Wireless Walking Pattern Detection Using Force Sensitive Resistor Pressure Sensors on Footwear

• Published in: 13th IEEE Symposium on Computer Applications and Industrial Electronics, ISCAIE 2023
• Main Author: Khairuddin N.Z.; Halim I.S.A.; Azreen A.N.; Musa A.; Hassan S.L.M.; Tulos N.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2023
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85165194884&doi=10.1109%2fISCAIE57739.2023.10165473&partnerID=40&md5=001f50bc849346a767e9932c3dcf48aa

Previous studies on gait analysis have been conducted on various human motions, such as walking and running. These studies are mostly based on the detection of abnormalities in walking patterns and were conducted using non-wearable prototypes. This imposes measurement constraints and is unsuitable for measuring free-living subjects. Therefore, this work aims to develop a wireless walking pattern detection system using force-sensitive resistors as the sensor on footwear. The important parameter for recognizing the walking pattern, also called the gait cycle, senses pressure distribution on the feet, which is then translated into resistivity. The second aim of this work is to train and classify the walking pattern using a support vector machine (SVM) and k-nearest neighbors (KNN) classifier. As a result, the accuracy with and without principal components analysis (PCA) is compared. The system mainly detects the walking pattern through pressure sensors embedded in the footwear, and the system is controlled by an Arduino Uno. These sensors are placed on three main areas of the shoe's insole: the toe, metatarsal, and heel. All the data obtained is transmitted wirelessly through the wireless module receiver NR24L01F and sent to the computer for further analysis. Several walking patterns were collected, such as standing still, walking, and ascending and descending stairs. The collected data is plotted into a MATLAB graph, and the gait cycle is studied. The highest accuracy of both classifiers is achieved with 73.9% for weighted KNN classifiers and 69.2% for fine Gaussian SVM classifiers. © 2023 IEEE.