#Comparing machine learning and deep learning for locomotion recognition

Able-bodied walkers can easily adjust to different terrains during daily activities, but amputees face challenges due to the loss of sensory feedback from their foot-ankle. Researchers have developed active foot prosthetics with gait phase detection methods to improve walking accuracy. Various algorithms, such as capacitive sensing signals and machine learning models, have been used for gait phase estimation. IMU sensors are commonly used for gait recognition due to their efficiency and stability. Deep learning models, such as RNN and CNN, have shown promising results in recognizing locomotion modes. Different classifiers, such as decision trees and SVM, have also been utilized for locomotion mode detection. The performance outcomes of these systems vary based on the sensors used and the algorithms applied. The proposed system aims to classify and detect four locomotion modes using deep learning and machine learning models. The study compares the performance of different models and highlights the importance of accurate locomotion mode recognition for prosthetic designs. The research provides insights into the development of efficient and cost-effective systems for locomotion assistance for amputees transitioning between different walking modes.

Source link

Source link: https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2022.923164/full