Sensor Fusion for Analysis of Gait under Cognitive Load: Deep Learning Approach

摘要: Human mobility requires substantial cognitive resources, thus elevated complexity in the navigated environment instigates gait deterioration due to naturally limited load capacity. This work uses deep learning methods for 116 sensors fusion study effects of on human healthy subjects. We demonstrate classifications, achieving 86% precision with Convolutional Neural Networks (CNN), normal as well 15 subjects’ under two types demanding tasks. Floor capturing multiples up 4 uninterrupted steps were utilized harvest raw spatiotemporal signals, based ground reaction force (GRF). A Layer-Wise Relevance Propagation (LRP) technique is proposed interpret CNN prediction terms relevance standard events cycle. LRP projects model predictions back input signal, generate a "heat map" over original training set, or an unknown sample classified by model. allows valuable insight into which parts signal have heaviest influence classification and consequently, gate are mostly affected load.