Imaging photoplethysmography (iPPG) is an emerging optical technique that allows for the contactless acquisition of arterial Blood Volume Pulse (BVP) signals from video recordings of the human skin. While iPPG offers a non-contact and convenient means for physiological monitoring, the accuracy of the extracted BVP signals remains limited. This limitation hinders its potential for advanced cardiovascular assessments, such as evaluations of arterial stiffness and cardiac function. To address this issue, we propose a novel physiologically informed Gaussian filtering method, based on the prior knowledge that the BVP waveform can be modeled as a mixture of multiple Gaussian components. Specifically, a set of physiological Gaussian kernels is employed to convolve the noisy iPPG signal, generating a Gaussian representation that emphasizes waveform components with physiological relevance. This representation is further refined by a Transformer-based neural network to reconstruct accurate BVP signals. Experimental results demonstrate a notable improvement in BVP accuracy, with the mean absolute error reducing from 0.25 to 0.08. This enhancement in iPPG precision highlights the potential of our approach for advanced medical applications.

The distributed deep learning architecture between front-deployed sensors and edge-deployed gateways attracts increasing interest. However, the inference performance of distributed deep models is also impacted by the delivery loss of intermediate representation in the wireless link, especially in the harsh industrial fading environments. Traditional communication systems usually focus on transmission errors at bit level, which treat all bits in the packets equally and fail to suit the varying importance in distributed deep models, which urges the essential evolution of the communication method to form a joint co-design paradigm for distributed deep models. This article then proposes to optimize the Mean Time To First Failure (MTTFF) of wireless link instead of traditional bit error rate, which enables a guaranteed transmission window. This paper first derives the analytical model of MTTFF under MIMO systems, then utilizes the kernel mixture distribution to obtain a closed-form solution of MTTFF, which forms a optimization algorithm minimizing the transmitted power while achieving the aiming MTTFF. Extensive reallife experiments show more than 70% satisfaction rate of MTTFF, which leads to more than 10 times higher inference accuracy than the original deep model.

In industrial environments, the wireless link of IoT systems often experiences complex channel fading effects, making accurate online estimation of link quality crucial for improving system performance. Using Gaussian Mixture Model (GMM) to fit I/Q symbols allows estimation of Rician channel parameters, but traditional GMMs typically rely on prior knowledge of the number of Gaussian components to ensure clustering accuracy, posing challenges for adaptive channel modulation schemes in industrial settings. This paper proposes an adaptive Gaussian mixture model based on Kullback-Leibler divergence (KLD), which autonomously determines the optimal number of clusters through iterative evaluation, achieving optimal clustering performance. Firstly, this study proposes the utilization of forward KLD as an optimization target, leveraging its known optimal prior of zero to avoid local optima. Secondly, the redundancy in the number of clusters is assessed using reverse KLD constructed with the single Gaussian distribution. These improvements ensure that the GMM converges correctly to the global optimum regardless of the initial cluster count settings.

Accurate online link quality metrics represented by the Rician parameter are critical to enhancing the reliability of industrial wireless networks subject to temporal fading channels. The Rician parameters can be estimated by fitting the received I/Q symbols with GMM (Gaussian Mixture Model). However, the classical Expectation-Maximization estimations of GMM rely on the preset hyper-parameter of kernel numbers to guarantee the convergence, making it hard to work under adaptive modulation schemes. To address this challenge, we first reveal that the derivative of likelihood is less capable of representing the global optimal, which leads to the well-known local optimal problem and the failure to recognize the false convergence caused by incorrectly configured kernel numbers. A new empirical metric derived from KLD (Kullback-Leibler divergence) has been proposed to identify the local optimal convergence, as well as a new metric tuple to discriminate redundant kernels. A novel estimation algorithm has then been designed to shift the number of kernels from the preset hyper-parameter to the adjustable parameter. This improvement guarantees the global optimal convergence of the GMM with any initial number of kernels. Extensive experiments demonstrate that the proposed method achieves over ten times better accuracy, while requires less than half the iterations.

The distributed deep learning architecture can support the front-deployment of deep learning systems in resource constrained IoT devices and is attracting increasing interest. However, most ready-to-use deep models are designed for centralized deployment without considering the transmission loss of the intermediate representation inside the distributed architecture. This oversight significantly affects the inference performance of distributed deployed deep models. To alleviate this problem, a state-of-the-art work chooses to retrain the original model to form an intermediate representation with ordered importance and yields better inference accuracy under constrained transmission bandwidth. This paper first reveals that this solution is essentially a pruning-like solution, where unimportant information is adaptively pruned to fit within the limited bandwidth. With this understanding, a novel scheme named Naturally Aggregated Intermediate Representation (NAIR) has been proposed, which aims to naturally amplify the difference of importance embedded in the intermediate representation from a mature deep model and reassemble the intermediate representation into a hierarchy of importance from high-to-low to accommodate the transmission loss. As a result, this method shows further improved performance in various scenarios, avoids compromising the overall inference performance of the system, and saves astronomical retraining and storage costs. The effectiveness of NAIR has been validated through extensive experiments, achieving a 112% improvement in performance compared to the state-of-the-art work.

