Paralinguistic and spectral feature extraction for speech emotion classification using machine learning techniques

Abstract Emotion plays a dominant role in speech. The same utterance with different emotions can lead to completely meaning. ability perform various of emotion during speaking is also one the typical characters human. In this case, technology trends develop advanced speech classification algorithms demand enhancing interaction between computer and human beings. This paper proposes approach based on paralinguistic spectral features extraction. Mel-frequency cepstral coefficients (MFCC) are extracted as feature, openSMILE employed extract feature. machine learning techniques multi-layer perceptron classifier support vector machines respectively applied into for emotions. We have conducted experiments Berlin database evaluate performance proposed approach. Experimental results show that achieves satisfied performances. Comparisons clean condition noisy respectively, indicate better scheme.