Removing ECG Noise from Surface EMG Based on Information Theory
An adaptive filtering method is presented which eliminates ECG artifact from EMG signals based on error entropy criterion. In this method, the error distribution is estimated and minimized in an adaptive manner. Mean squared error (MSE) criterion only minimizes 2nd order statistics of the error, so it is sufficient in cases where inherent noise is Gaussian. The error entropy (EE) criterion, used in the proposed algorithm, minimizes all moments of error distribution. So in EMG denoising, where ECG artifact is typically non-Gaussian, Minimum Error Entropy (MEE)-based adaptive algorithm will improve noise elimination performance. Simulation results show that proposed algorithm has better spectral coherence in low frequencies and improves the SNR of the denoised EMG signal (about 5dB), especially in low SNR inputs, compared to MSE based algorithms.
|Adaptive filtering, Entropy, LMS algorithm, MEE criterion, Noise cancellation|
|26th Iranian Conference on Electrical Engineering, ICEE 2018|
|Organisation||Department of Systems and Computer Engineering|
Darroudi, A. (Ali), Parchami, J. (Jaber), Sarbishaei, G. (Ghazaleh), & Rajan, S. (2018). Removing ECG Noise from Surface EMG Based on Information Theory. In 26th Iranian Conference on Electrical Engineering, ICEE 2018 (pp. 1403–1408). doi:10.1109/ICEE.2018.8472613