The Image Coding for Machine (ICM) approach effectively mitigates challenges posed by data transmission of machine vision tasks on edge devices with bandwidth-limited networks. However, existing ICM solutions consume tremendous storage in complex and multi-task scenarios, which is impractical for edge devices in the field of Internet of Things (IoT). To address this issue, we introduce VersaCodec, a versatile codec with variable compression rates, built upon the Swin Transformer architecture. VersaCodec utilizes a quality-map-based encoding-decoding scheme and an adapted prompt technique to maximize the preservation of semantic information for machine perception tasks via exploring the spatial importance. In VersaCodec, a quality-map generator is designed to extract the spatial semantic information from conditions and tasks. VersaCodec enables seamless task switching and incorporates a conditional input mechanism, allowing flexible adjustment of the compression rate. This makes VersaCodec adaptive to varying environmental conditions and thus suitable for specific application scenarios of IoT. Experiments demonstrate that our VersaCodec outperforms TransTIC tailored for specific compression rates across multiple downstream tasks.

In the field of time series prediction, runoff prediction is a critical subfield. Accurate runoff forecasting is essential for optimizing water resource allocation and mitigating flood risks. Conventional methods often fail to capture the complex patterns inherent in runoff formation due to its intricate mechanisms. While convolutional neural networks (CNNs) have shown promise in runoff prediction due to their robust feature extraction capabilities, existing studies predominantly focus on increasing network complexity, with limited emphasis on optimizing convolution operations. This study proposes a lightweight attention-based modern convolutional neural network (LMCNN) that enhances the efficiency of convolutional layers. LMCNN introduces a novel convolutional strategy that employs distinct convolutional layers to capture temporal features within channels, intra-variable inter-channel features, and inter-variable inter-channel features, significantly improving feature extraction while reducing computational complexity compared to traditional CNNs. Specifically, the model integrates meteorological and runoff data using a channel-independent strategy, followed by depthwise convolution (DWConv) to extract temporal features within individual channels. Two layers of pointwise and SE-enhanced convolution (PSConv), incorporating pointwise convolution for cross-channel fusion and improved Squeeze-and-Excitation Networks (improved SENet) for dynamic feature selection, are then applied to extract intra- and inter-variable channel features. The data is subsequently flattened and processed through a linear layer to produce predictions. The model’s performance is validated using data from three hydrological stations in the upper, middle, and lower reaches of the Yellow River Basin, China. A comparative analysis with other advanced prediction models validates the effectiveness of the proposed model.

Parallel decoding of backscatter improves communication throughput by enabling concurrent transmission of backscatter tags. In practical applications of parallel decoding, it is extremely difficult to distinguish collided signals in superclusters where multiple signal clusters overlap. Existing methods are usually effective for superclusters with uniformly distributed signals. Nevertheless, there are many more scenarios in which signals in superclusters tend to gather unevenly, and existing methods cannot work. Such uneven clustering of signals occurs due to the following two possible causes: (1) signal-strengthdifferences (SSDs) among tags; or (2) cluster drifting (CD) driven by interferences from other objects within communication environments. This paper proposes a novel scheme called SalixLeaf, which aims to identify the main veins of signal clusters to address this problem of superclusters with unevenly distributed signals. Salix-Leaf identifies the main vein of each signal cluster for fine-grained clustering so that the direction of the main veins can be used to verify the accuracy of clustering. In addition, SalixLeaf employs a supercluster decomposer that divides signals into different segments for clustering analysis, enhancing robustness and practicability. Experimental results show that Salix-Leaf achieves a 1.2-fold increase in throughput and a 25% reduction in bit error rate (BER) compared to the state-of-the-art.

Backscatter sensing has emerged as a significant technology within the Internet of Things (IoT), prompting extensive research interest. This paper presents LoMu, the first long-range multi-target backscatter sensing system designed for low-cost tags operating under ambient LoRa. LoMuintroduces an orthogonal sensing model that processes backscatter signals from multiple tags to extract motion information. The design addresses several practical challenges, including near-far interference among multiple tags, phase offsets from unsynchronized transceivers, and phase errors due to frequency drift in low-cost tags. To overcome these issues, we propose a conjugate-based energy concentration method to extract high-quality signals and a Hamming-window-based method to mitigate the near-far problem. Additionally, we exploit the relationship between excitation and backscatter signals to synchronize the transmitter (TX) and receiver (RX) and combine double sidebands of backscatter signals to eliminate tag frequency drift. Furthermore, a novel joint estimation algorithm is introduced to exploit both amplitude and phase information in target signals, enhancing frequency sensing results and robustness. Our implementation and extensive experiments demonstrate that LoMucan accurately sense up to 35 tags simultaneously and achieve an average frequency sensing error of 0.5% at a range of 400 meters, which is$4\times$the range of the state-of-the-art.

