Comparison of Ballistocardiogram Processing Methods Based on Fiber Specklegram Sensors

Abstract

The ballistocardiogram (BCG) is a graphic representation of the movements of the body associated with cardiac activity. In this article, a 10-min BCG has been captured for ten different volunteers with a polymer optical fiber (POF) specklegram sensor. This transducer, which is composed of a charge-coupled device (CCD) camera, a laser emitting diode, and two meters of POF, allows capturing the BCG by analyzing how the induced speckle pattern changes over time. These changes are related to cardiac activity. Several processing methods have been compared to determine which method achieves the best performance: complex cepstrum, power of spectral density (PSD), Pam–Tompkins algorithm, wavelet, autocorrelation, Savitzky–Golay filter, mean absolute deviation, and Hilbert transform. Accuracy and resource consumption have been characterized and compared for these methods. Hilbert, PSD, and Savitzky–Golay exhibit both small errors and computational costs. This article describes a baseline for the main frequency determination of POF-based BCG signals in real-time.