ECG reconstruction of body sensor network using compressed sensing based on overcomplete dictionary

Regarding to tackle the problem of demanding high accuracy reconstruction electrocardiogram (ECG) signal in a remote monitoring center of the body sensor network (BSN) and the low power problem of the body sensor network, this paper proposes a method of ECG reconstruction of body sensor network using compressed sensing based on overcomplete dictionary. The proposed method uses the compressed sensing theory and random binary matrices as the sensing matrix to measure the ECG signal on the sensor nodes. After the measured value is transmitted to the remote monitoring center, the overcomplete dictionary based on K-SVD algorithm training and the block sparse Bayesian learning reconstruction algorithm are used to reconstruct the ECG signal. Simulation results show that the SNR of the compressed sensing reconstruction ECG based on K-SVD overcomplete dictionary method is 5~22 dB higher than that of the method using discrete cosine transform when the ECG signal compression rate is at 70%~95%.The method has the advantages of high accuracy of signal reconstruction, low power, and easy hardware implementation. Copyright © 2014 Acta Automatica Sinica. All rights reserved.