EchoSense - (Audio Visual Framework for Active Speaker Isolation)

This paper presents EchoSense, a Google Chrome extension designed to enhance speech intelligibility in noisy environments through the fusion of audio and visual cues. It leverages computer vision (face-api.js) for speaker identification using lip movement detection and applies RNNoise, a recurrent neural network-based model, for real-time noise suppression. The system integrates the Web Audio API and WebRTC for in-browser audio-video processing, enabling speaker-focused filtering without the need for external hardware or cloud services. Preliminary evaluations show that EchoSense effectively suppresses background noise while maintaining the clarity of the target speaker’s voice. The proposed system is particularly useful for virtual meetings, interviews, and remote work applications.