A signal quality analysis method for electrocardiosignals in the CS domain

Abstract

Compressed sensing (CS) is an important cutting-edge low-power technology for wearable health monitoring systems. To solve the problem of how the quality of compressive sampled electrocardiosignal (ECG) affects the subsequent arrhythmia recognition and signal reconstruction, a novel ECG quality evaluation method in the CS domain is proposed. In order to achieve the evaluation effect in the CS domain, a modified bandpass filter method is designed to obtain the approximate frequency band energy and approximate wavelet entropy, which are used as the evaluation index for the support vector machine classifier to realize the CS ECG signal quality evaluation. Based on the simulation results from the MIT-BIH ECG database, this method can reach an accuracy of more than 92% under different compression ratios (2, 4, 8 and 16). Simulation experiments based on data from the PhysioNet/CinC Challenge 2011 database show that the accuracy of this method can reach above 83% under different compression ratios. In evaluation experiments using actual ECG signals, we show that this method can achieve good accuracy when the compression ratio is less than eight. The accuracy of this method needs to be further improved when the compression ratio is 16. When the compression ratio is less than 8 and a sparse matrix is not used as the measurement matrix, the method proposed in this paper can directly evaluate the quality of ECG signals after compression sampling without reconstruction. Compared with the signal quality evaluation algorithm in the uncompressed domain, the experimental results of the proposed method are highly competitive and, under a certain compression ratio, the amount of transmitted data can be reduced while maintaning accuracy. Meanwhile, the proposed method also provides a new idea for signal quality evaluation in the compressed domain.