Graph neural networks provide a powerful toolkit for embedding real-world graphs into low-dimensional spaces according to specific tasks. Up to now, there have been several surveys on this topic. However, they usually lay emphasis on different angles so that the readers cannot see a panorama of the graph neural networks. This survey aims to overcome this limitation and provide a systematic and comprehensive review on the graph neural networks. First of all, we provide a novel taxonomy for the graph neural networks, and then refer to up to 327 relevant literatures to show the panorama of the graph neural networks. All of them are classified into the corresponding categories. In order to drive the graph neural networks into a new stage, we summarize four future research directions so as to overcome the challenges faced. It is expected that more and more scholars can understand and exploit the graph neural networks and use them in their research community.

Abstract In the Global Navigation Satellite System, the traditional adaptive equalization algorithm re‐captures and tracks poor performance under the multipath channel for multipath errors caused by occlusion or reflection of buildings. A channel compensation multipath mitigation technique based on Kalman‐Based Least Mean Square based (KBLMS) on Kalman estimation. First, in the tracking loop of the receiver, a KBLMS‐based delay estimation module is designed to compensate for multipath distortion on the received signal. The coefficients of the filter are adaptively adjusted using the feedback signal. Second, a line of sight best estimate block is designed to generate a control error signal in the feedback loop for adaptively updating the filter coefficients. Finally, the performance of the proposed algorithm is verified by analyzing the performance of KBLMS, Recursive Least Squares (RLSs) and Least Mean Squares (LMS) algorithms through measured data and experimental simulation. The results show that KBLMS can converge quickly in multipath channels compared with RLS and LMS, and the code tracking error and carrier tracking error are reduced by 0.1chip, 0.2 cycle and final residual error is reduced by 0.015 compared with the RLS algorithm, which is 0.035 lower than the LMS algorithm, which shows that the KBLMS algorithm can make more accurate estimation.

Neuromuscular system disease can cause severe paralysis and even speech impairments. To communicate with others, the patients can use breathing to encode information. However, traditional solutions only allow patients to output several predefined sentences. Other solutions require patients to interact with the Windows system, which are not suitable for blind paralyzed patients. To address these problems, we present a text input system based on commercial-off-the-shelf RFID devices, named Respiration-Type (ReType). A passive lightweight RFID tag is attached to the user's chest area over the clothes. By interrogating the tag continuously, an RFID reader can output phase data stream, which will be modulated by the chest fluctuation caused by respiration. Based on this phenomenon, we let normal breath and deep breath represent binary 0 and 1, respectively. To extract the feature information related to breath in real time from the noisy phase data stream, we develop a novel algorithm that is robust to noise and baseline drift. To ensure the robustness of decoding, a breath-control model is established for decoding normal breath and deep breath. Comprehensive experiments show that ReType can achieve an average recognition accuracy of 92.22% and is robust against environment changes and user diversity.

In recent years, human motion recognition, as an important application of the intelligent perception of the Internet of Things, has received extensive attention. Many applications benefit from motion recognition, such as motion monitoring, elderly fall detection, and somatosensory games. Several existing RF‐based motion recognition systems are susceptible to multipath effects in complex environments, resulting in lower recognition accuracy and difficulty in extending to other scenarios. To address this challenge, we propose RF‐Motion, a device‐free commercial off‐the‐shelf (COTS) RFID‐based human motion recognition system that can detect human motion in complex multipath environments such as indoor environments. And when the environment changes, RF‐Motion still has high recognition accuracy, even without retraining. In addition, we use data slicing to solve the problem of discontinuity in the time domain of RFID communication and then use the synthetic aperture (SAR) algorithm to obtain the fingerprint feature matrix corresponding to each motion. Finally, the dynamic time warping (DTW) algorithm is used to match the prior motion fingerprint database to complete the motion recognition. Experiments show that RF‐Motion can achieve up to 90% accuracy for human motion recognition in an indoor environment, and when the environment changes, it can still reach a minimum accuracy of 87%.

