Seminar 1: Footstep recognition using vibration signals collected by Distributed Acoustic Sensing system (DAS)

The recent COVID-19 pandemic has caught people’s attention on using engineering control methods to separate people and the pathogen to reduce the risks of getting infected, especially in the indoor environment. Studying occupants’ footfall indicated by their trajectories in buildings can support detecting crowd gathering hazard. The first and the most important step of studying occupants’ trajectories is to observe their footsteps. Although the acquisition of occupants’ footsteps is straightforward to conceive in principle, it is difficult to be implemented in real cases considering privacy, cost and social acceptability. Hence, there is a need to propose an efficient method to observe occupants’ footsteps for the downstream research. This study focuses on the recognition of footsteps from the vibration signals collected through Distributed Acoustic Sensing system (DAS) that can be easily deployed especially for large scale scenarios. Data collected from a preliminary experiment done in NUS academic building including both single occupant’s movement and multiple occupants’ movement is analyzed. A machine learning method is proposed to recognize footsteps based on the characteristics of the data. An average accuracy of 73% has been achieved which is much better than the traditional signal processing methods. The findings of the study have proved the effectiveness and efficiency of using DAS in footstep recognition and provided opportunity for further gait profile distinguishment as well as trajectory extraction.