Speech Emotion Recognition Using MFCC and SVM Classification

Abstract

The analysis of human emotions from speech signals plays a major role in human-computer interaction through Speech Emotion Recognition (SER). The proposed system combines MFCCs as feature extraction elements with a CNN for deep feature learning which is then classified with an SVM to boost emotion recognition accuracy. The audio signal processing together with feature extraction operations rely on the Python-based Librosa library. The proposed approach uses the CNN to extract high-level speech data which SVM then classifies into categories effectively. The evaluation of this proposed method shows higher accuracy when implementing it on benchmark emotional speech datasets surpassing traditional MFCC-SVM systems. Deep learning united with SVM produces applications which suit actual usage through better generalization and more robustness for virtual assistant technology alongside sentiment evaluation and medical diagnostic systems. Experimental findings demonstrate that the model functions with high efficiency when identifying emotional patterns which strengthens its capacity for advanced applications in SER.