Prediction of Grasp Forces Using Multi-Channel EMG Signals

Biomedical Engineering

Yugesh Prasanna Balasubramian

Abstract

This project focused on predicting grasp force (FG) and manipulation force (FM) from high-density EMG signals
recorded during dexterous object manipulation tasks. A condition-wise CNN-LSTM framework was used to model the relationship between multi-channel EMG input and force output, supported by preprocessing steps such as bad-channel correction, interpolation, and down-sampling. The results showed that the model captured the overall force trends well, while also highlighting condition-dependent differences and opportunities for future improvement through CNN interpretability and expanded subject-level analysis

Video

Research poster

Faculty mentor

Portrait of Marco Santello

Marco Santello

Professor

School of Biological and Health Systems Engineering

[email protected]