Selection of suitable hand gestures for reliable myoelectric human computer interface.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Human-Computer Interaction

Background:

• Myoelectric prosthetic hands rely on identifying hand gestures using surface electromyogram (sEMG) signals from forearm muscles.
• Accurate automated recognition necessitates selecting a subset of hand gestures for improved performance.

Purpose of the Study:

• To develop and validate a method for selecting a subset of hand gestures to maximize sensitivity and specificity in myoelectric hand gesture recognition.
• To improve the reliability of myoelectric-based human-computer interface systems.

Main Methods:

• sEMG signals were recorded from forearm muscles during various hand gestures.
• The Positive-Negative Performance Measurement Index (PNM), derived from confusion matrices, was used to rank gesture performance.
• A subset of gestures was selected based on PNM ranking for optimized classification.

Main Results:

• Using all ten gestures yielded 80.0% sensitivity and 97.8% specificity.
• Selecting six key gestures (Hand open, Hand close, Little finger flexion, Ring finger flexion, Middle finger flexion, Thumb flexion) using the PNM method resulted in >95% sensitivity and specificity (96.5% and 99.3%).

Conclusions:

• Careful selection of hand gestures is crucial for reliable myoelectric-based human-computer interface systems.
• Gesture selection significantly enhances the accuracy and reliability of automated hand gesture recognition in prosthetic applications.