Recognition of Anger and Neutral Emotions in Speech with Different Languages

Abstract

Speech emotion recognition is a very interesting area. It has board applications in man-machine interaction. In this work, the influence of speech features on the recognition of anger and neutral emotions in different languages is studied. And on the other hand, the influence of anger and neutral emotions for classifying male and female gender in different languages is also studied. Four databases in different languages are used to achieve our purpose. These databases are Algerian Dialect Emotional Database (ADED), Berlin Database of Emotional Speech (EMO-DB), Sharif Emotional Speech Database (ShEMO) and Crowd-sourced Emotional Multimodal Actors Dataset (CREAMA-D). The databases are exploited for extracting the features that used in the recognition and classification systems. The features extracted are the pitch, intensity, formants and MFCCs (Mel Frequency Cepstral Coefficients) parameters. The results obtained show us that the use a combination of features improve the performance of recognition in all the databases. It was showed also in the results that the classification of gender classes is influenced by the type of emotion and the language of databases.