Imaging Photoplethysmography (IPPG) is an emerging and efficient optical method for non-contact measurement of pulse waves using an image sensor. While the contactless way brings convenience, the inevitable distance between the sensor and the subject results in massive specular reflection interference on the skin surface, which leads to a low Signal to Interference plus Noise Ratio (SINR) of IPPG. To ease this challenge, this work proposes a novel modulation illumination approach to measure the accurate arterial pulse wave via surface reflection interference isolation from IPPG. Based on the proposed skin reflection model, a specific modulation illumination is designed to separate the surface reflections and obtain the subcutaneous diffuse reflections containing the pulse wave information. Compared with the results under ambient illumination and constant supplemental illumination, the SINR of the proposed method is improved by 4.56 and 3.74 dB, respectively.

Over-The-Air (OTA) measurement is considered the preferred method for measuring the antenna system and end-to-end performance of Multiple-Input-Multiple-Output (MIMO) devices under test. Spatial correlation has been widely utilized as a key metric for evaluating the accuracy of MIMO OTA measurements. However, there is no guarantee that the standard signal streams convoluted with specified impulse responses will be ideally independent of each other in the implementation of the MIMO OTA testing system. Thus, it is envisaged that the spatial correlation in practical MIMO OTA testing systems may not be exactly equivalent to the expected value of the ideal theoretical model. In this paper, we propose a new evaluation framework for evaluating the spatial correlation performance of MIMO OTA testing system. This evaluation framework provides a novel observation method for spatial correlation, which reflects the non-ideal configuration of the MIMO OTA testing system and can be utilized to predict or cross-validate spatial correlation errors that deviate from the theoretical model. The experimental and simulation results have been verified against the theoretical model, demonstrating good consistency between the theoretical model and the proposed evaluation framework. Furthermore, several test scenarios have been verified with the different varying factors.

Imbalance of autonomic nerve is the cause of lots of cardiovascular and psychological diseases. Many studies have shown that Heart Rate Variability Biofeedback through respiratory training is an effective way to regulate the status of human autonomic nerves. However, due to the immature quantification of Respiratory Sinus Arrhythmia, the respiratory training process cannot be optimal personalized. In this paper, we propose the Cardiopulmonary Resonance Indices based on the Cardiopulmonary Coupling Model to calculate the personalized optimal breathing training frequency. Then the respiratory feedback training system is constructed to provide real-time audio-visual guidance for each subject. The results in the clinical trial of pregnant women in their third trimester showed that, compared to control group, with the intervention of personalized respiratory feedback training at home, pregnant women could stabilize their blood pressure and blood glucose, relieve mental stress, as well as exert positive effects on fetal development. This provides a new possibility of the autonomic imbalance treatment.

Based on an EMG-adjusted method in neuromusculoskeletal model, this study aims to predict the individual muscle force in shoulder and elbow during table tennis strokes. Muscle force estimation makes muscle activation analysis more physiological in sports. Twenty subjects, divided into professional group and amateur group, were adopted in this study. They were asked to do a basic stoke motion: backhand block. Surface electromyography (sEMG) of nine muscles was recorded, as well as the motion data collected by three inertial sensors. A Hill-type musculotendon model was then adopted to estimate individual muscle force by combining adjusted sEMG and motion data. The result shows that the method can estimate individual muscle force during table tennis strokes accurately, and the two groups show significant difference in muscle force of shoulders and elbows.

Miniaturized Inertial Measurement Unit (IMU) has been widely used in many motion capturing applications. In order to overcome stability and noise problems of IMU, a lot of efforts have been made to develop appropriate data fusion method to obtain reliable orientation estimation from IMU data. This article presents a method which models the errors of orientation, gyroscope bias and magnetic disturbance, and compensate the errors of state variables with complementary Kalman filter in a body motion capture system. Experimental results have shown that the proposed method significantly reduces the accumulative orientation estimation errors.

In the past decade, golf has stimulated people’s great interest and the number of golf players has increased significantly. Therefore, how to train a golfer to make a perfect swing has attracted extensive research attentions. Among these researches, the most important step is to capture and reconstruct the swing movement in a transportable and non-intrusive way. Restricted by the development of present depth imaging devices, the initial captured swing movement may not be acceptable due to occlusions and mixing up of body parts. In this paper, to restore motion information from self-occlusion and reconstruct 3D golf swing from low resolution data, a Dynamic Bayesian Network (DBN) model based golf swing reconstruction algorithm is proposed to increase the capture accuracy considering the spatial and temporal similarities of swing between different golfers. A Smart Motion Reconstruction system for Golf swing, SMRG, is presented based on the DBN model with a popular depth imaging device, Kinect, as capturing device. Experimental results have proved that the proposed system can achieve comparable reconstruction accuracy to the commercial optical motion caption (OMocap) system and better performance than state of art modification algorithms using depth information.

