Abstract:Monitoring the subject's physical activity intensity (PAI) is key to preventing sports injury and enhancing conditioning performance. Existing monitoring equipment can be cumbersome and is not always accessible in daily situations due to cost or usability. For ease of usage, we propose a motion robust remote photoplethysmography (rPPG) measurement during exercise for contactless PAI detection. To account for vigorous motion during physical exercise, we leverage signal holding with motion analysis to maintain the consistency of the signal source. We then obtain robust rPPG signals via a convolutional neural network (CNN)-based rPPG constructor module composed of a dual feature extractor with signal-based targets, a base span module, and a noise elimination module. In terms of the PAI detection application, as changes in heart rate are a dominant biomarker during exercise, this is applied to estimate the training level for PAI detection. In this study, training status is classified into three levels and is estimated via a residual-learning-based architecture with sequential heart rate input. In terms of heart rate measurements, the proposed method outperforms the state-of-the-art approaches in motion scenarios on both ECG-Fitness and our own exercise datasets. The mean absolute error (MAE) is reduced to about one-third of the comparison methods in real-world treadmill running scenarios, and the error level goes down by about 30% in the ECG-Fitness dataset under a cross-dataset evaluation. In addition, for follow-up PAI detection applications, the heart rate values estimated with the proposed method approximate the heart rate ground truth. It is capable to represent the trend between training load and PAI changes. We propose a novel framework for the motion robust rPPG measurement during exercise applied to PAI detection, with which overall athlete assessment can be accomplished as a convenient alternative to existing athlete moni- oring equipment.

Deep Recurrent Neural Network for Extracting Pulse Rate Variability from Photoplethysmography During Strenuous Physical Exercise

A Feasibility Study of a Video-Based Heart Rate Estimation Method with Convolutional Neural Networks

Heart Rate Estimation From Facial Videos Using a Spatiotemporal Representation With Convolutional Neural Networks

A novel one-stage framework for visual pulse rate estimation using deep neural networks

A LSTM-Based Realtime Signal Quality Assessment for Photoplethysmogram and Remote Photoplethysmogram

Convolutional neural network with spatio-temporal-channel attention for remote heart rate estimation

Efficient Real-Time Camera Based Estimation of Heart Rate and Its Variability

PulseGAN: Learning to Generate Realistic Pulse Waveforms in Remote Photoplethysmography

PPGnet: Deep Network for Device Independent Heart Rate Estimation from Photoplethysmogram

Real-time Webcam Heart-Rate and Variability Estimation with Clean Ground Truth for Evaluation

Recognising Cardiac Abnormalities in Wearable Device Photoplethysmography (PPG) with Deep Learning

Stochastic Modeling Based Nonlinear Bayesian Filtering for Photoplethysmography Denoising in Wearable Devices

Efficient Spatiotemporal Attention Network for Remote Heart Rate Variability Analysis

Motion Robust Remote Photoplethysmography Measurement During Exercise for Contactless Physical Activity Intensity Detection

Classification of Photoplethysmographic Signal Quality with Deep Convolution Neural Networks for Accurate Measurement of Cardiac Stroke Volume

Remote Photoplethysmograph Signal Measurement from Facial Videos Using Spatio-Temporal Networks

i-PRExT: Photoplethysmography Derived Respiration Signal Extraction and Respiratory Rate Tracking Using Neural Networks

PulseNet: A multitask learning network for remote heart rate estimation

Beat-to-beat continuous blood pressure estimation with optimal feature set of PPG and ECG signals using deep recurrent neural networks

Robust modelling of arterial blood pressure reconstruction from photoplethysmography

Prediction of arterial blood pressure waveforms from photoplethysmogram signals via fully convolutional neural networks