Frequent blood glucose monitoring is crucial for patients with diabetes. However, current blood glucose measurement methods are primarily invasive and, thus, cause discomfort and infection risks. Accordingly, this study developed a noninvasive real-time glucose monitoring system based on a deep learning (DL) model (comprising convolutional neural network and long short-term memory network models) and multiwavelength near-infrared (NIR) light technology. The system is equipped with a portable finger gripper with NIR light-emitting diodes emitting at three distinct wavelengths (810, 860, and 940 nm) to illuminate the finger. A triad spectroscopy sensor is then used to capture finger photoplethysmography (PPG) signals within only 6 s. The captured signals are then transmitted to a server-side DL model for glucose prediction. This DL model analyzes the three-wavelength PPG data to predict a user’s blood glucose value. The predicted glucose value is subsequently displayed on a dedicated smartphone app. In contrast to previously proposed systems relying on machine learning for feature extraction, the proposed system uses a DL model to automatically extract glucose-related features from the three-band PPG signals, ultimately leading to blood glucose predictions with a root-mean-square error of 6.62. Furthermore, the proposed system prioritizes user comfort, portability, and stability, thereby offering a convenient and accessible blood glucose monitoring experience.

Although high heels are commonly worn for their perceived aesthetic value, they can have negative effects on the health of the foot and lower limbs. Wearing excessively high heels for long periods can negatively affect the wearer's ankles, knees, and spine. In this study, a smartphone app based on the You Only Look Once (YOLO) v8 model was developed to provide users with personalized maximum height recommendations for high-heeled shoes to reduce these health risks. Specifically, YOLOv8n identified heels in videos of participants transitioning from standing flat-footed to tip-toe, and a weighting algorithm was used to identify the maximum appropriate high-heel height. Extreme values were suppressed to ensure that the predicted value was safe and sufficiently conservative. The system was optimized for rapid prediction, and strategies for ensuring accurate predictions against backgrounds of different colors were developed. The system achieved a lower mean absolute error (MAE) of 4.33 mm on white backgrounds than previous studies and is the first reported system to also achieve high accuracy for backgrounds of other colors, namely green and black backgrounds with MAEs of 5.25 and 6.84 mm, respectively. The system has excellent performance and practical applicability. Moreover, unlike our previously developed client-server system, this lightweight system can be implemented locally on a smartphone, greatly improving convenience and privacy.

Energy expenditure (EE) is often used to quantify physical activity. Currently, EE is estimated with data collected by inertial measurement units or depth cameras and verified by oxygen consumption data. Due to the different data collection time spans in this system, raw data were split into minute-by-minute windows, and summary statistics for each window were computed. However, using summary statistics to aggregate data might be influenced by redundant noise or result in the loss of valuable information. This article presents a modeling method using functional analysis to characterize the trajectory of the collected skeletal data, thus enabling the effective use of the complete data. Next, the fitted values of the skeletal data can be aligned to the overall EE data and used to predict the overall EE as well as the task-based EE. The study results revealed for metabolic equivalent of task prediction that the root-mean-square error (RMSE) derived for the proposed method was$< $0.5 and that the mean absolute error (MAE) was approximately 0.3. Models for estimating task-based EE, including EE related to standing and walking task, also exhibited low RMSE and MAE values. Accordingly, the proposed modeling approach is superior to summary statistics for estimating EE in depth camera systems.

Spasticity is a common complication for patients with stroke, but only few studies investigate the relation between spasticity and voluntary movement. This study proposed a novel automatic system for assessing the severity of spasticity (SS) of four upper-limb joints, including the elbow, wrist, thumb, and fingers, through voluntary movements. A wearable system which combined 19 inertial measurement units and a pressure ball was proposed to collect the kinematic and force information when the participants perform four tasks, namely cone stacking (CS), fast flexion and extension (FFE), slow ball squeezing (SBS), and fast ball squeezing (FBS). Several time and frequency domain features were extracted from the collected data, and two feature selection approaches based on recursive feature elimination were adopted to select the most influential features. The selected features were input into five machine learning techniques for assessing the SS for each joint. The results indicated that using CS task to assess the SS of elbow and fingers and using FBS task to assess the SS of thumb and wrist can reach the highest weighted-average F1-score. Furthermore, the study also concluded that FBS is the optimal task for assessing all the four upper-limb joints. The overall result shown that the proposed automatic system can assess four upper-limb joints through voluntary movements accurately, which is a breakthrough of finding the relation between spasticity and voluntary movement.

