Falling is one of the causes of accidental death of elderly people over 65 years old in Taiwan. If the fall incidents are not detected in a timely manner, it could lead to serious injury or even death of those who fell. General fall detection approaches require the users to wear sensors, which could be cumbersome for the users to put on, and misalignment of sensors could lead to erroneous readings. In this paper, we propose using computer vision and applied machine-learning algorithms to detect fall without any sensors. We applied OpenPose real-time multi-person 2D pose estimation to detect movement of a subject using two datasets of 570 × 30 frames recorded in five different rooms from eight different viewing angles. The system retrieves the locations of 25 joint points of the human body and detects human movement through detecting the joint point location changes. The system is able to effectively identify the joints of the human body as well as filtering ambient environmental noise for an improved accuracy. The use of joint points instead of images improves the training time effectively as well as eliminating the effects of traditional image-based approaches such as blurriness, light, and shadows. This paper uses single-view images to reduce equipment costs. We experimented with time series recurrent neural network, long- and short-term memory, and gated recurrent unit models to learn the changes in human joint points in continuous time. The experimental results show that the fall detection accuracy of the proposed model is 98.2%, which outperforms the baseline 88.9% with 9.3% improvement.

In recent years, wearable monitoring devices have been very popular in the health care field and are being used to avoid sport injuries during exercise. They are usually worn on the wrist, the same as sport watches, or on the chest, like an electrocardiogram patch. Common functions of these wearable devices are that they use real time to display the state of health of the body, and they are all small sized. The electromyogram (EMG) signal is usually used to show muscle activity. Thus, the EMG signal could be used to determine the muscle-fatigue conditions. In this study, the goal is to develop an EMG patch which could be worn on the lower leg, the gastrocnemius muscle, to detect real-time muscle fatigue while exercising. A micro controller unit (MCU) in the EMG patch is part of an ARM Cortex-M4 processor, which is used to measure the median frequency (MF) of an EMG signal in real time. When the muscle starts showing tiredness, the median frequency will shift to a low frequency. In order to delete the noise of the isotonic EMG signal, the EMG patch has to run the empirical mode decomposition algorithm. A two-electrode circuit was designed to measure the EMG signal. The maximum power consumption of the EMG patch was about 39.5 mAh. In order to verify that the real-time MF values measured by the EMG patch were close to the off-line MF values measured by the computer system, we used the root-mean-square value to estimate the difference in the real-time MF values and the off-line MF values. There were 20 participants that rode an exercise bicycle at different speeds. Their EMG signals were recorded with an EMG patch and a physiological measurement system at the same time. Every participant rode the exercise bicycle twice. The averaged root-mean-square values were 2.86 ± 0.86 Hz and 2.56 ± 0.47 Hz for the first and second time, respectively. Moreover, we also developed an application program implemented on a smart phone to display the participants’ muscle-fatigue conditions and information while exercising. Therefore, the EMG patch designed in this study could monitor the muscle-fatigue conditions to avoid sport injuries while exercising.

In recent years, privacy awareness is concerned due to many Internet services have chosen to use encrypted agreements. In order to improve the quality of service (QoS), the network encrypted traffic behaviors are classified based on machine learning discussed in this paper. However, the traditional traffic classification methods, such as IP/ASN (Autonomous System Number) analysis, Port-based and deep packet inspection, etc., can classify traffic behavior, but cannot effectively handle encrypted traffic. Thus, this paper proposed a hybrid traffic classification (HTC) method based on machine learning and combined with IP/ASN analysis with deep packet inspection. Moreover, the majority voting method was also used to quickly classify different QoS traffic accurately. Experimental results show that the proposed HTC method can effectively classify different encrypted traffic. The classification accuracy can be further improved by 10% with majority voting as K = 13. Especially when the networking data are using the same protocol, the proposed HTC can effectively classify the traffic data with different behaviors with the differentiated services code point (DSCP) mark.

As transmission speeds increase faster than processing speeds, the packet processing time (PPT) of a host is becoming more significant in the measurement of different network parameters in which packet processing by the host is involved. The PPT of a host is the time elapsed between the arrival of a packet at the data-link layer and the time the packet is processed at the application layer (RFCs 2679 and 2681). To measure the PPT of a host, stamping the times when these two events occur is needed. However, time stamping at the data-link layer may require placing a specialized packet-capture card and the host under test in the same local network. This makes it complex to measure the PPT of remote end hosts. In this paper, we propose a scheme to measure the PPT of an end host connected over a single- or multiple-hop path and without requiring time stamping at the data-link layer. The proposed scheme is based on measuring the capacity of the link connected to the host under test. The scheme was tested on an experimental testbed and in the Internet, over a U.S. inter-state path and an international path between Taiwan and the U.S. We show that the proposed scheme consistently measures PPT of a host.

Current schemes for configuration of input-queued three-stage Clos-network (IQC) switches involve port matching and path routing assignment, in that order. The implementation of a scheduler capable of matching thousands of ports in large-size switches is complex. To decrease the scheduler complexity for such switches (e.g., 1024 ports or more), we propose a configuration scheme for IQC switches that hierarchizes the matching process. In a practical scenario our scheme performs routing first and port matching thereafter. This approach reduces the scheduler size and the configuration complexity of IQC switches. We show that the switching performance of the proposed approach using weight-based and weightless selection schemes is high under uniform and nonuniform traffic