論文

In this article, a deep-Q network (DQN)-based approach is proposed to improve the harvested energy and the secrecy capacity by optimizing the beamforming vectors in multiple-input-multiple-output (MIMO) networks with simultaneous wireless information and power transmission (SWIPT). Here, in the presence of multiantenna eavesdroppers, a multiantenna hybrid access point (HAP) serves energy harvesting Internet of Things (EH-IoT) nodes and information decoding (ID) receiver. The EH-IoT nodes and the ID receiver are separated from each other and experience different channels from the HAP. Two cases of signal models are studied to maximize the harvested energy. The first case is the consolidated signal case, where the HAP transmits an information-bearing signal to both the EH-IoT nodes and the ID receiver. The second case is the split signal case, where the HAP sends both information-carrying signal and energy signals to the EH-IoT nodes and the ID receiver. For both these cases, the optimization strategies for maximizing the harvested energy under Quality-of-Service (QoS) constraints are formulated with the secrecy capacity and the harvested energy requirements. The harvested energy maximization problems are complex to handle due to the presence of high-dimensional variables. An efficient, robust, and secure DQN-based algorithm is proposed, where Q-learning and deep neural network (DNN) methods are extended to solve these problems. Simulation results show that the proposed DQN approach can achieve a balance between harvested energy and security. By incorporating action and reward spaces that are adaptable to the formulated problem, significant performance gains are achieved.

Malware traffic classification (MTC) is one of the important techniques to ensure the security of cyberspace, which aims to detect anomalies and classify different types of network traffic. Recently, MTC methods based on deep learning (DL) have shown their excellent performance. However, these DL-based methods rely on data sets with manually labeled samples for training, which are costly and hard to obtain. To address this problem, this article proposes a novel self-supervised MTC method based on the framework of masked auto-encoder (MAE). Specifically, MAE first constructs a reasonable unsupervised pretext task with a random masking strategy, which reduces the redundant information in samples and speeds up the pretraining process. The transformer-based backbone network then efficiently extracts features from the nonredundant traffic data efficiently. The proposed MTC-MAE method employs self-supervised learning on a large-scale unlabeled data set to acquire unbiased features, and fine-tunes on specific data sets to adapt to diverse traffic classification scenarios. Simulation experiments show that our proposed MTC-MAE method is able to learn universal features with high quality and has excellent classification performance on various downstream data sets. The data sets we used, code implementation, and pretrained models are available on GitHub. Code available at https://github.com/TsuiHark/Self-supervised_MTC.

In health monitoring systems for the elderly, a crucial aspect is unobtrusively and continuously monitoring their activities to detect potentially hazardous incidents such as sudden falls as soon as they occur. However, the effectiveness of current non-contact sensor-based activity detection systems is limited by obstacles present in the environment. To overcome this limitation, a straightforward yet highly efficient approach involves utilizing multiple sensors that collaborate seamlessly. This paper proposes a method that leverages 2D Light Detection and Ranging (Lidar) technology for activity detection. Multiple 2D Lidars are positioned in an indoor environment with varying obstacles such as furniture, working cohesively to create a comprehensive representation of ongoing activities. The data from these Lidars is concatenated and transformed into a more interpretable format, resembling images. A convolutional Long Short-Term Memory (LSTM) Neural Network is then used to process these generated images to classify the activities. The proposed approach achieves high accuracy in three tasks: activity detection, fall detection, and unsteady gait detection. Specifically, it attains accuracies of 96.10%, 99.13%, and 93.13% for these tasks, respectively. This demonstrates the efficacy and promise of the method in effectively monitoring and identifying potentially hazardous events for the elderly through 2D Lidars, which are non-intrusive sensing technology.