Long-term poor sitting posture leads to physical injuries such as muscle soreness and waist and neck alignment problems. In this study, we proposed an intelligent sitting posture detection system that uses depth cameras fixed on a chair to capture depth images of the user's sitting posture, and then applies a trained artificial intelligence (AI) model on an embedded Raspberry Pi board to recognize the user's sitting posture from the image data. Finally, through Bluetooth on the Raspberry Pi, the results are sent to the user's smartphone application for display and recording to achieve rapid detection of sitting posture and warning of poor sitting posture. The contribution of this study is its use of two depth cameras mounted on a chair, thereby eliminating the problem of cumbersome sensors that compromise user comfort or are prone to damage. The detection of the user's entire sitting posture was completed on an edge computing platform, which leads to power savings and offers privacy protection. Furthermore, because of the low battery power usage, the system is portable. To perform quick AI calculations, we developed a lightweight EfficientNet model and programmed it for the Raspberry Pi. The system achieved an accuracy of 99.71% and an execution speed of almost one posture result per second.

Electrocardiogram (ECG) is one of the most important information for cardiovascular diseases (CVDs) diagnosis. In recent year, several dry electrode-based smart clothes have been widely developed to improve the skin allergic reaction and gel-drying issue from conventional Ag/AgCl electrode under long-term measurement. However, most of these dry electrodes still have to contact with skin and may encounter the risk of skin irritation, and many smart clothing systems lack of automatic CVDs detection. In this article, a novel smart clothing was designed to automatically detect CVDs in daily life. Based on the technique of capacitive electrodes, the proposed smart clothing could access the bio-potential across the clothes to prevent the skin from irritation and discomfort, and could adapt to different body sizes by the specific belt mechanical design. Moreover, the CVDs detection algorithm was also designed and implemented in the field programmable gate array (FPGA) based ECG analysis module. The experiment results show that the proposed smart clothing could effectively real-time extract ECG features (P-, R-, and T-waves) and detect CVDs state via the front-end circuit, including bradycardia, tachycardia, atrial fibrillation, left ventricular hypertrophy, first-degree atrioventricular block, and hyperkalemia. The proposed FPGA architecture is also beneficial for future revisions or additions of CVD algorithms to improve more accurate diagnosis and monitoring of heart disease. It might reduce huge ECG data collected in daily life via only transmitting the abnormal ECG segment, and improve the diagnostic efficiency of CVDs in the future.

Arteriosclerosis is a cardiovascular disease that can cause calcification, sclerosis, stenosis, or obstruction of blood vessels and may further cause abnormal peripheral blood perfusion or other complications. In clinical settings, several approaches, such as computed tomography angiography and magnetic resonance angiography, can be used to evaluate arteriosclerosis status. However, these approaches are relatively expensive and require an experienced operator and often the injection of a contrast agent. In this article, a novel smart assistance system based on near-infrared spectroscopy was proposed that can noninvasively assess blood perfusion and thus indicate arteriosclerosis status. In this system, a wireless peripheral blood perfusion monitoring device simultaneously monitors changes in hemoglobin parameters and the cuff pressure applied by a sphygmomanometer. Several indexes extracted from changes in hemoglobin parameters and cuff pressure were defined and can be used to estimate blood perfusion status. A neural network model for arteriosclerosis evaluation was constructed using the proposed system. The relationship between the blood perfusion indexes and arteriosclerosis status was investigated, and the neural network model for arteriosclerosis evaluation was validated. Experimental results indicated that the differences in many blood perfusion indexes for different groups were significant and that the neural network model could effectively evaluate arteriosclerosis status (accuracy = 80.26%). By using a sphygmomanometer, the model can be employed for simple arteriosclerosis screening and blood pressure measurements. The model offers real-time noninvasive measurement, and the system is relatively inexpensive and easy to operate.

Bedside falls and pressure ulcers are crucial issues in geriatric care. Although many bedside monitoring systems have been proposed, they are limited by the computational complexity of their algorithms. Moreover, most of the data collected by the sensors of these systems must be transmitted to a back-end server for calculation. With an increase in the demand for the Internet of Things, problems such as higher cost of bandwidth and overload of server computing are faced when using the aforementioned systems. To reduce the server workload, certain computing tasks must be offloaded from cloud servers to edge computing platforms. In this study, a bedside monitoring system based on neuromorphic computing hardware was developed to detect bedside falls and sleeping posture. The artificial intelligence neural network executed on the back-end server was simplified and used on an edge computing platform. An integer 8-bit-precision neural network model was deployed on the edge computing platform to process the thermal image captured by the thermopile array sensing element to conduct sleep posture classification and bed position detection. The bounding box of the bed was then converted into the features for posture classification correction to correct the posture. In an experimental evaluation, the accuracy rate, inferencing speed, and power consumption of the developed system were 94.56%, 5.28 frames per second, and 1.5 W, respectively. All the calculations of the developed system are conducted on an edge computing platform, and the developed system only transmits fall events to the back-end server through Wi-Fi and protects user privacy.

