October 10, 2025Open Access

TinyML Implementation of CNN-Based Gait Analysis for Low-Cost Motorized Prosthetics: A Proof-of-Concept

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Proof-of-concept demonstrates gait analysis using CNNs in motorized prosthetics, highlighting efficiency and accuracy.

Achieved an average RMSE of 2.05° in knee joint angle prediction, surpassing a generalist baseline by over 35%.
Specialized CNN architectures tailored for diverse stride types were developed, optimizing model performance and efficiency.
Real-time gait data were collected from four healthy participants using inertial measurement unit signals, ensuring varied data for training models.
Lightweight classification allowed rapid selection between models, emphasizing the use of low-cost embedded platforms for practical applications.

Yamashita et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark