Novel near E-Field Topography Sensor for Human-Machine Interfacing in Robotic Applications

Area of Science:

• Robotics
• Biomedical Engineering
• Sensor Technology

Background:

• Human-machine interfaces (HMIs) require intuitive control methods.
• Sensing muscle activity is crucial for advanced robotic applications.
• Existing methods for muscle sensing can be invasive or computationally intensive.

Purpose of the Study:

• To introduce and validate a novel non-contact near E-field sensing technology for robotic HMIs.
• To assess the sensor's performance in detecting subtle changes in limb topography due to muscle actuation.
• To demonstrate the sensor's potential for real-time intention detection and joint angle prediction.

Main Methods:

• Utilized near E-field sensing to measure limb surface topography changes.
• Evaluated sensor characteristics: accuracy, hysteresis, and resolution.
• Analyzed sensor output against hand and finger movements for intention detection.
• Employed a convolutional neural network for joint angle prediction across nine degrees of freedom.

Main Results:

• The near E-field sensors demonstrated non-contact, low-noise, and low-computational-cost muscle activity sensing.
• Sensor validation confirmed reliable performance in accuracy, hysteresis, and resolution.
• Raw sensor data showed high relevance for muscle activation detection.
• Achieved root-mean-square error (RMSE) below 6 degrees for thumb/wrist and 11 degrees for finger joint angle prediction.

Conclusions:

• The novel near E-field sensing technology shows significant promise for robotic HMI applications.
• The sensor's ability to provide targeted muscle activation data facilitates effective intention detection.
• This approach offers a viable, high-performance solution for advanced human-robot interaction systems.