MUSIC PLAYER BASED ON SPEECH EMOTION USING KNN, SVC, RANDOM FOREST AND GRADIENT BOOSTING

Abstract

In recent years, human-computer interaction systems are gradually entering our lives. Speech Emotion Recognition (SER) plays as one of the key technologies in human-computer interaction systems. SER system is a collection of methodologies that process and classify each signal to detect emotions embedded in them. Emotions are subjective and it plays a vital role in communication. Emotion detection is a challenging task. This study is based on AI and Machine Learning. For SER, the simulations are performed using Ryson Audio-Visual Database of Emotional Speech and Song (RAVDESS). Emotions are classified as happy, neutral and sad. This study presents a comparison of classifiers such as K-Nearest Neighbour (KNN), Support Vector Classifier (SVC), Random Forest and gradient boosting algorithm for recognition of emotions in speech. And a set of features such as MFCC (Mel Frequency Cepstral Coefficients), Chromagram and MEL Spectrogram Frequency (mel) are considered for training the proposed classifiers. Similarly, the feature of music will be extracted using librosa python library and its emotional weight is detected by using multi-class classification of SVM i.e. SVC. SVC classifier generates the probabilistic measure for the music emotion through extracted features, then the mood randomizer compares the both values, i.e. probability value of speech emotion and probability value of music emotion. And based on the equivalency of both values, music is played accordingly.