Emotion Stream: A CNN-Based Speech Emotion Recognition System for Real-Time Music Recommendation

Emotions expressed in speech play a vital role in how people communicate and interact. Teaching machines to recognize these emotions can make digital systems more adaptive and engaging. This study introduces Emotion Stream, a web-based application that identifies the emotional state of a speaker and suggests suitable music or podcasts in real time. The system relies on Mel Frequency Cepstral Coefficients (MFCCs) to extract meaningful features from audio signals, while a Convolutional Neural Network (CNN) performs multi-class classification of emotions such as happiness, sadness, anger etc. A simple interface allows users to provide voice input, which is analyzed instantly, with results linked to Spotify’s recommendation service for personalized content delivery. Testing indicates that CNN models capture the subtle variations in speech that reflect emotional tone, enabling reliable classification. By combining deep learning with real-time deployment, the application demonstrates how speech emotion recognition can be used to create emotion-aware recommendation systems. The work highlights the growing role of artificial intelligence in enhancing multimedia experiences and offers a framework that could be extended to other human–computer interaction domains.