Recent radar data generation studies for human activity recognition (HAR) always focus on statistic-driven methods guided by statistical parameters, such as mean, variance, or latent network features. However, they lack physical interpretability and are unable to generate targeted data with a specified position or direction. To address this issue, a novel motion-driven radar data generation method is proposed for Any-position Any-Direction (APAD) HAR using monostatic radar. First, a 5D point cloud, including time, range, Doppler, pitch, and azimuth angle features, is achieved as a human activity model sample from a radial recording by Fast Fourier Transform (FFT), Constant False Alarm Rate Detector (CFAR), and Direction of Arrival (DOA) angle estimation processing. In particular, a speed-aware Density-based Spatial Clustering of Application with Noise (VA-DBSCAN) algorithm is designed for point reduction under the consideration of non-uniform point cloud distributions. Then, new human activity samples from any starting position with any moving direction can be generated by translating and rotating the above model sample. The generated Doppler and energy features are corrected according to the viewing position and direction. Finally, a radar-based PointNet Long Short-Term Memory (PointNet-LSTM) network is adopted for the APAD-HAR. In contrast experiments, only 48 recorded samples were utilized as a data source to generate a training group with 27888 samples. Our proposed method demonstrated its feasibility and superiority with an average recognition accuracy rate of 92.11% by more than 25% improvement compared to other state-of-the-art methods.

Obstructive sleep apnea (OSA) is associated with psychophysiological impairments, and recent studies have shown the feasibility of using speech and craniofacial images during wakefulness for severity estimation. However, the inherent limitations of unimodal data constrain the performance of current methods. To address this, we proposed a novel hypergraph-based multimodal fusion framework (HMFusion) that integrates psychophysiological information from audio-visual data. Specifically, we employ long short-term memory (LSTM)-based encoders to extract modality-specific temporal dynamics from pre-trained audio-visual embeddings and remotely photoplethysmography (rPPG)-derived heart rate sequences. A hypergraph neural network is then utilized to capture critical cross-modal interactions for OSA severity estimation. Evaluation on a dataset of 159 participants from a clinical sleep center demonstrates that the proposed model achieves area under the receiver operating characteristic curves (AUCs) of 88.26%, 86.07%, and 85.29%, with corresponding F1-scores of 92.91%, 85.50%, and 85.30% at Apnea-Hypopnea Index (AHI) thresholds of 5, 15, and 30 events/hour, respectively, outperforming state-of-the-art approaches. This study highlights the potential of psychophysiological data in enhancing OSA severity estimation during wakefulness, offering new avenues for clinical research in this field.

The high-frequency hybrid sky-surface wave radar (HSSWR) utilizes the ionosphere–ocean hybrid propagation channel and can detect targets over the horizon while preserving high detection precision. Owing to the uneven and ever-changing electron density distribution in the ionosphere, the phase path of electromagnetic waves propagating through it fluctuates irregularly, resulting in the widening, even splitting of the spectrum, and ultimately affecting the capacity of radar for detecting slow-moving targets. In this article, a novel parametric estimation and compensation method that combines the cubic phase function (CPF) and the high-order ambiguity function (HAF), referred to as the hybrid CPF-HAF method, is adopted to solve the problem of phase path contamination of the HSSWR signals. The direct wave is selected as the calibration signal and modeled by piecewise polynomial phase signals (PPSs). The maximum likelihood (ML) principle is adopted for selecting the PPSs’ orders, and the hybrid CPF-HAF method is employed to estimate the parameters of PPSs and reconstruct the correction signal. It is ultimately employed to compensate for the ionosphere-induced phase path contamination. Simulation results indicate that the CPF-HAF outperforms the HAF in terms of PPSs’ coefficients estimating accuracy. The processing results of measured data further demonstrate the efficiency of the proposed decontamination algorithm. After phase compensation, the dilated spectra of the sea clutter were sharpened, and the peak amplitudes significantly increased.

Sleep monitoring plays a vital role in various scenarios such as hospitals and living-assisted homes, contributing to the prevention of sleep accidents as well as the assessment of sleep health. Contactless camera-based sleep monitoring is promising due to its user-friendly nature and rich visual semantics. However, the privacy concern of video cameras limits their applications in sleep monitoring. In this paper, we explored the opportunity of using a defocused camera that does not allow identification of the monitored subject when measuring sleep-related parameters, as face detection and recognition are impossible on optically blurred images. We proposed a novel privacy-protected sleep parameters measurement framework, including a physiological measurement branch and a semantic analysis branch based on ResNet-18. Four important sleep parameters are measured: heart rate (HR), respiration rate (RR), sleep posture, and movement. The results of HR, RR, and movement have strong correlations with the reference (HR: R = 0.9076; RR: R = 0.9734; Movement: R = 0.9946). The overall mean absolute errors (MAE) for HR and RR are 5.2 bpm and 1.5 bpm respectively. The measurement of HR and RR achieve reliable estimation coverage of 72.1% and 93.6%, respectively. The sleep posture detection achieves an overall accuracy of 94.5%. Experimental results show that the defocused camera is promising for sleep monitoring as it fundamentally eliminates the privacy issue while still allowing the measurement of multiple parameters that are essential for sleep health informatics.

