Finger motion classification by forearm skin surface vibration signals.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Human-Computer Interaction

Background:

• Prosthetic hand systems aim to restore function for amputees, often relying on myoelectric potentials from surface electromyogram (EMG) signals.
• Surface EMG signals have limitations in recognizing fine, independent finger motions crucial for dexterous hand activities.

Purpose of the Study:

• To investigate the feasibility of using motion-related surface vibrations for detecting independent finger motions.
• To develop an alternative interface for prosthetic hands that bypasses the limitations of EMG signal recognition.

Main Methods:

• Accelerometers were utilized to capture mechanical vibration patterns during a finger tapping experiment.
• Feature extraction was performed using norm-based, correlation coefficient-based, and power spectrum-based methods.
• Back-propagation neural networks were employed to classify distinct finger motions based on extracted vibration features.

Main Results:

• The study successfully demonstrated that distinct finger motions generate recognizable vibration patterns.
• Neural network classification based on vibration features proved effective in identifying individual finger movements.
• This approach offers a viable alternative to EMG-based control for prosthetic hands.

Conclusions:

• Independent finger motion identification is achievable through the analysis of surface vibration patterns.
• This vibration-based method holds promise for improving the dexterity and functionality of prosthetic hands.
• Further research can explore integrating this technique into advanced prosthetic control systems.