Speech emotion recognition by mel frequency magnitude coefficient and multistage classification with two step majority voting

Abstract

newlineSpeech emotion recognition is an important research area in the newlinefield of affective computing. In human-machine interaction systems, emotion newlinerecognition from human speech plays a vital role. To effectively identify the newlineemotions, it is necessary to have a better feature and efficient classification newlinemethod. In speech emotion recognition, it is difficult to differentiate anger newlineand happiness emotions since they differ only in valence dimension. Also, newlinefear-surprise and boredom-sadness emotion pairs also closely related. These newlineemotions are difficult to classify by single step classification. While newlineextracting features, extracting highly emotional information is one of the most newlineimportant steps. These problems are addressed in the proposed methods. The newlineproposed methods aim to provide a better recognition rate by a novel spectral newlinefeature namely, Mel frequency magnitude coefficient and a multistage newlineclassification method.