Research On The Extraction Of Non-contact Heart Rate Variation Feature Under Different Illumination

Heart Rate Variability(HRV)is widely used in the fields of medical health and emotion recognition.At present,most of the HRV features are extracted by contact,and the subjects have more inconvenience during the collection process.However,when the HRV features are extracted in a non-contact manner through the face video based on the IPPG principle,it is easily affected by environmental factors such as illumination.This article focuses on the extraction of HRV features under different illuminances.Because different illuminances have different effects on HRV feature extraction,this article first observes the HRV features extracted under three illuminance environments of low,medium and high through experiments.The results show that the lower the illuminance,the greater the error of the extracted HRV features.In order to reduce the error of extracting HRV features based on face video under different illuminances,this paper focuses on signal denoising processing and enhancement processing.The heart rate curve waveforms extracted under different illuminances are different,and a single noise reduction method cannot meet the requirements of adaptability and less distortion at the same time.For this reason,this paper proposes an improved noise reduction method of Ensemble Empirical Mode Decomposition(EEMD)combined with wavelet threshold.First,it is adaptively decomposed into multiple components by EEMD,and then the components outside the pulse wave frequency range are processed by wavelet threshold noise reduction.It solves the self-adaptation problem of a single wavelet threshold,and calculates the average absolute error of HRV features such as Amean,ALF,AHF,ALF/HF,etc.The improved noise reduction is reduced by 2.73%,1.97%,2.86%,3.33% under low illumination compared with single wavelet threshold processing;the error was reduced by 5.78%,6.22%,6.18%,and 5.72% under daily illumination;the error was reduced by 1.43%,5.65%,4.96%and5.60% under strong illumination,respectively.At the same time,it solves the problem of more distortion of a single EEMD.The experimental results show that the improved noise reduction is1.12%,2.06%,1.49%,and 3.11% lower than the single EEMD noise reduction process under low illumination;under daily illumination the errors were reduced by 4.39%,6.73%,5.34%,and 5.73%;the errors were reduced by 1.24%,6.0%,5.75%,and 5.82%under strong illumination.Because the strength of the heart rate signal extracted in a low illumination environment is weak.For this reason,this paper proposes a processing scheme for signal enhancement using the Maximum Ratio Combining(MRC)algorithm.Divide the entire face into multiple local areas,calculate the signal-to-noise ratio of each area as a weight,re-weight and combine the regional signals to obtain the heart rate curve,and finally calculate the HRV feature.A comparative experiment under three illuminances shows that: after using the MRC algorithm,the HRV feature error in the time domain and frequency domain of the HRV feature under each illuminance is reduced,especially in a low-illumination environment,the MRC algorithm is used to enhance the extraction after processing The average absolute error of HRV features was reduced by 5.68%,5.95%,5.13%,6.82%,respectively.