Ambulatory blood pressure is critical in predicting some major cardiovascular events; therefore, cuff-less and noninvasive beat-to-beat ambulatory blood pressure measurement is of great significance. Machine-learning methods have shown the potential to derive the relationship between physiological signal features and ABP. In this paper, we apply random forest method to systematically explorer the inherent connections between photoplethysmography signal, electrocardiogram signal and ambulatory blood pressure. To archive this goal, 18 features were extracted from PPG and ECG signals. Several models with most significant features as inputs and beat-to-beat ABP as outputs were trained and tested on data from the Multi-Parameter Intelligent Monitoring in Intensive Care II database. Results indicate that compared with the common pulse transit time method, the RF method gives a better performance for one-hour continuous estimation of diastolic blood pressure and systolic blood pressure under both the Association for the Advancement of Medical Instrumentation and British Hyper-tension Society standard.

Quantitative measures of autonomic regulation of cardiovascular system have clinical and prognostic value in a variety of cardiovascular diseases. This paper proposes a model-based method in the measurement of baroreflex sensitivity, sympathetic and parasympathetic activity. The method measures the continuous blood pressure and heart rate in orthostatic scenario, models the baroreflex and sympathetic regulation process, solves for personalized model parameters by optimization using measured blood pressure and heart rate variations. Experimental results have shown the validation of the quantitative measures and the effectiveness of the method.

Commercial optical human motion capture systems perform well in studio-like environments, but they do not provide solution in daily-life surroundings. Micro-sensor motion capture has shown its potentials because of its ubiquity and low cost. We present an ambulatory low-cost real-time motion capture system using wearable micro-sensors (accelerometers, magnetometers and gyroscopes), which can capture and reconstruct human motion in real-time almost everywhere. It mainly consists of three parts: a sensor subsystem, a data fusion subsystem and an animation subsystem. The sensor subsystem collects human motion signals and transfers them into the data fusion subsystem. The data fusion subsystem performs sensor fusion to obtain motion information, i.e., the orientation and position of each body segment. Using the motion information from the data fusion subsystem, the animation subsystem drives the avatar in the 3D virtual world in order to reconstruct human motion. All the processes are accomplished in real-time. The experimental results show that our system can capture motions and drive animations in real-time vividly without drift and delay. And the output from our system can be made use of in film-making, sports training and argument reality applications, etc.

Accurate human movement estimation techniques are widely used in various applications, such as robotics, human-machine interaction, sports, and rehabilitation. With rapid advances in microsensors, human movement estimation using wearable micro inertial sensors has become an active research topic. The main challenges for the wearable sensor motion estimation are the inertial sensor drift problem and the linear acceleration interference problem. Because of the agility in movement, upper limb motion estimation has been regarded as the most difficult problem in human motion estimation. In this paper, we take the upper limb as our research subject and present a novel upper limb movement estimation algorithm to cope with these two challenges by adaptive fusion of sensor data and human skeleton constraint. In the sensor fusion part, a quaternion-based unscented Kalman filter is invoked to fuse the gyroscope, accelerometer, and magnetometer measurement information. In the Kalman filter framework, an acceleration interference detection scheme is implemented based on the exponentially discounted average of the normalized innovation squared (NIS). According to the detection results, the process and measurement noise levels are scaled up or down automatically. To further compensate for the drift, we present a novel solution by modeling geometrical constraint in the elbow joint and fuse the constraint to revise the sensor fusion results and improve the estimation accuracy. The experimental results have shown that the proposed algorithm can provide accurate results in comparison to the BTS SMART-D optical motion tracker.

Tracking multiple maneuvering targets remains a challenge because of clutter and spurious targets. We propose a Signature Driven multiple target Tracking (SDT) method which uses target signature in spectral, spatial and temporary spaces as well as the Markov property of target movement, so that the data association process in SDT is very efficient and effective. The experimental results have shown outstanding performance.

Hip angle is a major parameter in gait analysis while gait analysis plays important role in health-care, animation and other applications. Accurate and robust estimation of hip angle in ambulatory environment remains a challenge because the non-linear nature of thigh movement has not been well studied yet. Although piece-wise linear model is effective to approximate the non-linear model, the current solutions, Gaussian Particle Filter (GPF), is suffering from heavily computation load, which makes the ambulatory hip angle estimation in real time impossible. In this paper, we propose to use Discrete Wavelet Transform to detect major gait events from the measurements of the wearable accelerometer that are attached to the thigh. Based on the detection result, a corresponding linear hip angle dynamic is selected and an Unscented Kalman Filter (UKF) is invoked to estimate the hip angle. The experimental results have shown that the proposed methods can achieve robust and accurate hip angle estimation, and with much less computation loads over the previous work on the ambulatory gait analysis.