Voice Mood Detector

The Voice Mood Detector is a web-based system designed to analyze speech and recognize human emotions using machine learning and signal processing. Unlike traditional speech-to-text systems, it focuses on tone, pitch, rhythm, and energy to classify emotions such as happy, sad, angry, calm, fearful, and disgust. The system integrates a React and TypeScript frontend with a Flask backend, offering real-time audio processing through feature extraction methods like MFCC, Chroma, and Mel-spectrogram. A Multi-Layer Perceptron (MLP) classifier, trained on the RAVDESS dataset, achieves an accuracy of over 85%, ensuring reliable results. Key features include drag-and-drop audio upload, secure authentication, real-time prediction, confidence scoring, and history tracking. The system prioritizes privacy by processing data locally rather than relying on cloud services. With applications in mental health monitoring, customer service, and education, the project demonstrates how artificial intelligence can enhance human–computer interaction and pave the way for advanced emotion-aware technologies.