ML Based Speech Emotion Recognition Framework For Music Therapy Suggestion System

Emotions are sudden rushes; they are unexpected factors dependent on the living situations. Generally, there are two types of emotions: positive and negative. The most impactful emotions are the negative ones. Nowadays, depression has become a common disorder. But seeking professional help isn't something everyone can afford. To overcome this, many people go through different therapies. Music therapy is the most common type and it helps calm the nerves affected due to social pressure, anxiety, depression etc. In recent era, the interaction between machine and human has become very convenient. Therefore, we have proposed a speech emotion recognition-based framework for a music therapy suggestion system using machine learning. Three emotion classes considered are happy, angry and depressed. Experimentation is carried out on the English language RAVDESS Dataset, augmented RAVDESS dataset, and combined dataset. With the SVM classifier and PCA feature selection algorithm average accuracy for RAVDESS 80.21%, for an augmented dataset 98.09%, while that for Combination (RAVDESS + Augmented) an accuracy of 91.11% is obtained.