Speech based boredom verification approach for modern education system

Owing to the emotion deficiency problem in many of the conventional education systems, emotion sensing has become a new recent research trend as it is possible to provide useful strategies to enhance the learning effectiveness of a student. Among the various modalities for emotion sensing, this paper a speech-based emotion verification system. In particular, boredom verification is addressed herein. We present an emotion feature set comprising mel-frequency cepstral coefficients (MFCCs), Legendre-based trend coefficients (LBTCs) of MFCCs and 4th subband power, PCA transformed LBTCs, spectral flatness, and RSS. The proposed feature set is fed into a 2-class support vector machine (SVM) for boredom emotion verification. The proposed system has been demonstrated an emotional speech database with a 67.79% verification rate.