19-0059 Algorithm to provide real-time decomposition from electromyography signals

Adaptive real-time decomposition algorithm for detecting muscle activation.New motor units are identified and improved the accuracy of decomposition is achieved with the algorithm.
  • Adaptive real-time decomposition algorithm for detecting muscle activation.
  • New motor units are identified and improved the accuracy of decomposition is achieved with the algorithm.

Abstract

Motor units (MU) are a basis of muscle activation and transmit firing activities into contractions of muscle fibers. MU can provide insight into the conditions around neuromuscular control. MU activity can be captured using HD-EMG and extracted using source separation algorithms. These algorithms are computationally intensive, and the signals, therefore, need to be processed offline. Real-time approaches have been developed but have limitations in the number of MU that can be detected and the quality of the signal. An improved real-time decomposition algorithm is needed for use in research and clinical settings.

Researchers in the Department of Biomedical Engineering has developed an adaptive real-time decomposition algorithm for detecting muscle activation. This is a parallel-double-thread computation algorithm, with the backend performing separation matrix updates, while the frontend performs the real-time decomposition. The separation matrix updates alleviate the performance degradation of the decomposition. Studies with simulated signals were used to evaluate the performance of the real-time algorithm. This demonstrated that the real-time algorithm identified new motor units (3-4 fold over 30 minutes) and improved the accuracy of decomposition up to 10% in real-time, compared to approaches without real-time decomposition.

Website

https://unc.flintbox.com/technologies/6CDFCCFA0D17470D8AD15F381C489381

Advantages

  • Adaptive real-time MU decomposition
  • Enhanced identification of new MU
  • Improved accuracy of decomposition compared to non-real-time approaches

Potential Applications

Research and clinical EMG studies. Myoelectric control of rehabilitation or assistive devices.

Contact Information

Name: Matthew Howe

Email: matthew.howe@unc.edu

Phone: 919.966.3929