Statistical Feature Set Calculation using Teager Energy Operator on Emotional Speech Signals

In this paper a novel, modified statistical feature vector for classification of various emotions from speech signals based on Teager Energy Operator-Mel Frequency Cepstral Coefficients (TEO-MFCC) is calculated. The speech signal is processed through the Teager Energy operator, local properties of signal are extracted and Mel Frequency Cepstral Coefficient feature is extracted from the Teager Energy operated signal. The basic statistical features like mean, standard deviation, mean absolute deviation and inter-quartile range are calculated on the extracted MFCC from various emotional speech signals. Since the feature vector is minimized into statistical feature set, complexity and computation is reduced.