Hand motion strength forecasting using Extreme Learning Machine for post-stroke rehabilitation

Abstract

Stroke or Cerebrovascular accident (CVA) can cause weakness in one side of the body, including the upper limbs such as the hand. Rehabilitation is needed to restore the function of the hand. Rehabilitation should also measure the strength of the movements carried out. This article aims to forecast the strength of movement based on Electromyography (EMG) signals using the Extreme Learning Machine (ELM). This study collected EMG signal data and movement strength, carried out data pre-processing and data extraction using various extraction features, applied ELM for forecasting strength based on EMG signals, and applied created models in stroke therapy robots. The forecasting model is evaluated by measuring the Mean Squared Error (MSE). The average value of the best MSE in offline testing is 1.77, while the real-time testing is 0.79. A small MSE value indicates that the model is good enough. The resulted value of strength can be applied to make the stroke therapy robots actuating properly.