A growing focus among scientists has been on researching the techniques of automatic detection of dementia that can be applied to the speech samples of individuals with dementia. Leveraging the rapid advancements in Deep Learning (DL) and Natural Language Processing (NLP), these techniques have shown great potential in dementia detection. In this context, this paper proposes a method for dementia detection from the transcribed speech of subjects. Unlike conventional methods that rely on advanced language models to address the ability of the subject to make coherent and meaningful sentences, our approach relies on the center of focus of the subjects and how it changes over time as the subject describes the content of the cookie theft image, a commonly used image for evaluating one’s cognitive abilities. To do so, we divide the cookie theft image into regions of interest, and identify, in each sentence spoken by the subject, which regions are being talked about. We employed a Long Short-Term Memory (LSTM) neural network to learn different patterns of dementia subjects and control ones and used it to perform a 10-fold cross validation-based classification. Our experimental results on the Pitt corpus from the DementiaBank resulted in a 82.9% accuracy at the subject level and 81.0% at the sample level. By employing data-augmentation techniques, the accuracy at both levels was increased to 83.6% and 82.1%, respectively. The performance of our proposed method outperforms most of the conventional methods, which reach, at best, an accuracy equal to 81.5% at the subject level.

Fetal heart rate (FHR) monitoring, typically using Doppler ultrasound (DUS) signals, is an important technique for assessing fetal health. In this work, we develop a robust DUS-based FHR estimation approach complemented by DUS signal quality assessment (SQA) based on unsupervised representation learning in response to the drawbacks of previous DUS-based FHR estimation and DUS SQA methods. We improve the existing FHR estimation algorithm based on the autocorrelation function (ACF), which is the most widely used method for estimating FHR from DUS signals. Short-time Fourier transform (STFT) serves as a signal pre-processing technique that allows the extraction of both temporal and spectral information. In addition, we utilize double ACF calculations, employing the first one to determine an appropriate window size and the second one to estimate the FHR within changing windows. This approach enhances the robustness and adaptability of the algorithm. Furthermore, we tackle the challenge of low-quality signals impacting FHR estimation by introducing a DUS SQA method based on unsupervised representation learning. We employ a variational autoencoder (VAE) to train representations of pre-processed fetal DUS data and aggregate them into a signal quality index (SQI) using a self-organizing map (SOM). By incorporating the SQI and Kalman filter (KF), we refine the estimated FHRs, minimizing errors in the estimation process. Experimental results demonstrate that our proposed approach outperforms conventional methods in terms of accuracy and robustness.

Noncontact heartbeat detection and heart rate (HR) estimation have been hot topics of research over the last few years. By employing wireless sensors, such as Doppler sensors, the goal of these tasks is to detect subtle movements associated with heart movements. The detection is either used to reconstruct the R-peaks, thus the heartbeat detection, or simply identify the HR. However, due to the nature of the faint movements of the chest caused by the heart beats, unless the person is still, such movements are overwhelmingly obscured by the motion of the person. That being the case, we are interested in this work in identifying such instances where the person is not in motion, in which case, we measure their heartbeats and HR. To do so, we use a combination of a 3-D light detection and ranging (LiDAR) and a multiple-input multiple-output (MIMO) Doppler Radar. The former's objective is to recognize when the person is still, in which case their position and distance from the sensors are reported. The latter's objective is to create a beam directed toward the person's detected chest and measure the reflected signal, for heartbeat detection, HR, or respiration rate (RR) estimation. Our experiments demonstrate that determining the subject's location and distance and identifying their chest position using LiDAR, then collecting the data accordingly using the Doppler radar leads to better HR detection and R-R intervals (RRIs) estimation. For five different scenarios, the RRI estimation error reaches values between 112 and 231 ms.

We consider a service cost minimization problem for resource-constrained small-cell networks with caching, where the challenge mainly stems from (i) the insufficient backhaul capacity and limited network bandwidth and (ii) the limited storing capacity of small-cell base stations (SBSs). Besides, the optimization problem is NP-hard since both the users' mobility patterns and content preferences are unknown. In this paper, we develop a novel mobility-Aware joint routing and caching strategy to address the challenges. The designed framework divides the entire geographical area into small sections containing one SBS and several mobile users (MUs). Based on the concept of one-stop-shop (OSS), we propose a federated routing and popularity learning (FRPL) approach in which the SBSs cooperatively learn the routing and preference of their respective MUs and make a caching decision. The FRPL method completes multiple tasks in one shot, thus reducing the average processing time per global aggregation of learning. By exploiting the outcomes of FRPL together with the estimated service edge of SBSs, the proposed cache placement solution greedily approximates the minimizer of the challenging service cost optimization problem. Theoretical and numerical analyses show the effectiveness of our proposed approaches.