Inflammation of the long head of the biceps tendon is a common cause of shoulder pain. Bicipital peritendinous effusion (BPE) is the most common biceps tendon abnormality and is related to various shoulder injuries. Physicians usually use ultrasound imaging to grade the inflammation severity of the long head of the biceps tendon. However, obtaining a clear and accurate ultrasound image is difficult for inexperienced attending physicians. To reduce physicians' workload and avoid errors, an automated BPE recognition system was developed in this article for classifying inflammation into the following categories-normal and mild, moderate, and severe. An ultrasound image serves as the input in the proposed system; the system determines whether the ultrasound image contains biceps. If the image depicts biceps, then the system predicts BPE severity. In this study, two crucial methods were used for solving problems associated with computer-aided detection. First, the faster regions with convolutional neural network (faster R-CNN) used to extract the region of interest (ROI) area identification to evaluate the influence of dataset scale and spatial image context on performance. Second, various CNN architectures were evaluated and explored. Model performance was analyzed by using various network configurations, parameters, and training sample sizes. The proposed system was used for three-class BPE classification and achieved 75% accuracy. The results obtained for the proposed system were determined to be comparable to those of other related state-of-the-art methods.

Speech is easily affected by different background noise in real environment to reduce the speech intelligibility, in particular, for hearing impaired listeners. In order to improve the above issue, several hearing aids have been developed to enhance the speech signal in noisy environment. Most of current hearing aids were designed to enhance the component of speech and suppress the component of noise. However, it is difficult to separate other speech sources. Adaptive signal enhancement with the beamforming technique might improve the above issue. However, how to distinguish the location of the desired speaker effectively is still a difficult challenge for adaptive beamforming method. A novel concept of hearing aid was proposed in this study. Different from the beamforming-based hearing aids, which use the crosscorrelation-coefficient method to estimate time difference of arrival (TDOA), an image recognition technology was used to estimate the location of the desired speaker to obtain the more precise TDOA. An adaptive signal enhancement was also used to enhance the noisy speech sound. From the experimental results, the proposed system could provide a smaller absolute error of TDOA less than 1.25 × 10-4ms, and a clear speech sound from the target speaker who the user wants to listen to.

Energy expenditure (EE) monitoring is crucial to tracking physical activity (PA). Accurate EE monitoring may help people engage in adequate activity and therefore avoid obesity and reduce the risk of chronic diseases. This study proposes a depth-camera-based system for EE estimation of PA in gyms. Most previous studies have used inertial measurement units for EE estimation. By contrast, the proposed system can be used to conveniently monitor subjects' treadmill workouts in gyms without requiring them to wear any devices. A total of 21 subjects were recruited for the experiment. Subjects' skeletal data acquired using the depth camera and oxygen consumption data simultaneously obtained using the K4b2 device were used to establish an EE predictive model. To obtain a robust EE estimation model, depth cameras were placed in the side view, rear side view, and rear view. A comparison of five different predictive models and these three camera locations showed that the multilayer perceptron model was the best predictive model and that placing the camera in the rear view provided the best EE estimation performance. The measured and predicted metabolic equivalents of task exhibited a strong positive correlation, with r = 0.94 and coefficient of determination r2 = 0.89. Furthermore, the mean absolute error was 0.61 MET, mean squared error was 0.67 MET, and root mean squared error was 0.76 MET. These results indicate that the proposed system is handy and reliable for monitoring user's EE when performing treadmill workouts.

In this paper, a higher-order-statistics (HOS)-based radial basis function (RBF) network for signal enhancement is introduced. In the proposed scheme, higher order cumulants of the reference signal were used as the input of HOS-based RBF. An HOS-based supervised learning algorithm, with mean square error obtained from higher order cumulants of the desired input and the system output as the learning criterion, was used to adapt weights. The motivation is that the HOS can effectively suppress Gaussian and symmetrically distributed non-Gaussian noise. The influence of a Gaussian noise on the input of HOS-based RBF and the HOS-based learning algorithm can be mitigated. Simulated results indicate that HOS-based RBF can provide better performance for signal enhancement under different noise levels, and its performance is insensitive to the selection of learning rates. Moreover, the efficiency of HOS-based RBF under the nonstationary Gaussian noise is stable.

This paper demonstrates the design and implementation of a real-time wireless physiological monitoring system for nursing centers, whose function is to monitor online the physiological status of aged patients via wireless communication channel and wired local area network. The collected data, such as body temperature, blood pressure, and heart rate, can then be stored in the computer of a network management center to facilitate the medical staff in a nursing center to monitor in real time or analyze in batch mode the physiological changes of the patients under observation. Our proposed system is bidirectional, has low power consumption, is cost effective, is modular designed, has the capability of operating independently, and can be used to improve the service quality and reduce the workload of the staff in a nursing center.