A Low-power Dry Electrode-based ECG Signal Acquisition with De-noising and Feature Extraction
Deepa Kota,Nishat Tasneem,Karthik Kakaraparty,Ifana Mahbub,Gayatri Mehta,Kamesh Namuduri
DOI: https://doi.org/10.1007/s11265-021-01681-z
2021-07-29
Journal of Signal Processing Systems
Abstract:Electrocardiogram (ECG) is widely used by physicians to detect heart ailments. The multi-resolution analysis offered by the Discrete Wavelet Transform (DWT) coupled with its noise-removal capability makes it a suitable tool for ECG signal analysis. In this paper, a DWT-based approach is presented for ECG signal decomposition and noise removal. The algorithm is tested on two ECG signals acquired using dry electrodes and a low-power on-chip amplifier designed in a standard 130-nm CMOS process. It is also tested on four ECG signals acquired from the MIT-BIH database; these measure the subject both at rest and in motion. For the ECG signal acquired using the dry electrode, the DWT-based denoising algorithm shows a 53.1% higher SNR compared to the Savitzky Golay de-noising algorithm. It outperforms the same algorithm and is effective in motion-artifact removal for ECG signals while walking, jogging, and exercising by 33%, 25%, and 16.1%. The proposed DWT-based feature extraction algorithm augmented with a minimum R-R interval threshold check feature enhances the performance and improves the feature extraction by 36.5%. The peak values of P, Q, R, S, and T, as well as the average RR, QRS, PP, and TT intervals of the ECG signals, are extracted and compared with the typical values for healthy adults. In case of any abnormalities in the detected features, the subject is recommended to consult the physician for further diagnosis. Comparison between the DWT and the Pan-Tompkins algorithm yields F1 score metric to be equal to one for both methodologies.