Heart Rate Signal Separation Algorithm Based On Human Face Video

Heart rate(HR)is the number of beats per minute,and the human HR detecting has become an important measure of heart health.The existing style of human HR detecting is divided into contact and non-contact,which the contact HR detecting accuracy is better.However,the contact human HR detecting needs to be in contact with the skin and not suitable for special situations such as large area burns due to the use of sophisticate.In recent years,non-contact human HR estimation based on face-videos has attracted widespread attentions and reasearchs.In real application,the extracted PPG(Photoplethysmogram)signals from the face-videos usually include different types of noised signals.How to separate the HR signals from the mixed PPG signal is the focus and difficulty of HR detection researchs.To address this problem,this paper proposed a HR signals separation algorithm based on morphological component analysis(MCA)and sparse reconstruction,including three parts: dictionary learning,signal separation and sparse reconstruction.Specifically,the dictionary of controllable environments and the dictionary of background noise containing HR signals are separately learnt to form a combined dictionary.Then the mixed signals are sparely decomposed in the combined dictionary.Finally,the HR separation signal is obtained through the sparse coefficient reconstruction with the corresponding dictionary.Experiments are implemented on the actual data collected and the results demonstrate the effectiveness and robustness of the proposed method.