Continuous noise estimation using time-frequency ECG representation

Common use of telemedical recordings performed in home care conditions and interpreted automatically justifies the need for a reliable signal-to-noise estimate. We propose new noise measurement technique based on a time-frequency model of noise computed in a quasi-continuous way. The proposed method is dedicated to ECG and uses automatically recognized cardiac components for temporal adaptation of the local bandwidth estimate. This noise is captured in each particular scale as non-uniformly sampled series and next is interpolated to the regions where components of cardiac representation are expected. Our approach yields a quasi-continuous model of the noise with a maximum value of measured to calculated data points ratio. The measurement of the noise level may be specified as temporal function being local ratio of energies from signal and from noise TF zones. The dynamic response of the model to rapid noise changes and thus the temporal precision of the SNR estimation are limited only by the resolution of TF representation. The accuracy of noise estimation for noise model-based and baseline-based methods are similar (0.64dB and 0.69dB respectively) as long as the noise level is stable. However in case of dynamic noise, the proposed algorithm outperforms the baseline-based method (0.95dB and 2.90dB respectively).

[1]  K. Egiazarian,et al.  Suppression of electromyogram interference on the electrocardiogram by transform domain denoising , 2001, Medical and Biological Engineering and Computing.

[2]  M. Ramasubba Reddy,et al.  A transform domain SVD filter for suppression of muscle noise artefacts in exercise ECG's , 2000, IEEE Transactions on Biomedical Engineering.

[3]  Arthur J. Moss,et al.  Noninvasive Electrocardiology: Clinical Aspects of Holter Monitoring , 1996 .

[4]  H. Feichtinger,et al.  Exact iterative reconstruction algorithm for multivariate irregularly sampled functions in spline-like spaces: The $L^p$-theory , 1998 .

[5]  José M. F. Moura,et al.  Biomedical Signal Processing , 2018, Series in BioEngineering.

[6]  P. Augustyniak.  Moving window signal concatenation for spectral analysis of ECG waves , 2010, 2010 Computing in Cardiology.

[7]  R. M. Rangayyan,et al.  Automatic de-noising of knee-joint vibration signals using adaptive time-frequency representations , 2006, Medical and Biological Engineering and Computing.

[8]  W. Abelmann,et al.  The Harvard-MIT Division of Health Sciences and Technology , 2004 .

[9]  P. Augustyniak.  Controlling the Distortions Distribution in a Wavelet Packet-Based ECG Compression , 2001 .

[10]  Atanas P. Gotchev,et al.  Denoising of ECG signals using wavelet shrinkage with time-frequency dependant treshold , 1998, 9th European Signal Processing Conference (EUSIPCO 1998).

[11]  Piotr Augustyniak,et al.  How a human perceives the electrocardiogram - the pursuit of information distribution through scanpath analysis , 2003, Computers in Cardiology, 2003.

[12]  J H van Bemmel,et al.  A reference data base for multilead electrocardiographic computer measurement programs. , 1987, Journal of the American College of Cardiology.