In this paper, we optimize the secrecy capacity of the legitimate user under resource allocation and security constraints for a multi-antenna environment for the simultaneous transmission of wireless information and power in a dynamic downlink scenario. We study the relationship between secrecy capacity and harvested energy in a power-splitting configuration for a nonlinear energy-harvesting model under co-located conditions. The capacity maximization problem is formulated for the vehicle-to-vehicle communication scenario. The formulated problem is non-convex NP-hard, so we reformulate it into a convex form using a divide-and-conquer approach. We obtain the optimal transmit power matrix and power-splitting ratio values that guarantee positive values of the secrecy capacity. We analyze different vehicle-to-vehicle communication settings to validate the differentiation of the proposed algorithm in maintaining both reliability and security. We also substantiate the effectiveness of the proposed approach by analyzing the trade-offs between secrecy capacity and harvested energy.

Specific emitter identification (SEI) plays an important role in secure Industrial Internet of Things (IIoT). In recent years, many SEI methods based on machine learning (ML) and deep learning (DL) have been proposed due to their great performance. However, DL-based SEI methods are accompanied by huge computation overhead, which is not suitable for IIoT applications. In addition, the existing ML-based SEI methods rely on feature extraction and a heavy and redundant classifier, which do not ensure optimal feature combination and efficient computation. To solve the above problem, we propose an improved DL-based SEI method using a signal feature embedded knowledge graph (KG) composed of universal features. To the best of our knowledge, this is the first attempt to apply KG for SEI technology. Specifically, we explore an adaptive feature combination (AFC) strategy through the attention mechanism to realize an efficient SEI classifier. The simulation results show that the proposed KG-AFC algorithm outperforms existing SEI methods in identification performance and computation overhead. At the same time, under the optimal compression rate, the average accuracy of the proposed SEI algorithm is higher than 99.2% and can effectively reduce complexity. The code and the data set can be downloaded from https://github.com/Lollipophua/KG-AFC.

Alzheimer’s disease (AD) is a type of dementia that is more likely to occur as people age. It currently has no known cure. As the world’s population is aging quickly, early screening for AD has become increasingly important. Traditional screening methods such as brain scans or psychiatric tests are stressful and costly. The patients are likely to feel reluctant to such screenings and fail to receive timely intervention. While researchers have been exploring the use of language in dementia detection, less attention has been given to face-related features. The paper focuses on investigating how face-related features can aid in detecting dementia by exploring the PROMPT dataset that contains video data collected from patients with dementia during interviews. In this work, we extracted three types of features from the videos, including face mesh, Histogram of Oriented Gradients (HOG) features, and Action Units (AU). We trained traditional machine learning models and deep learning models on the extracted features and investigated their effectiveness in dementia detection. Our experiments show that the use of HOG features achieved the highest accuracy of 79% in dementia detection, followed by AU features with 71% accuracy, and face mesh features with 66% accuracy. Our results show that face-related features have the potential to be a crucial indicator in automated computational dementia detection.

In this article, we develop a deep learning (DL)-guided hybrid beam and power allocation approach for multiuser millimeter-wave (mmWave) networks, which facilitates swift beamforming at the base station (BS). The following persisting challenges motivated our research: (i) User and vehicular mobility, as well as redundant beam-reselections in mmWave networks, degrade the efficiency; (ii) Due to the large beamforming dimension at the BS, the beamforming weights predicted by the cutting-edge DL-based methods often do not suit the channel distributions; (iii) Co-located user devices may cause a severe beam conflict, thus deteriorating system performance. To address the aforementioned challenges, we exploit the synergy of supervised learning and super-resolution technology to enable low-overhead beam- and power allocation. In the first step, we propose a method for beam-quality prediction. It is based on deep learning and explores the relationship between high- and low-resolution beam images (energy). Afterward, we develop a DL-based allocation approach, which enables high-accuracy beam and power allocation with only a portion of the available time-sequential low-resolution images. Theoretical and numerical results verify the effectiveness of our proposed framework.