Research On Speech And Emotional Recognition Of Specific Speakers

With the fast development in artificial intelligence, people are not only want thecomputer can play media, but also want it to get the ability of feeling their emotions.Therefore, they paid attention to the research of speech emotion recognition gradually.This article focused on the speech emotion recognition and the research workincluded content as follows:1. A speech emotion database faced to call centers was built. We invited17students to make conversations like operators and clients in call centers. Theyperformed by using telephone according to the script designed. The database has5emotions such as angry, worry, neutral, happy and joyful.2. We generalized many speech emotion features that other researchers used intheir experiments and extracted them. Selecting effective emotional features is veryimportant in recognition. About the global statistical features, we read many papers toget emotional features which usually used and14types of features were extracted,such as pitch, energy and formant. We also calculated their statistical features anddelta-features.3. A feature selection method based on PCA was proposed. For the featuresextracted, we must use a method to assess which of them are good. We proposed afeature selection method based on PCA. First, make transformation using PCA, thenanalyze the transformation matrix and feature selection can be done.4. An emotion recognition system has been constructed. This system can make aspeaker’s emotional speech training and recognizing and a good performance we canget. The system also trained a model that everyone can use it to distinguish theiremotions.