Radar-based solutions have attracted great attention in human activity recognition (HAR) for their advantages in accuracy, robustness, and privacy protection. The conventional approaches transform radar signals into feature maps and then directly process them as visual images. While effective, these image-based methods may not be the best solutions in terms of representation efficiency to encode the relevant information for classification. This article proposes a novel HAR method combining sparse theory and PointNet network, with both operations in the time-Doppler (TD) and range-Doppler (RD) domains. First, sparsity-based feature extraction is introduced to use a limited number of sparse solutions to characterize human activities in the form of TD sparse point clouds (TDSP) or dynamic RD sparse point clouds (DRDSP). This new representation is validated by comparing the reconstructed and original signals. Then, PointNet networks are adopted to summarize multidomain features and predict human activity labels by a sparse set of input point clouds. Comprehensive experiments were conducted to demonstrate that the proposed method can yield a higher representation efficiency, classification accuracy, and better generalization capability than existing ones.

Radar-based fall detection technology has attracted much attention for its high accuracy, robustness, and privacy preservation potential. Detection of “Soft Fall”, i.e., high-freedom fall, is the key to practical application. This paper proposes a novel height-tracking method based on a multiple-input multiple-output (MIMO) radar system to address this problem. First, the received signal was segmented into a time sequence with a sliding window along slow time. Next, Fast Fourier Transform (FFT) and Multiple Signal Classification (MUSIC) algorithms were applied to estimate the general trend of the human body’s time-varying range and pitch angle information. Then, they were fused with a geometrical relationship to describe height changes during fall motions using a height trajectory map. Two height-based features were extracted as input to Support Vector Machine (SVM) to distinguish soft fall and fall-similar motions. Finally, experiments, including four soft fall and five typical fall-similar motions, were conducted to demonstrate its feasibility and superiority.

Moving-target indication (MTI) has been widely researched in urban sensing. The existing research work mainly concentrates on a single moving target, and there exist difficulties to research multiple moving targets, such as estimating the number of targets and distinguishing the targets that are close to each other. In this letter, based on compressive sensing (CS), we consider the indication of multiple moving targets in through-the-wall radar imaging (TWRI). Using the method of change detection, stationary targets and clutter are removed. Then, according to the reasonable state of motion, we delete the ghost points and combine the tracing points of each moving target. Comparison with the existing CS-based MTI method shows that with the proposed method, ghosts can be suppressed for about 60% at most, which can better estimate the number of moving targets and show the indication of each moving target.

Sleep staging is to score the sleep state of a subject into different sleep stages such as Wake and Rapid Eye Movement (REM). It plays an indispensable role in the diagnosis and treatment of sleep disorders. As manual sleep staging through well-trained sleep experts is time consuming, tedious, and subjective, many automatic methods have been developed for accurate, efficient, and objective sleep staging. Recently, deep learning based methods have been successfully proposed for electroencephalogram (EEG) based sleep staging with promising results. However, most of these methods directly take EEG raw signals as input of convolutional neural networks (CNNs) without considering the domain knowledge of EEG staging. Apart from that, to capture temporal information, most of the existing methods utilize recurrent neural networks such as LSTM (Long Short Term Memory) which are not effective for modelling global temporal context and difficult to train. Therefore, inspired by the clinical guidelines of sleep staging such as AASM (American Academy of Sleep Medicine) rules where different stages are generally characterized by EEG waveforms of various frequencies, we propose a multi-scale deep architecture by decomposing an EEG signal into different frequency bands as input to CNNs. To model global temporal context, we utilize the multi-head self-attention module of the transformer model to not only improve performance, but also shorten the training time. In addition, we choose residual based architecture which makes training end-to-end. Experimental results on two widely used sleep staging datasets, Montreal Archive of Sleep Studies (MASS) and sleep-EDF datasets, demonstrate the effectiveness and significant efficiency (up to 12 times less training time) of our proposed method over the state-of-the-art.

