| PhD Seminar

Name of the Speaker: Mr. Shubhanshu Sharma (EE20D413)
Guide: Prof. Boby George
Online meeting link: https://meet.google.com/uun-bako-erd
Date/Time: 20th March 2024 (Wednesday), 2:00 PM
Title: Pathway Classification Through Integrated Shoe Sensor Systems

Abstract

Automated pathway classification helps to enhance safe mobility of individuals such as the visually impaired, lower leg amputees, and diabetic patients facing mono-filament issues. Additionally, it facilitates guided navigation for robots and autonomous vehicles. Many existing pathway classification schemes rely on properties such as color or image texture to categorize pathways. However, this reliance on limited information often leads to restricted classification accuracy. In addition, it fails to sense the pathway condition, such as wet or dry. In this talk, a sensing mechanism and the associated system that provides data to help significantly improve the pathway classification accuracy will be presented. It is achieved by using the electrical impedance information of the floor or pathway. Obtaining the current data on the floor impedance using a wearable device with sufficient accuracy is challenging. Modified shoes with flexible electrodes are used to achieve this. Using those, the impedance of the floor between the user's shoes is measured. This talk will present the details of the proposed approach and a thorough scientific study of the scheme. In addition, a detailed practical evaluation of the sensor prototype on various pedestrian pathways like soil, mud, road, grass, etc., and its application in aiding the classification will be presented. An analytical model is derived using conformal mapping to compute the electrical properties of the pathway between the shoes. The model's accuracy is verified using finite element analysis (FEA) and actual measurements. The existing image-based classification techniques use deep learning algorithms. The new information, i.e., the impedance of the floor, is provided as an additional input and integrated into this algorithm. The results proved the practical use of the proposed scheme. The details will be presented in the talk.