Research On Emotion Monitoring System On Speech Signal

Emotion recognition is an important part of artificial intelligence,human-computer interaction and so on.The purpose of emotion recognition research is to make the machine more humanized,which can make people's life more intelligent and simple.Speech is one of the most accurate and direct media of human emotion transmission.Speech emotion recognition has become an important branch of speech signal processing,and an imoprtant part of human-computer interacion.Based on the analysis of the background and Research on speech emotion recognition technology,aiming at some key problems of the current speech emotion recognition involved,including the selection of audio signal preprocessing,feature extraction,feature selection,and emotion recognition model based on speech signal research.The optimization of experimental sequential floating forward selection strategy for speech signal feature parameters in this paper,the small scale and classification of speech signal feature subset with high accuracy.on this basis,thie paper proposes an improved KNN speech emotion recognition method,the core of algorithm is the facus from the original feature parameters,and select some different dimension option subset.Then,a single optimal subset uses the distance funcion to solve,and uses the nearest neighbot to measure the emotion neighbor criterion of intitical classification decision,ant the end of the first classification results were obtained to meansure the emotional attributes of voting the object.For the performance of speech emotion recognition strategy is verified,this paper has created a classification model of speech signal based on MATLAB experimental platform and based on the speech emotion recognition Microsoft Azure machine learning platform,through the voice of the corpus in the classification experiment,the classification label accuracy as the evaluation index,obtained satisfactory classification results.This paper studies the feature selection of emotion in speech recognition and classification method,the corresponding algorithm is proposed,and the experiments were carried out to obtain a satisfactory result,and provides a way for the future of online real-time emotion recognition research,has certain the oretical value and application value.