Human Activity Recognition via Garment-embedded EMG Sensors

Loading...
Thumbnail Image

Date

2024-05

Journal Title

Journal ISSN

Volume Title

Publisher

The Ohio State University

Research Projects

Organizational Units

Journal Issue

Abstract

Human activity recognition (HAR) is a fast-growing field due to the increased desire to include this technology in many commercial electronic devices and its applicability to healthcare. The most common approach to HAR is to train a machine learning model with sensor data from human activity. Currently, accelerometers, inertial sensors, and gyroscopic sensors are the most common devices used to collect data for HAR. Another type of sensor that can provide data on human activity is an electromyography (EMG) sensor. EMG measures the electric signals due to person’s muscles activation. This thesis explores the possibility of using EMG sensors for HAR. To investigate this possibility, garments with embedded EMG sensors from ExoForce. The garments, when worn by participants, are designed to measure the activation of triceps, biceps, and calf muscles. The study consisted of participants who were asked to do a series of defined exercises (pushups, jumping jacks, curls, etc.) while wearing the EMG garments. A subset of the data from the participants was used to train a multinomial logistic regression machine learning model for classification. The rest of the data from the participants was used to test the accuracy of the model. The accuracy of the model was 45.31% on the test data (much higher than random guessing, which would be 12.5%), 98.44% on the training data, and 71.88% on the whole data set. The results from this study show that EMG sensors can be used as the basis for a device that accurately identifies various human activities, while future work could focus on improving accuracy and exploring specific applications. This finding could lead the way to improved commercial or medical devices aimed at HAR.

Description

Keywords

Human Activity Recognition, EMG Sensors, Machine Learning, Wearable Devices, Feature Extraction

Citation