Study On Heart Rate Detection Method Based On Subtle Head Motions

With the number of vehicles per capita increased year by year,the traffic accident rate is getting higher and higher,in which the proportion of accidents caused by the fatigue driving is growing.When the human body is in a state of fatigue,a series of physiological parameters will change,including the heart rate signal.Studies have shown that when a human was in a stationary state,his head appears subtle motions caused by the cardiac contraction,and those subtle motions contain heartbeat signal,so there is an important significance for further study of drivers' fatigue state by detecting heart rate according to the process on head subtle motions.This paper proposed an improved method named Pre-decision Pyramid Layers of Eulerian Video Magnification for Subtle Movements,using this improved methed to amplify the subtle movements of head in the input video,then processing the amplied movements to detective the heart rate signal.The main contents of the paper is as following:First,amplify subtle motions.The advantages and disadvantages of three methods including Motion Magnification,Eulerian Video Magnification and Phase-Based Video Motion Processing algorithm has been studied for amplify subtle motions,finally,Eulerian Video Magnification method has been chosen to magnify subtle motions and color changes for the input video.Second,Proposed the Pre-decision Pyramid Layers of Eulerian Video Magnification Algorithm For Subtle Motions.The video quality after small motions being magnified was affected by the amplitude of subtle motions and spatial decomposition pyramid layers.The quality of magnified video was better at the high number pyramid layers when the amplitude of small motion is large,on the contrary,a better magnified video was achieved at the low number of pyramid layers.Detect the movement foreground before doing Eulerian Video Magnification processing,find the video frame which have the minimum proportion of non-zero pixels to the total pixels from the foreground video,this is the frame which have the biggest amplitude of movements,then determine the suitable pyramid layers according to the settled threshold,then do the Eulerian Video Magnification processing.The results shown that this method effectively solve the question of amplified result is not ideal due to the signal pyramid layer on space decomposition.Finally,Extracted heart rate signal based on those small motions on head.Detect face on the video with amplified small movements,removed the region of eyes as to the interference on the experiment,then extract and track feature points in ROI,analyze the trajectory on vertical direction,selecte the frequency component that best matches the heartbeat signal,finally get the heart rate signal.