the background of this research. With the popularity of mobile Internet, another focus of this thesis is to study how to apply the speech recognition technology in mobile Internet, which can make speech recognition technology to have a widespread use.This thesis has completed the work of the following several aspects based speech recognition technology:1. This thesis studys the speech recognition technology, and the study focuses on the main technology points of the speech recognition, and studys the process of speech recognition, including framing,adding-windows,pre-processing, endpoint detection, feature extraction, model training and voice recognition technology, etc.2. Studying the hidden markov model and three very important algorithms when we using the hidden markov model, including the forward algorithm, the Viterbi algorithm, as well as the Baum-Welch algorithm, they are very important in the speech recognition system based on hidden markov model.3. We have learned and compared a variety of speech recognition technology already existing, and improve the robustness of speech recognition in the HMM by joining the Posterior Probability Model into the HMM model.4. We have implement an improved speech recognition system based on HMM, which can identify a sequence of numbers which are said by anyone.5. Because of the popularity of mobile Internet, smart devices based on Android System become more and more popular. This thesis will transplant the improved speech recognition system into the Android system platform, and finally implement a speech recognition application on the Android system platform.6. We have analysised and compared the efficiency of the speech recognition system based the improved HMM with the speech recognition system based the common HMM in the end.