Samueli School of Engineering

One in four people will experience a stroke in their lifetime, and technical advancements have decreased stroke deaths worldwide, correlating to an increase in stroke survivors. About 80% of survivors experience upper extremity impairment, and half of stroke survivors will experience muscle spasticity, paresis, and/or contractures after recovery, hindering activities of daily living (ADLs). To address this growing concern and particular needs in care standardization, psychological support, and effectiveness, we sought to leverage machine learning to provide personalized care. Our approach is two-fold: movement characterization and targeted stimulation. We aim to characterize different patient movements using a random forest-based machine learning model and electromyography signals. Then, using the characterization and the relative strength of the EMG signal, we will direct electrical stimulation to the arm with individualized locations and intensities. Our project focuses on two muscle groups: flexor carpi radialis & ulnaris and extensor carpi radialis & digitorum. Our main achievement thus far is creating our characterization machine learning model with a 93% accuracy. We have also developed an initial functional electrical stimulation (FES) prototype, showing our capability to customize and safely house the FES within the greater solution pathway. With further optimization and integration of our prototype, we will be able to demonstrate the impact of our solution in providing individualized support and care for common ADLs.