Visually impaired people are often unaware of dangers in front of them, even in familiar environments. Furthermore, in unfamiliar environments, such people require guidance to reduce the risk of colliding with obstacles. This study proposes a simple smartphone-based guiding system for solving the navigation problems for visually impaired people and achieving obstacle avoidance to enable visually impaired people to travel smoothly from a beginning point to a destination with greater awareness of their surroundings. In this study, a computer image recognition system and smartphone application were integrated to form a simple assisted guiding system. Two operating modes, online mode and offline mode, can be chosen depending on network availability. When the system begins to operate, the smartphone captures the scene in front of the user and sends the captured images to the backend server to be processed. The backend server uses the faster region convolutional neural network algorithm or the you only look once algorithm to recognize multiple obstacles in every image, and it subsequently sends the results back to the smartphone. The results of obstacle recognition in this study reached 60%, which is sufficient for assisting visually impaired people in realizing the types and locations of obstacles around them.

In the clinic, the wheezing sound is usually considered as an indicator symptom to reflect the degree of airway obstruction. The auscultation approach is the most common way to diagnose wheezing sounds, but it subjectively depends on the experience of the physician. Several previous studies attempted to extract the features of breathing sounds to detect wheezing sounds automatically. However, there is still a lack of suitable monitoring systems for real-time wheeze detection in daily life. In this study, a wearable and wireless breathing sound monitoring system for real-time wheeze detection was proposed. Moreover, a breathing sounds analysis algorithm was designed to continuously extract and analyze the features of breathing sounds to provide the objectively quantitative information of breathing sounds to professional physicians. Here, normalized spectral integration (NSI) was also designed and applied in wheeze detection. The proposed algorithm required only short-term data of breathing sounds and lower computational complexity to perform real-time wheeze detection, and is suitable to be implemented in a commercial portable device, which contains relatively low computing power and memory. From the experimental results, the proposed system could provide good performance on wheeze detection exactly and might be a useful assisting tool for analysis of breathing sounds in clinical diagnosis.

Capturing hand motions for hand function evaluations is essential in the medical field. Various data gloves have been developed for rehabilitation and manual dexterity assessments. This study proposed a modular data glove with 9-axis inertial measurement units (IMUs) to obtain static and dynamic parameters during hand function evaluation. A sensor fusion algorithm is used to calculate the range of motion of joints. The data glove is designed to have low cost, easy wearability, and high reliability. Owing to the modular design, the IMU board is independent and extensible and can be used with various microcontrollers to realize more medical applications. This design greatly enhances the stability and maintainability of the glove.

This paper proposes a data glove system integrated with six-axis inertial measurement unit sensors for evaluating the hand function of patients who have suffered a stroke. The modular design of this data glove facilitates its use for stroke patients. The proposed system can use the hand's accelerations, angular velocities, and joint angles as calculated by a quaternion algorithm, to help physicians gain new insights into rehabilitation treatments. A clinical experiment was performed on 15 healthy subjects and 15 stroke patients whose Brunnstrom stages (BSs) ranged from 4 to 6. In this experiment, the participants were subjected to a grip task, thumb task, and card turning task to produce raw data and three features, namely, the average rotation speed, variation of movement completion time, and quality of movement; these features were extracted from the recorded data to form 2-D and 3-D scatter plots. These scatter plots can provide reference information and guidance to physicians who must determine the BSs of stroke patients. The proposed system demonstrated a hit rate of 70.22% on average. Therefore, this system can effectively reduce physicians' load and provide them with detailed information about hand function to help them adjust rehabilitation strategies for stroke patients.

Progress in wearable techniques makes the long-term daily electrocardiogram (ECG) monitoring possible. However, the long-term wearable ECGs can be significantly contaminated by various noises, which affect the detection and diagnosis of cardiovascular diseases (CVDs). The situation becomes more serious for wearable ECG screening, where the data are huge and doctors have no way to visually check the signal quality episode-by-episode. Therefore, automatic and accurate noise rejection for the wearable big-data ECGs is craving. This study addressed this issue and proposed a noise rejection method for wearable ECGs based on the combination of modified frequency slice wavelet transform (MFSWT) and convolution neural network (CNN). Wearable ECGs were recorded using the newly developed 12-lead Lenovo smart ECG vest with a sample rate of 500 Hz and a resolution of 16 bit. One thousand 10-s ECG segments were picked up and were manually labeled into three quality types: clinically useful segments with good signal quality (type A), clinically useful segments with poor signal quality (type B) and clinically useless segments (pure noises, type C). Each of the 1,000 10-s ECG segments were transformed into a 2-D time-frequency (T-F) image using MFSWT, with a pixel size of 200×50. Then, the 2-D grayscale images from MFSWT were fed into a 13-layer CNN model for training the classification models. Results from the standard 5-folder cross validation showed that, the proposed combination method of MFSWT and CNN achieved a highest classification accuracy of 86.3%, which was higher than the comparable methods from continuous wavelet transform (CWT) and artificial neural networks (ANN). In addition, CNN classifier required more calculation cost than comparable ANN model. This study indicated that combination of MFSWT and CNN is a potential method for automatic identification of noisy segments from wearable ECG recordings.

Falls are the second leading cause of death from unintentional injuries in older adults. Although many systems have been used to detect falls, they are limited by the computational complexity of the algorithm. The images taken by the camera must be transmitted through a network to the back-end server for calculation. As the demand for Internet of Things increases, this architecture faces such problems as high bandwidth costs and server computing overload. Emerging methods reduce the workload of servers by transferring certain computing tasks from cloud servers to edge computing platforms. To this end, this study developed a fall detection system based on neuromorphic computing hardware, which streamlines and transplants the neural network model of the back-end computer to the edge computing platform. Through the neural network model with integer 8 bit precision deployed on the edge computing platform, the object photos obtained by the camera are converted into human motion features, and a support vector machine is then used for classification. After experimental evaluation, an accuracy of 96% was reached, the detection speed of the overall system was 11.5 frames per second, and the power consumption was 0.3 W. This system can monitor the fall events of older adults in real time and over a long period. All data were calculated on the edge computing platform. The system only reports fall events via Wi-Fi, thereby protecting the privacy of the user.