Spatiotemporal Analysis by Deep Learning of Gait Signatures From Floor Sensors

The recognition of gait pattern variation is high importance to various industrial and commercial applications, including security, sport, virtual reality, gaming, robotics, medical rehabilitation, mental illness diagnosis, space exploration, others. purpose this paper study the nature variability in more detail, by identifying intervals responsible for variations individuals, as well between using cognitive demanding tasks. This work uses deep learning methods sensor fusion 116 plastic optical fiber (POF) distributed sensors recognition. floor system captures spatiotemporal samples due varying ground reaction force (GRF) multiples up 4 uninterrupted steps on a continuous 2×1 m area. We demonstrate classifications signatures, achieving 100% F1-score with Convolutional Neural Networks (CNN), context 21 subjects, imposters clients. Classifications under load, induced different dual tasks, manifested lower F1-scores. Layer-Wise Relevance Propagation (LRP) are employed decompose trained neural network prediction relevant standard events cycle, generating “heat map” over input used classification. allows valuable insight into which parts signal have heaviest influence classification consequently, events, such heel strike or toe-off, mostly affected load.