Design Of A Video-based Heart Rate Detection System In A Smart Home

In recent years,artificial intelligence physiological parameter collection technology has developed rapidly.The use of camera to perform non-sensing heart rate detection can determine the health of the human body and detect sudden cardiovascular diseases in time.However,this technology requires stable acquisition of fixed areas of the face.Video images,real-time dynamic heart rate detection research has become one of the research hotspots in this field.The traditional contact heart rate measurement method has high measurement accuracy,but the test cost is high,the equipment is highly specialized,the test method is cumbersome,and it is not suitable for the daily heart rate detection of the general population at home.The video-based non-contact heart rate measurement method has unique advantages,but it is easily affected by sports and ambient light,and the accuracy is difficult to meet the requirements.According to these two factors,this article optimizes the video-based heart rate detection method and designs a method based on The indoor heart rate detection system tracking the key points of the face,combined with the image processing technology in deep learning,effectively crops the skin area,and uses the signal separation and amplification technology to purify the heart rate signal and calculate the heart rate.This paper mainly uses the MTCNN network for face recognition,uses the optimized GOTURN network to track the face,uses the optimized PFLD network to detect the key points of the face,and optimally crops the image signal of the face region of interest according to the position relationship of the face organs.Use the Euler color amplification method to amplify the frequency band of the image signal heart rate signal,and use the complementary set empirical mode decomposition to decompose the source signal into multiple groups of signals,use the blind source separation technology to purify the heart rate signal,and transfer the signal from the time domain to the frequency domain,Calculate the real-time heart rate.This design combines multiple technologies such as face recognition and tracking,key point detection,signal amplification,decomposition,and synthesis,and is applied to human heart rate detection.It is suitable for indoor daily scenes,effectively reducing costs and improving accuracy.System experiments show that,This design has met expectations and has a certain practicability.