Implementation of Novel Music Player Based on Speech and Text Emotion Recognition for Mood Uplift

Abstract

It is observed that music possesses human mood regulation ability. Establishing congruence between music and mood uplift is an important domain of study for facilitating emotional elevation. The work implements a novel speech and text emotion recognition based music player for mood uplift. The study takes into consideration five important human emotions namely Happy, Sad, Angry, Fear and Bored. The novelty is introduced in the implementation of music player due to its ability to recognize emotions through speech and text inputs provided by the user. Emotion recognition through speech is based on identification of pitch and amplitude of the speech and later subjecting it to algorithmic technique of Probabilistic Neural Network (PNN). Text input is processed using Support Vector Machine (SVM) algorithm for analysing emotions in the text. The working database consists of about 100 songs per emotion in Hindi Language. The most widely used scales to measure mood, Positive and Negative Affect Schedule (PANAS) mood scale has been employed to design the database of around 500 Hindi Songs.  Result reveals accuracy around 95.76% of the music player for recognizing human emotion based on speech and text and playing the songs accordingly for mood uplift. On comparison with other algorithms, it is statistically revealed that the proposed model outperforms the other existing techniques used for speech and text emotion recognition.