METHOD FOR ESTIMATING PHYSIOLOGICAL EVENTS FROM PHYSIOLOGICAL SIGNALS, A NON-TRANSITORY COMPUTER-READABLE MEDIUM, AND, AN APPARATUS

Accurate peak detection in physiological signals is fundamental for several tasks related to health monitoring. A method for fine-tuning candidate peak positions and detecting peaks of interest in signals is provided. The fine-tuning method addresses the problem of low signal resolution and reduces the error with respect to the gold-standard reference signal usually collected at higher sampling frequencies. Obtaining accurate peak positions without modifying the sampling frequency is essential in the context of wearable devices, which often present limited computational resources and storage. Furthermore, the method enables selection of the peaks of interest by classifying their tuned positions according to a set of features extracted from morphological characteristics of the signal. The present pipeline is illustrated through inter-beat interval (IBI) estimation from wrist-PPG signals collected from smartwatches. The method may also be suited to the refinement and detection of different fiducial points, including peaks and valleys of interest.