CRF Continuous Speech Recognition Research And SVM

With the rapid development of computer technology, the computer has become an integral part of day-to-day tools in the way people live, work and learning. People eager to interact with the computer, and look forward to the interaction with the computer more friendly, more natural. This ideal has prompted speech recognition research boom. Currently, there are a variety of voice recognition technologies and has made great achievements in the field of speech recognition. In practical applications, however, still there are many problems, especially in continuous speech recognition, there are still many difficulties and inadequacies. Therefore, in order to achieve a better, more practical significance of continuous speech recognition theory and technology related to the field of speech recognition and the algorithm thought, researchers still need to continue to study, so as to realize the ideal.In this thesis, first give an overview of speech recognition, speech recognition and the status quo at home and abroad, and its classification and the existing difficulties analysis and research; then give a detailed introduction to acoustic base and front-end speech signal analysis and processing, learned how to pretreatment, endpoint detection and feature parameter extraction voice signals collected; then studied three relatively more important recognition algorithms in the field of speech recognition:dynamic time warping (Dynamic Time Warping DTW), HMM (Hidden Markov Model, HMM) and artificial neural network (Artificial Neural Network, ANN), analyze their principles, characteristics and implementation process; And then again with the CRF (Conditional Random Fields, CRF) model and support vector machine (Support Vector Machine SVM) in the analysis of conditions on the basis of these two models, combined with the advantages of both proposed based on the combination of CRF and SVM continuous speech recognition introduction of the overall framework of the system, and specific experimental analysis verify identify the effect of this approach in the speech recognition system, experiments prove the algorithm to improve the correct rate of continuous speech recognition; Finally, the paper carried a summary of prospects and future work to further improve and enhance.