Extraction of speech Emotion Features Using MLP Classifier

Abstract

Speech Emotion Recognition is a thriving research topic. Speech emotion recognition use MLP classifier to categorize the emotions from the speech. This Speech is also used as the medium, through which one can express their feelings and mind state in human to machine interaction. The case is very easy where two humans communicate along with their emotions as by nature, they can recognize each other’s emotions. But for computer, if it can able to understand the human emotions, then it can work even more effective on the work given by user. Emotions like calm, fear, happy, disgust are considered. The emotions of a person will be based on pitch, breath, speech, its loudness and tone, energy, heart rate. Though the emotions of an individual are similar in nature, the understanding and analysis is different. Multilayer Perceptron, a fully connected multi-layer neural network is used to recognize the emotion according to the inputs given. Here RAVDESS dataset is used for input. It takes the MFCC, Chroma, Mel Spectrograph Frequency, which are extracted from the data (audio file) as input, analyses them and visualize the corresponding emotion related to the data.