Comparison of Several Classifiers for Emotion Recognition from Noisy Mandarin Speech

Abstract

Automatic recognition of emotions in speech aims at building classifiers for classifying emotions in test emotional speech. This paper presents an emotion recognition system to compare several classifiers from clean and noisy speech. Five emotions, including anger, happiness, sadness, neutral and boredom, from Mandarin emotional speech are investigated. The classifiers studied include KNN WCAP GMM HMM and W-DKNN. Feature selection with KNN was also included to compress acoustic features before classifying the emotional states of clean and noisy speech. Experimental results show that the proposed W-DKNN outperformed at every SNR speech among the three KNN-based classifiers and achieved highest accuracy from clean speech to 20dB noisy speech when compared with all the classifiers.