An Enhanced Human Speech Emotion Recognition Using Hybrid of PRNN and KNN

In a system comprising of human-machine interaction, the emotion recognition of speech has always been a wide area of research since the machines can never analyze the emotion of a speaker on its own. To recognize the speaker's emotion, numerous systems were developed and tested. In this research study, an enhanced human speech emotion recognition system using a hybrid of PRNN and KNN algorithms is designed. The six basic emotions like neutral, anger, happiness, sadness, surprise and fear over the speech emotions are classified and studied for their accuracy with other previously developed systems. The database for this study is taken as the emotional speech samples of numbers. A cascaded system of Mel Frequency Cepstral Coefficient (MFCC) and Gray Level Co-occurrence Matrix (GLCM) was used for feature extraction process along with a Wiener filter for filtering the noise in speech. Also, a hybrid of Pattern Recognition Neural Network (PRNN) and K-Nearest Neighbour (KNN) is used for prediction accuracy of outcomes. The outcomes are compared with previously developed recognition systems and better efficiency is observed.