Stockwell Transform Detector for Photoplethysmography Signal Segmentation

Abstract

Real-time embedded analysis of physiologic waveforms is critical to predict impending pathophysiology. While electrocardiogram (ECG) data is often analyzed to assess cardiovascular disease, there is recent evidence that photoplethysmography (PPG) can track blood loss and thereby alert to hypovolemia. In this work we present a Stockwell transform inspired filter bank to segment PPG waveforms. The Stockwell transform allows for computationally efficient frequency analysis. The proposed Stockwell filter bank utilizes a sparse time-frequency spectrum and is coupled to the Shannon energy envelope to extract PPG peaks. Finally, we demonstrate that the described method is tolerant to the presence of additive Gaussian noise.