Speech Emotion Verification Using Emotion Variance And Discriminate Scale Frequency Maps

Abstract

The problems of emotion variances and emotion blending are solved using the emotion verification system. In this methodology consists of two parts. The first part, feature extraction is used to lessen the huge input data into smaller data. The second part, emotion verification can automatically verify and authenticate users from their voice or speech. For each sound frames, vital atoms from the Gabor dictionary are selected by the using of vector quantization. Each atom within the dictionary takes the form of a Gabor function which includes frequency, scale etc.Sparse representation is used to convert scale-frequency maps into sparse coefficients to enhance the robustness towards emotion variance.After the feature extraction, a fusion mechanism is applied the subsequent two strategies. The first is the sparse representation verification based on Gaussian modeled residual errors which can be accommodate emotion variance within the voices of speakers. Such a classifier can be reduce emotion variance and improve recognition accuracy. The second, is to involves the use of an indicator, the EAI, for the measuring the degree of blended emotions. The two scores have been calculated they are fused with each other. It also verifies the performance as well as feasibility.