Emotional Intelligence-Based Music Recommendation System Using Hybrid Deep Neural Network
Bhawani Sankar Panigrahi, K J Hannah Joyce, Tejasri Punna, Baby Keerthana Varala, B. K. Madhavi · 2024
This research focuses on advancing emotion recognition techniques, particularly in predicting arousal and valence from multi-channel signals. This study aims to advance techniques in emotion recognition, with a specific focus on the prediction of arousal and valence from multi-channel signals. The resulting emotional states serve as valuable input, seamlessly integrated into collaborative or content-based recommendation engines to enrich their capabilities with nuanced emotional information. The research contributes significantly to the broader field of affective computing, striving to create more sophisticated and emotionally intelligent systems that effectively cater to user preferences and needs. To achieve this, we propose a hybrid deep neural network architecture that combines Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Feedforward Neural Networks to learn intricate patterns of emotion. Emotions, being affective states reflecting an individual's response to mental stimuli, play a central role in this investigation. The primary objective is facilitating the transition towards neutral or positive emotional states, such as Joy and Trust, particularly in patients. For the identification of targeted emotions, we employ a Convoluted Neural Network (CNN) model, trained on meticulously curated datasets containing annotated emotion-labeled audio clips. These clips transform Mel Frequency Spectrograms (MFS) for effective classification based on the specific emotions they represent.