Artificial Neural Network to Detect Human Hand Gestures for a Robotic Arm Control

Area of Science:

• Biomedical Engineering
• Robotics
• Neuroscience

Background:

• Assistive technology is crucial for individuals with muscular impairments, such as Cerebral Palsy and Duchenne Muscular Dystrophy, to improve their activities of daily living (ADL).
• Existing robotic arm interfaces often lack intuitiveness and accuracy, especially for higher degrees of freedom (DOF).

Purpose of the Study:

• To develop an intuitive control system for a 6 DOF robotic arm using artificial neural network (ANN) classification of surface electromyography (sEMG) signals.
• To enable control of the robotic arm through nine distinct hand gestures detected in real-time.

Main Methods:

• Real-time surface electromyography (sEMG) signals from the user's forearm were captured.
• An artificial neural network (ANN) model, specifically a scaled conjugate gradient backpropagation network, was trained to classify nine hand gestures.
• The trained ANN model was integrated into a control system for a 6 DOF robotic arm within a Simulink environment.

Main Results:

• The ANN model achieved a classification accuracy of 85% for detecting hand gestures from sEMG signals.
• Comparison with other machine learning classifiers showed ensemble-bagged trees (90.3%) and cubic SVM (89.6%) had higher accuracies, but the ANN was chosen for usability.
• Preliminary testing in a virtual environment indicated that the robotic arm's forward kinematic control system performed well for most hand poses.

Conclusions:

• The proposed ANN-based sEMG control system offers a viable and intuitive interface for robotic manipulators.
• This approach has the potential to significantly enhance the independence and quality of life for individuals with neuromuscular disorders.
• Future work will focus on advanced sEMG signal processing and developing a generalized ANN model trained on data from multiple subjects.