A Novel Approach for Pulse Signals Extraction and Heart Rate Estimation With Hilbert Space Orthogonalization Using Event-Based Camera

This paper presents a novel signal processing pipeline for estimating pulse signals, onset peaks, and heart rate using an event-based camera (EVS). EVS capture changes in brightness as a series of events, including both spatially induced events triggered by motion artifacts and temporally induced events caused by changes in skin luminance. When reconstructing pulse signals, spatially induced ON- and OFF-events within the region of interest (ROI) often occur simultaneously, canceling each other out and compromising signal stability. Additionally, biases between ON- and OFF-events can cause peak inversions under certain conditions, further destabilizing the signal. The proposed method addresses these challenges and temporal characteristics of EVS by defining event streams in a Hilbert space and applying Gram-Schmidt orthogonalization to isolate the temporally induced event stream. This enables robust estimation of pulse signals, heart rate through frequency analysis, and onset peak detection. To our knowledge, this is the first study that applies orthogonalization in Hilbert space to improve the signal processing of event-based pulse waves. The performance of the proposed approach is evaluated by comparing the extracted parameters with ground truth photoplethysmography (PPG) signals under three different lighting conditions. In the setup combining a green LED and a green optical filter, the method achieves a mean absolute error (MAE) of 1.13 BPM, a root mean square error (RMSE) of 1.69 BPM, and a peak detection accuracy of 92.72%. These results demonstrate the effectiveness of the proposed event-based method and highlight the potential of EVS for non-contact heart rate monitoring.