Automatic Identification of Signal Quality for Heart Beat Detection in Cardiac MEMS Signals

Rapidly growing number of heart monitoring systems based on microelectromechanical sensor (MEMS) devices demands highly accurate processing algorithms for indicating heartbeat location. In clinical applications based on the analysis of cardiovascular mechanical motion, the signal may vary due to a variety of reasons including motion artifacts, location of the sensor and posture of the person being tested. Also reasons related to the internal device characteristics affect to the signal formation. The purpose of this paper is to address the problem of selecting the best axis when using 3-axis MEMS accelerometer and gyroscope sensors. The formation of noise and artifacts in the observed data depends, for example, on the posture of the monitored person. In this paper, the application for the axis selection is heartbeat detection, and we show that by selecting the optimal axis benefits can be obtained in terms of detection accuracy. With 10-fold cross-validation and with KSVM classifier we obtained the sensitivity and specificity of 83.9% and 86.1%, respectively, in using seismocardiogram waveform for the sepatation between good quality and low quality data. With using gyrocardiogram waveforms we obtained the sensitivity and specificity of 95.9% and 80.0%, respectively.