Performance Evaluation of Learning Classifiers for Speech Emotions Corpus using Combinations of Prosodic Features

This paper introduces the speech emotion corpus, a multilingual speech emotion database recorded in the provincial languages of Pakistan: Urdu, Punjabi, Pashto and Sindhi for analyzing the speech emotions present in the recorded speech signals with the four different emotions (Anger, Sadness, Comfort and Happiness). The objective of this paper is to evaluate the performance of the learning classifiers (MLP, Naive Bayes, J48, and SMO) for speech emotion corpus recorded in the provincial languages of Pakistan with different combinations of prosodic features in term of classification accuracy and time taken to build models. The experimental results clearly show that the J48 classifier performs far better than all other classifiers in term of both classification accuracy and model building time. SMO indicates slightly better classification accuracy than Naïve Bayes classifiers whereas; Naïve Bayes exhibit minimum model building time as compared to MLP.