Evaluation of Denoising Algorithms for Footsteps Sound Classification in Noisy Environments

Identifying a person using footsteps sounds is part of the recent research in developing biometrics, systems designed to identify an individual in a group using body measurements. The sound of footsteps has a short history in this field, and present particular challenges. One of the most important is the background noise, given that any microphone installed on the floor with the purpose of recording footstep sounds will eventually record background noise and many other sounds as well. In this paper, we evaluate the combination of several denoising and classification algorithms for a person’s identification under several noisy conditions so as to establish a baseline in the field of distant sound recognition of footsteps. The results show the convenience of applying the denoising algorithms only in cases where the signal is affected by the high-noise level, which indicates the convenience of using real-time adaptive filters or more robust algorithms for both denoising and classification.