Using Speech Signal for Emotion Recognition Using Hybrid Features with SVM Classifier

Abstract

Emotion recognition is a hot topic that has received a lot of attention and study,owing to its significance in a variety of fields, including applications needing human-computer interaction (HCI). Extracting features related to the emotional state of speech remains one of the important research challenges.This study investigated the approach of the core idea behind feature extraction is the residual signal of the prediction procedure is the difference between the original and the prediction .hence the visibility of using sets of extracting features from speech single when the statistical of local features were used to achieve high detection accuracy for seven emotions. The proposed approach is based on the fact that local features can provide efficient representations suitable for pattern recognition. Publicly available speech datasets like the Berlin dataset are tested using a support vector machine (SVM) classifier. The hybrid features were trained separately. The results indicated that some features were terrible. Some were very encouraging, reaching 99.4%. In this article, the SVM classifier test results with the same tested hybrid features that published in a previous article  will be presented, also a comparison between  some related works  and the proposed technique  in speech emotion recognition techniques.