People counting is one of the hottest issues in sensing applications. Impulse radio ultrawideband radar has been extensively adopted to count people because it provides a device-free solution without illumination and privacy concerns. However, current solutions have limited performances in congested environments due to signal superpositions and obstructions. In this letter, a hybrid feature extraction method based on the curvelet transform and the distance bin is proposed. First, 2-D radar matrix features are extracted at multiple scales and multiple angles by applying the curvelet transform. Then, the distance bin concept is introduced by dividing each row of the matrix into several bins along the propagating distance to select features. A radar signal data set is constructed for three density scenarios, including people randomly walking in a constrained area at densities of three and four persons per square meter and people in a queue with an average between-person distance of 10 cm. The number of people in the data set scenarios varies from 0 to 20. Four classifiers--a decision tree, an AdaBoost classifier, a random forest, and a neural network--are compared to validate the hybrid features. The random forest achieves the highest accuracy of above 97% in the three density scenarios. To further investigate the reliability of the hybrid features, they are compared with three other features: cluster features, activity features, and features extracted by a convolutional neural network. The comparison results reveal that the proposed hybrid features are stable, and their performance is substantially more effective than that of the others.

Vital signs monitoring in physical activity (PA) is of great significance in daily healthcare. Impulse Radio Ultra-WideBand (IR-UWB) radar provides a contactless vital signs detection approach with advantages in range resolution and penetration. Several researches have verified the feasibility of IR-UWB radar monitoring when the target keeps still. However, various body movements are induced by PA, which lead to severe signal distortion and interfere vital signs extraction. To address this challenge, a novel joint chest–abdomen cardiopulmonary signal estimation approach is proposed to detect breath and heartbeat simultaneously using IR-UWB radars. The movements of target chest and abdomen are detected by two IR-UWB radars, respectively. Considering the signal overlapping of vital signs and body motion artifacts, Empirical Wavelet Transform (EWT) is applied on received radar signals to remove clutter and mitigate movement interference. Moreover, improved EWT with frequency segmentation refinement is applied on each radar to decompose vital signals of target chest and abdomen to vital sign-related sub-signals, respectively. After that, based on the thoracoabdominal movement correlation, cross-correlation functions are calculated among chest and abdomen sub-signals to estimate breath and heartbeat. The experiments are conducted under three kinds of PA situations and two general body movements, the results of which indicate the effectiveness and superiority of the proposed approach.

The monitoring of human activity and vital signs plays a significant role in remote health-care. Radar provides a non-contact monitoring approach without privacy and illumination concerns. However, multiple people in a narrow indoor environment bring dense multipaths for activity monitoring, and the received vital sign signals are heavily distorted with body movements. This paper proposes a framework based on Frequency Modulated Continuous Wave (FMCW) and Impulse Radio Ultra-Wideband (IR-UWB) radars to address these challenges, designing intelligent spatial-temporal information fusion for activity and vital sign monitoring. First, a local binary pattern (LBP) and energy features are extracted from FMCW radar, combined with the wavelet packet transform (WPT) features on IR-UWB radar for activity monitoring. Then the additional information guided fusing network (A-FuseNet) is proposed with a modified generative and adversarial structure for vital sign monitoring. A Cascaded Convolutional Neural Network (CCNN) module and a Long Short Term Memory (LSTM) module are designed as the fusion sub-network for vital sign information extraction and multisensory data fusion, while a discrimination sub-network is constructed to optimize the fused heartbeat signal. In addition, the activity and movement characteristics are introduced as additional information to guide the fusion and optimization. A multi-radar dataset with an FMCW and two IR-UWB radars in a cotton tent, a small room and a wide lobby is constructed, and the accuracies of activity and vital sign monitoring achieve 99.9% and 92.3% respectively. Experimental results demonstrate the superiority and robustness of the proposed framework.

Further applications of impulse radio ultra-wideband radar in mobile health are hindered by the difficulty in extracting such vital signals as heartbeats from moving targets. Although the empirical mode decomposition based method is applied in recovering waveforms of heartbeats and estimating heart rates, the instantaneous heart rate is not achievable. This paper proposes a Heartbeat Estimation And Recovery (HEAR) approach to expand the application to mobile scenarios and extract instantaneous heartbeats. Firstly, the HEAR approach acquires vital signals by mapping maximum echo amplitudes to the fast time delay and compensating large body movements. Secondly, HEAR adopts the variational nonlinear chirp mode decomposition in extracting instantaneous frequencies of heartbeats. Thirdly, HEAR extends the clutter removal method based on the wavelet decomposition with a two-parameter exponential threshold. Compared to heart rates simultaneously collected by electrocardiograms (ECG), HEAR achieves a minimum error rate 4.6% in moving state and 2.25% in resting state. The Bland–Altman analysis verifies the consistency of beat-to-beat intervals in ECG and extracted heartbeat signals with the mean deviation smaller than 0.1 s. It indicates that HEAR is practical in offering clinical diagnoses such as the heart rate variability analysis in mobile monitoring.

To enhance electrocardiogram (ECG) monitoring systems in personalized detections, deep neural networks (DNNs) are applied to overcome individual differences by periodical retraining. As introduced previously [4], DNNs relieve individual differences by fusing ECG with impulse radio ultra-wide band (IR-UWB) radar. However, such DNN-based ECG monitoring system tends to overfit into personal small datasets and is difficult to generalize to newly collected unlabeled data. This paper proposes a self-adjustable domain adaptation (SADA) strategy to prevent from overfitting and exploit unlabeled data. Firstly, this paper enlarges the database of ECG and radar data with actual records acquired from 28 testers and expanded by the data augmentation. Secondly, to utilize unlabeled data, SADA combines self organizing maps with the transfer learning in predicting labels. Thirdly, SADA integrates the one-class classification with domain adaptation algorithms to reduce overfitting. Based on our enlarged database and standard databases, a large dataset of 73200 records and a small one of 1849 records are built up to verify our proposal. Results show SADA's effectiveness in predicting labels and increments in the sensitivity of DNNs by 14.4% compared with existing domain adaptation algorithms.

In-vehicle people sensing has received considerable attention for preventing overloading and forgotten children. Impulse radio ultrawideband (IR-UWB) radar is widely applied in noncontact sensing applications. However, metal-rich environment and narrow space with multiple objects bring dense multipaths in received radar signals, posing a great challenge to in-vehicle sensing, especially people counting and motion recognition. This article proposes novel spatial-temporal-circulated gray level co-occurrence matrix (STC-GLCM) and physiological features for people sensing, to detect the presence, number, and motions of people in the vehicle. Considering temporal consistency and spatial continuity of highly associated in-vehicle radar signals, STC-GLCM feature is designed to describe the 2-D circular spatial-temporal distribution, and extracted at individual, associated, and global scales. In addition, the singular value ratio (SVR) is introduced to select the physiological feature at each individual scale, representing the possibility and intensity of cardiopulmonary activity. Three different classifiers are applied on these combined features for classification, and the stochastic gradient descent (SGD) is adopted for performance analysis. A radar signal dataset is constructed for various in-vehicle scenarios, including different number of people in a stationary and moving vehicle with several motions. The accuracies reach 97.5% and 97.0% in counting the number of people in the stationary and moving vehicle, and 99.0% in recognizing the motions, demonstrating the effectiveness and robustness of these proposed features. The proposed in-vehicle radar signal dataset is available at https://github.com/yangxiuzhu777/In-Vehicle-Radar-Data .