Deep Learning based Audio Processing Speech Emotion Detection

This paper represents the usage of deep learning model to differentiate the audio emotion classification from a given speech, basically it is called as Speech Emotion Recognition (SER). The intensity of voice is useful to determine the pitch and tone of voice helps to differentiate the emotions based on the audio. In this paper the Multilayer Perceptron model helps in classifying the emotions from the audio. RAVDESS emotional audio speech dataset is used in this work. Feature selection methods are applied to select required features. The data set is more comfortable for extracting features. Model is trained using extracted features and the predicted results are verified based on accuracy parameter. Web application is designed to access the results at user end.