Speech Emotion Recognition using Machine Learning

Humans communicate most naturally and quickly through speech. Speech Emotion Recognition (SER) attempts to identify human emotions and affective states from speech. This capitalizes on voice tone and pitch, reflecting underlying emotion. The voice signal is a rich source of information and a powerful medium for communicating approach and eliciting emotions. Audio emotion recognition needs feature extraction and classifier training. The feature vector must contain audio signal elements that characterize speaker-specific properties, including tone, pitch, and energy, to train the classifier model to classify an emotion effectively. RAVDESS, an English language open source dataset of male and female acted speech corpus, is carefully trained and tested. We retrieved MFCC coefficients from training dataset audio samples to reflect speaker vocal tract…