Application Of Intelligent Analysis Method In Sound Recognition

With the development of big data,more and more multimedia data is filling our life.As an important part of multimedia data,sound contains a lot of information.By processing and analyzing the collected sound data,we can analyze and excavate the information that is useful to us.So the voice signal processing and analysis have been the focus of scholars at home and abroad,and voice recognition,as an important application direction of sound signal processing and analysis,has been studied extensively.Voice recognition is to extract the sound characteristics of the sound signal to be recognized and match it with the sample sound characteristics,so as to obtain the judgment of whether the sound is consistent with the sample sound.Voice recognition is widely used in many fields such as speaker identification,audio data retrieval,abnormal sound detection and so on.The research on voice recognition includes pretreatment,feature extraction and pattern matching.This paper focuses on the research of feature extraction and pattern matching.Firstly,pre-process the sound signal,and then the feature extraction of sound signal has been studied extensively.This paper describes the extraction method of short-term energy,linear prediction cepstral coefficients(LPCC)and mel frequency cepstral coefficients MFCC.A combination parameter extraction algorithm based on correlation distance Fisher ratio is proposed.And the short-time energy,linear prediction cepstral coefficients(LPCC),MFCC and the extraction of composite parameters were realized by simulation.Secondly,in order to improve the recognition rate of voice recognition system,after the comparison of the different methods of intelligent analysis,this paper adopts an improved Support Vector Machine(Support Vector Machine,SVM)pattern recognition methods.Because compared with other pattern recognition methods,SVM adopts the optimal classification hyperplane with the largest classification interval for classification,thus minimizing the structural risk.Kernel function is the core mechanism of support vector machine model,kernel function type and parameter selection are very important to support vector machine classification accuracy.In this paper,genetic algorithm is adopted to improve the selection of the type and parameter of SVM kernel function,so that SVM parameters are optimized,so as to improve the recognition rate of voice recognition.