The Key Technology Research Of Isolated Word Speech Recognition

The speech recognition is to make machine to understand what the people say,it is one of the important branches of speech signal processing.At the same time,speech recognition is an interdisciplinary,involving digital signal processing,artificial intelligence,physiology pattern recognition,psychology and cognitive science,and many other fields,it has far-reaching research value.At present,though speech recognition technology has achieved some success,most systems are limited to laboratory,and still didn’t meet to the requirement of practical.The paper discusses the basic principles and various algorithms of speech recognition.First,we use dynamic time warping(DTW)technology to achieve the recognition of speech signals,through the matlab to achieve simulation.Before identification,pre-emphasis the speech signals in order to enhance its part of the high frequency,through the double threshold detection method to achieve the endpoint detection,extract frequency spectrum parameters which reflect ear hearing characteristics,finally,establish of a complete speech recognition system.Due to the advantage of artificial neural networks reflected in speech recognition,put forward a speech recognition method based on a common feed-forward network.Using matlab and VC mixed program,respectively construct the corresponding model for the two algorithms in the speech recognition,get simulation results and experimental analysis.Compared to dynamic time warping techniques,the neural network shows very attraction in terms of classification gathered.