Indoor Multi-Person Detection and Recognition through Footsteps: A Deep Learning Approach with Acoustic Signal Analysis

In the era of intelligent environments, detecting and recognising multi-person is crucial across various applications like smart homes, security surveillance, and human-computer interaction. This paper introduces a novel methodology integrating acoustic signal processing with deep learning to identify and detect multi-persons based on indoor footsteps. Anchored in a simulated family setting of three individuals, the research establishes a comprehensive audio repository, demonstrating the efficacy of the proposed approach. Presently, the study focuses on family recognition and activity detection, which can be applied to prevent unauthorised intrusions and facilitate the issuance of intelligent commands based on identified activities, thereby enhancing the convenience of daily life. Furthermore, the paper outlines prospective avenues for refining and expanding this pioneering methodology, opening new vistas for future research endeavours in this domain.