Acoustic gait recognition on a staircase

Our work involves monitoring and learning methods for inhabitant identification in smart home environments. Inexpensive audio monitoring equipment is used to gather sets of samples of footstep patterns on an interior staircase. Samples are then automatically processed to identify some basic features of the resulting wave pattern. Labeled examples are then utilized in training a Neural Network for purposes of identification of different individuals. Our results show significant ability to pick out different occupants of a real-world home environment based on the patterns of their gaits; indeed the use of the staircase for testing provides a natural way of limiting data variability and successfully classifying multiple individuals.