Human pulse wave acquisition method based on face video

Abstract

The invention discloses a human body pulse wave acquisition method based on human face video. This method first uses a camera to record a video of a face, then performs motion tracking on the face to obtain face motion information, and performs skin detection on the face area, and extracts skin chroma information from the relevant area based on the tracked face to construct a multi-region original image. Pulse wave signal; then according to the design mask of face motion information, filter out the distorted part of the initial pulse wave; then according to the designed signal dictionary, use the method of joint reconstruction to select a group of atoms similar to the original pulse wave from the dictionary, Add these atoms according to the weight to get the fitted pulse wave signal, and finally reduce the dimensionality of the fitted pulse wave matrix to get the pulse wave waveform. The invention can effectively filter out the noise caused by human face movement factors such as head twisting and expression changes and external light changes, and realizes reliable photocapacitive human body pulse wave signal acquisition in the actual environment.

Description

technical field [0001] The invention relates to a non-contact pulse wave acquisition method for a human body, in particular to a photoplethysmography acquisition method based on human face video images, which is applied to the fields of human physiological state detection and physical health state monitoring. Background technique [0002] By monitoring the pulse wave signal of the human body, the working conditions of the cardiovascular system of the human body can be understood. In the medical field, the potential risk of cardiovascular disease can be found through the pulse wave signal, which can provide a reference for treatment; in the field of health monitoring, the pulse wave signal can also be used to reflect the health status of the detected person. When the pulse wave signal is abnormal, prompt Cardiovascular health risks for users. [0003] When the light beam irradiates the skin surface, the blood in the skin absorbs and attenuates the light beam, and the attenua...

AI Technical Summary

Problems solved by technology

The authorized announcement number is CN103702014B, and the invention patent titled "non-contact physiological parameter detection method, system and device" is a blind source separation method, which cannot effectively filter out the noise introduced by human face movement and illumination

The authorized announcement number is CN105266787A, and the invention patent titled "a non-contact heart rate detection method and system" is a color difference algorithm, which cannot eliminate the different noises in each color channel

The authorized announcement number is CN110191675A, and the invention patent titled "System and Method for Non-Contact Determination of Blood Pressure" uses a pre-trained machine learning model of hemoglobin concentration change, which is inconvenient to obtain data sets and mark data for actual use. train

"Pulse wave detection based on mobile phone video in real scenes" takes into account the influence of uneven illumination in various areas of the face, and uses small area signals for weighted average processing, but there are noises caused by local movement of the face in the small area signals. As a result, motion noise in some areas is introduced into the final weighted signal. It does not consider that the source of each small area signal is caused by the beating of the human heart. It only uses the characteristics of the small area signal itself to simulate the small area signal. Combined reconstruction fails to take advantage of the correlation between signals in each small area, and cannot effectively remove local noise during signal reconstruction

Images