Improved Real-time EMG Decomposition via Signal Enhancement Using Fitzhugh-Nagumo Model

Motor unit (MU) activity obtained via real-time electromyogram (EMG) decomposition has a promising prospect in different applications. However, the non-stationary of EMG recordings can lead to a decreased decomposition accuracy under the real-time condition compared with the offline condition. In the independent component analysis (ICA)-based algorithms, the degraded performance was mainly caused by the decreased distance between the peaks at the time of firings and the baseline activity in the source signal, i.e. independent component. Therefore, a method that utilized the Fitzhugh-Nagumo (FHN) model to enhance the peaks in the source signal was developed in order to improve the real-time decomposition performance. The proposed method was compared with the conventional methods using synthetic EMG signals of which the MU spike trains were known a priori. The results showed that the proposed method can increase the decomposition accuracy significantly via enhancing the peaks in the source signal, especially when the EMG signals had a low signal-to-noise ratio. Our preliminary testing demonstrated the feasibility and effectiveness of the FHN model to enhance the peaks at the time of firings in order to improve the performance of the ICA-based EMG decomposition algorithms under the real-time condition. Further development of the proposed method can possibly promote the application of the real-time EMG decomposition in a variety of scenarios.