Face Recognition In Harsh Conditions: An Acoustic Based Approach

This project develops an innovative acoustic-based facial recognition system that is designed to maintain robust performance even in situations where facial masks obstruct visibility. The system operates through two key components: a signal processing module that translates raw acoustic signals into explicit 3D facial representations, and a deep learning model for recognition and discrimination, ensuring high accuracy even when facial masks are present. 

The uploaded codebase accompanying this paper provides a full-stack implementation of the proposed system, comprising the signal processing code for extracting facial spectrums and the implementation of the deep learning model for spectrum recognition. Additionally, this artifact includes a self-constructed dataset, enabling evaluation of system performance across various usage scenarios. 

Please access and download the dataset with this link: https://mega.nz/folder/hwA23I7Y