Feasibility of Accelerometer-Based Arterial Pulse Measurement: Part I—Theoretical Analysis

Despite their commercial availability and portability, accelerometers (ACCs) are much less frequently used for arterial pulse waveform (APW) measurement, and the resulting acceleration pulse waveform (APWa) has been much less studied than the APW measured by a tactile sensor. This paper, the first of two parts, presents a theoretical analysis of the feasibility of ACC-based arterial pulse measurement at rest, focusing on the measurement-related technical issues underlying the differences from tactile-based measurement. The tissue–contact–sensor (TCS) stack between the artery and ACC is modeled as a single-degree-of-freedom (SDOF) system to collectively capture the dynamic behavior of all the factors involved during pulse measurement, where motion artifact (MA) is quantified as low-frequency (<0.7 Hz) mechanical motion of the ACC and alters tissue prestress, resulting in time-varying system parameters (TVSPs) in the SDOF system. Theoretical relations are established among the arterial pulse signal, MA, TCS system parameters, and measured acceleration. The analysis reveals that, even in the absence of MA, the measured acceleration differs fundamentally from the displacement-domain APW because of the dynamic response of the TCS stack and the nonstationary nature of the arterial pulse signal. In the presence of MA, the measured acceleration is affected through two mechanisms: MA itself, represented as baseline drift (BD), and TVSP-induced distortion. While MA itself can be removed using conventional filtering, TVSP-induced distortion is embedded within individual arterial pulse harmonics. Numerical simulations illustrate these effects by comparing displacement- and acceleration-domain signals with and without MA and further demonstrate the distinct effects of TVSP-induced distortion on the instantaneous parameters of a harmonic. These findings provide a theoretical basis for the experimental investigation and evaluation in Part II and a mechanistic basis for MA and sensor-noise removal in ACC-measured pulse signals.