Respiratory sound is able to differentiate sleep stages and provide a non-contact and cost-effective solution for the diagnosis and treatment monitoring of sleep-related diseases. While most of the existing respiratory sound-based methods focus on a limited number of sleep stages such as sleep/wake and wake/rapid eye movement (REM)/non-REM, it is essential to detect sleep stages at a finer level for sleep quality evaluation. In this paper, we for the first time study a sleep stage detection method aiming at classifying sleep states into four sleep stages: wake, REM, light sleep, and deep sleep from the respiratory sound. In addition to extracting time-domain features, frequency-domain features of respiratory sound, non-linear features of snoring sound are devised to better characterize snoring-related signals of respiratory sound. To effectively fuse the three sets of features, a novel feature fusion technique combining the generalized canonical correlation analysis with the ReliefF algorithm is proposed for discriminative feature selection. Final stage detection is achieved with popular classifiers including decision tree, support vector machines, K-nearest neighbor, and the ensemble classifier. To evaluate our proposed method, we built an in-house dataset, which is comprised of 13 nights of sleep audio data from a sleep laboratory. Experimental results indicate that our proposed method outperforms the existing related ones and is promising for large-scale non-contact sleep monitoring.

Radar-based human motion recognition is crucial for many applications, such as surveillance, search and rescue operations, smart homes, and assisted living. Continuous human motion recognition in real-living environment is necessary for practical deployment, i.e., classification of a sequence of activities transitioning one into another, rather than individual activities. In this paper, a novel dynamic range-Doppler trajectory (DRDT) method based on the frequency-modulated continuous-wave (FMCW) radar system is proposed to recognize continuous human motions with various conditions emulating real-living environment. This method can separate continuous motions and process them as single events. First, range-Doppler frames consisting of a series of range-Doppler maps are obtained from the backscattered signals. Next, the DRDT is extracted from these frames to monitor human motions in time, range, and Doppler domains in real time. Then, a peak search method is applied to locate and separate each human motion from the DRDT map. Finally, range, Doppler, radar cross section (RCS), and dispersion features are extracted and combined in a multidomain fusion approach as inputs to a machine learning classifier. This achieves accurate and robust recognition even in various conditions of distance, view angle, direction, and individual diversity. Extensive experiments have been conducted to show its feasibility and superiority by obtaining an average accuracy of 91.9% on continuous classification.

In this paper, a noncontact breathing disorder recognition system has been proposed for identifying irregular breathing patterns. The proposed system consists of a Doppler radar-based sensor module and a machine-learning-based breathing disorder recognition module. A custom-designed 2.4-GHz continuous wave digital-IF Doppler radar is utilized as the radar sensor module to accurately capture the time-domain breathing waveform. Then, a recognition module is designed with selected features and optimized classifiers. Four sets of experiments have been carried out to evaluate the proposed system comprehensively. For the laboratorial experiments, the proposed system achieves 94.7% classification accuracy using the linear support vector machine classifier with seven selected features. Results of clinical experiments demonstrate the feasibility of long-term breathing disorder recognition with good accuracy and robustness, and illustrate the potential of the proposed solution for the auxiliary diagnosis of diseases.

Noncontact techniques for detecting vital signs have attracted great interest due to the benefits shown in medical monitoring and military applications. A rapid remote evaluation on physiological signal frequencies is needed in search and rescue operations as well as intensive care. However, the presence of respiration harmonics causes aliasing problems to heart-rate estimation, especially when the data volume is limited. By taking advantage of the simple pattern of physiological signals, we propose a stepwise atomic norm minimization method (StANM) to accurately assess the respiration and heartbeat frequencies with a limited data volume. First, the respiration frequency is estimated by the conventional atomic norm minimization. Then the frequencies of respiration harmonics are generated based on the inherent relationship between the fundamental tone and the harmonics. Finally, with the pre-estimated frequencies, we locate the heartbeat frequency by solving a modified atomic norm minimization problem. Simulations and experiments show that the proposed method can accurately estimate physiological frequencies from 6.5-second-long raw data with a 4-Hz sampling rate.

A method based on ensemble empirical mode decomposition (EEMD) is proposed for accurately detecting the time varying pitch of speech in tonal languages. Unlike frame-, event-, or subspace-based pitch detectors, the time varying information of pitch within the short duration, which is of crucial importance in speech processing of tonal languages, can be accurately extracted. The Chinese Linguistic Data Consortium (CLDC) database for Mandarin Chinese was employed as standard speech data for the evaluation of the effectiveness of the method. It is shown that the proposed method provides more accurate and reliable results, particularly in estimating the tones of non-monotonically varying pitches like the third one in Mandarin Chinese. Also, it is shown that the new method has strong resistance to noise disturbance.