Multimodal sensor fusion of cardiac signals via blind deconvolution: A source-filter approach

Sensor fusion is a growing field within the medical signal processing community. Traditionally, it is performed implicitly by the physician when diagnosing the state of a patient from various measurement modalities such as electrocardiography (ECG), arterial blood pressure (ABP) or photoplethysmography (PPG). These may represent different physical quantities like voltage, pressure or scattering properties and are modulated by various physiological conditions and artifacts. Still, they originate from a single source, the heart. In signal processing, this is known as a single-input-multiple-output (SIMO) system and several approaches to estimate the source are known. In this paper, a blind deconvolution approach is chosen and adapted for physiological signals. The feasibility and robustness is shown using simulated data. Moreover, the approach is validated on real data recorded in a polysomnography setting, fusing PPG and Ballistocardiography (BCG).