Classifying Gait Alterations Using an Instrumented Smart Sock and Deep Learning

This article presents a noninvasive method of classifying gait patterns associated with various movement disorders and/or neurological conditions, utilizing unobtrusive, instrumented socks and deep-learning network. Seamless were fabricated using three accelerometer-embedded yarns, positioned at the toe (hallux), above heel, on lateral malleolus. Human trials conducted 12 able-bodied participants, an sock was worn each foot. Participants asked to complete seven consisting their typical six different types that mimicked conditions. Four neural networks SVM tested ascertain most effective automatic data classification. The bi-long short-term memory (LSTM) generated accurate results illustrates use accelerometers per foot increased classification accuracy compared single accelerometer by 11.4%. When only utilized for classification, ankle in comparison other two. network able correctly classify five types: stomp (100%), shuffle (66.8%), diplegic (66.6%), hemiplegic “normal walking” (58.0%). incapable differentiating slap (21.2%) steppage (4.8%). work demonstrates textile incorporating yarns capable generating sufficient allow distinguish between specific patterns. may enable clinicians therapists remotely alterations observe changes during rehabilitation.