Abstract
Emotions play a vital role during verbal communication in our daily life as only the textual information cannot convey the complete information. Emotions in human speech is a complex phenomenon, which vary from person to person based on gender, anger, varying activities and spoken languages. In this work, a novel technique based on artificial neural networks (ANN) is proposed to recognize real-time emotions such as anger, disgust, fear, happiness, sadness and surprise. First, the noise and silence are filtered from recorded speech using adaptive filtering. Secondly, the acoustic and statistical features are extracted from the filtered speech. Set of uncorrelated features are obtained by using principal component analysis (PCA). The input and target features are used to train the feed forward neural network (FFNN), generalized regression neural network (GRNN), Elman network and radial basis feed forward neural network (RBFNN). Performance analysis based on test results indicates that the RBFNN gives better performance and recognition rate than FFNN, GRNN and Elman network.