Spatio-Temporal Video Processing for Respiratory Rate Estimation

In this paper, we present a wire-free, low-cost video processing-based technique for respiratory rate (RR) estimation. The proposed method blends together two recently presented techniques, with the purpose of emphasizing small movements, such as respiratory movements possibly present in a video stream, in order to detect them. Initially, the system performs a spatial decomposition of the video frames in a pyramidal representation, in which each layer contains different spatial details. The levels are then pixel-wise temporally filtered with an infinite impulse response (IIR) filter, purposely designed to extract components having a periodicity compatible with the respiratory rate. Afterwards a single motion signal is extracted from each level. Finally, the extracted signals are jointly analyzed according to the maximum likelihood (ML) criterion in order to estimate the respiratory rate. The parameters extracted by our algorithm show a good agreement with those indicated by a gold-standard polysomnographic system. Therefore, our results, although preliminary, are encouraging and show that the respiratory rate can be reliably measured and monitored by a low-cost, wire-free, video processing-based system.