Research Of Heart Rate Measurement Based On Multi-channel PPG Sensor Signals

As we all know,the number of heart beats per minute is defined as heart rate from the first sound.Among the vital signs to detection,heart rate is regarded as one objective evaluation for controlling training load,and it can also provide a reference for medical diagnosis.Therefore,heart rate has been widely applied to fitness exercises,competitive race,and sports training,to name a few.With the emerging of smart watches,smart wristbands and other wearable intelligent devices,people pay more attention to their own health.Heart rate measurement using photoplethysmography(PPG)has attracted more and more attention both in industry and academia.More importantly,PPG sensors are simple,low-cost,easy-to-use,and do not require reference sensors.However,PPG sensor signal is a feeble biological signal from tissue surface.Its signal intensity is weak,so it is sensitive and vulnerable to motion artifacts which caused by intense physical activities.And then the phenomenon strongly affects the reliability and accuracy of heart rate measurement.Hence,removal of motion artifacts in raw PPG sensor signals is a challenging task.On the basis of a large number of literature research and summarize,this paper proposes three algorithms of robust heart rate measurement based on multi-channel PPG sensor signals,which can satisfactorily solve the shortcomings of the current existing state-of-the-art algorithms.(1)Considering the sparsity of the PPG sensor signals and the strong relativity between the PPG sensor signals and the acceleration signals,we propose an algorithm of heart rate measurement based on multi-channel spectral matrix decomposition.The algorithm first constructs a spectral matrix using multi-channel PPG sensor signals and acceleration signals during the same time period.Next,the spectral matrix is divided into a motion artifacts matrix and real PPG sensor signals matrix.Then we combine the compressive sensing theory to model the motion artifacts removing process in PPG sensor signals as a multi-channel spectral matrix decomposition model.And we adopt Accelerated Proximal Gradient(APG)method to optimize the objective function of the model.Finally,we combine a novel spectral peak tracking method to estimate the heart rate in real PPG spectra and achieve the precise heart rate measurement.(2)To the best of our knowledge,most existing algorithms adopt heuristic methods to find the spectral peaks corresponding to heart rate,which involve many user-tuning parameters.Thus,these algorithms have low generalization ability and robustness.For the above problem,we propose an algorithm of heart rate measurement based on Support Vector Machine(SVM).The algorithm first generate the reference motion artifacts signal components by dealing with the acceleration signals by Principle Component Analysis(PCA)prior to adaptive filtering.Then a part of motion artifacts can be reduced by updating weight of the adaptive least mean square filter continually.Next,we use the compressive sensing theory and the rows sparse characteristics of the spectral matrix to model the further motion artifacts removing process in PPG sensor signals as a sparse signal reconstruction model.And we utilize the Regularized M-FOCUSS algorithm to estimate the solution of this model.Combined with spectrum subtraction,the denoised multi-channel PPG spectra can be obtained.Finally,we make use of SVM-based spectral peak tracking method to select the correct spectral peaks corresponding to heart rate.(3)To further improve the denoising performance and the accuracy of heart rate measurement,we also research an offline algorithm of heart rate measurement based on joint sparse spectrum reconstruction.The algorithm first uses multi-channel PPG sensor signals and simultaneous acceleration signals to construct a spectral matrix.The sparse characteristics of the spectral matrix and its rows are extracted which based on the sparsity of the PPG sensor signals and the strong relativity between the PPG sensor signals and the acceleration signals.Then we combine the compressive sensing theory to model the motion artifacts removing process in PPG sensor signals as a joint sparse spectrum reconstruction model.And we exploit Imprecise Augmented Lagrangian Method(IALM)to optimize the objective function of the model.Finally,we utilize the spectrum subtraction to elimination the motion artifacts in PPG sensor signals and output the precise estimated heart rate.The above proposed algorithms were verified on the well-known PPG datasets,and comparing with the current mainstream heart rate measurement algorithms.The results demonstrate that the algorithms proposed in this paper can effectively remove the motion artifacts and ensure the accuracy of heart rate estimates.