Extraction of Heartbeat Feature Based on Ballistocardiogram Signal From Multichannel Piezoelectric Ceramic Sensors

The fast pace of life has made the incidence rate and mortality rate caused by cardiovascular diseases increase. It is of great significance to detect and treat cardiovascular problems as early as possible. Due to inconvenience and uncomfortable reasons, wearable electrocardiogram (ECG) monitoring devices are unsuitable to be applied in daily healthcare, especially during sleep at night. It is necessary to provide a noncontact heart health monitoring method for those at risk of heart disease. In this article, we propose a multiinstance learning (MIL)-based algorithm to extract cardiac characteristics from ballistocardiogram (BCG) signals collected by piezoelectric ceramic sensors. Time and frequency domain heart rate variability (HRV) characteristics are obtained and compared with that extracted from ECG signals. The results show that the proposed method has the advantages of high detection accuracy compared with ECG method. Therefore, noncontact characteristics make BCG monitoring convenient to be used in daily healthcare for real-time alarm of heart disease, so as to achieve the aim of in-time risk detection and early treatment.