Signal detection accuracy of digital accelerometers for ballistocardiographic propose

The Ballistocardiography is a method to gain detailed information about cardiac function parameters using heart's transferred movements on the body surface. In our current work, we try to achieve such a precision and sensitivity with digital accelerometers. We created a test set-up under lab conditions in which we used a piezo-vibration generator (PVG) with four times four different high-sensitive triaxial accelerometers on four boards. We used 1 Hz and 2 Hz for physiological-like signals, as well as 10 Hz for comparison proposes. Furthermore, we recorded neutral signals for noise information. All signals were measured by sensor's z-axis. Altogether we got 128 physiological-like-signal data sets. For the analysis, we used the auto-correlation via sliding-window. With this method, we determined the mean periodic intra-signal-correlation, which is used as a base value. Furthermore, we used the cross-correlation (CC) for inter-signal-corr. between similar/diff. sensors, synthetic double-differentiated voltage-like-signals and diff. signal-morphologies (incl. neutral). To conclude, such low-frequency-signals, based on intra-signal periodicity, can be distinguished properly by CC. The intra-signal-correlation is on average. 0.73, values up to 0.98 are reached. In contrast, inter-signal comparisons have correlation of <0.04 max (neutral) and 0.22 between comparable signals. Due to the low accel. triangle signals show low correlation, but are clearly distinguishable from noise.