Abstract Ambient backscatter communication (AmBC) has great application prospects in the green Internet of Things due to its shared nature of energy and spectrum. In this paper, we have done research on the signal detection problem in the AmBC system and proposed to use the k-means clustering analysis algorithm to detect the received backscattered signal. At the same time, in order to obtain a more obvious clustering center, this paper introduces a multi-antenna receiver channel detection mechanism. The proposed method can compare and analyze the energy received by the receiver antennas in an extremely short time slots. The channel state information between the reader and the backscatter tag is obtained, and the optimal communication channel is selected, thereby effectively improving the clustering effect. Finally, a large number of experimental simulations are provided to compare and analyze the corresponding BER performance to confirm our theoretical research.

Summary Recently, in a variety of Internet of Things (IoT) application scenarios, fingerprint image as a common biological information has been widely collected. For example, in safe city projects, civil fingerprints representing unique identification are vastly assembled, stored, and transmitted wirelessly. Under such situations, data privacy protection becomes increasingly valued. In our study, this fundamental problem was centered. In our work, the method to remain secure in such an universal cloud based IoT system and the approach to hide important information in biologically collected data such are fully discussed. A new lightweight algorithm in protecting the privacy and integrity of data in a cloud based IoT system is badly needed. We believe that it is a way to hide the core information in the surrounding environment just like “Chameleon” to protect the data security. Thus, in this paper, we propose “CHAMELEON” by using least significant bits (LSB) to hide important information in fingerprint images and chaotic sequence encryption (CSE) to securely transmit useful data. The experimental results of this algorithm show that Chameleon can resist static attacks and protect the privacy of users regarding to the information security and integrity.

Summary In complex urban environment, GPS signals are susceptible to occlusion and interference; the positioning accuracy will be reduced or even impossible to locate. Today, Wireless Local Area Network (WLAN) localization technology can effectively supplement the blind spot of GPS signals. However, due to the instability of received signal strength (RSS), the positioning accuracy of WLAN is seriously affected. To solve the aforementioned problems, a GPS/WLAN integrated location algorithm based on particle filter is proposed. In our method, the federated Kalman filter is used to fuse the location information of GPS and WLAN, and then the important sampling function of the particle filter is generated by the fusion result. The updating of the particle weights is accomplished by the results of federated Kalman filter and map information. With the complementary characteristics of GPS and WLAN positioning technology, the experimental results show that our algorithm can effectively improves the coverage rate of GPS and the positioning accuracy of pedestrians in complex urban environment. Compared with GPS and WLAN single system and traditional fusion algorithm, the positioning accuracy of the proposed algorithm is significantly improved.

In urban canyons, multipath (MP) signals have some problems such as short time delay, high dynamic, and uncontrollable. To solve this problem and mitigate the interference of MP effect on positioning, the authors proposed a new method which uses gated recurrent unit (GRU) neural networks (NNs) aided by compressed sensing algorithm to estimate parameters of MP signal accurately. They showed how to build the GRU – deformation of recurrent NNs (RNNs) – in a dynamic environment, and estimated the MP signals parameters accurately. Also, considering the special cases in a dynamic environment, they used the adaptive wavelet filter to optimise the NNs, compensated pseudo‐distance, and improved the accuracy of positioning. Like other NNs, the learning process of this machine learning method in the dynamic environment can be summarised as obtaining data from different environmental conditions. For this, they collected signal data from different speeds to train GRU–RNN model, which correctly estimated the MP signal's parameter value in a different environment and different speeds. By experimental verification, when the speed is slower than 20 km/h, GRU–RNN can reduce code tracking error to 0.08 chips, and increased the positioning accuracy by about 13%.

Compared with traditional radio frequency identification (RFID), computational RFID (CRFID) tag has more powerful computing capability, but it shows poor performance when it follows EPC Class‐1 Generation‐2 protocol especially transmitting large amounts of data. In fact, the operation of the CRFID tag entirely depends on the state of energy. For this point, this study proposes an optimised protocol called LILAC. LILAC allows tags to select the communication time slot according to their current voltage value measured by using an analogue‐to‐digital converter rather than randomly selecting, the authors proposed a more reasonable time slot mapping algorithm for LILAC that increases the success rate of tag responding. In addition, they design a data transmission format that can be retransmitted to improve the uplink throughput. Finally, they implemented LILAC on a CRFID platform and practically measured several parameters to compare with the existing well‐known protocol. The results of experiments show that LILAC increases the maximum communication distance by more than half in the access phase and doubles the goodput of the backscatter link in both the inventory and access phase.

Summary The main challenge of access control system is how to allow authorized users to access quickly and accurately. In reality, most access control system cannot run smoothly with the adverse impact of a security problem, it is because the access control card is easy to duplicate. To improve system security, recent efforts introduced biological technologies, such as fingerprint identification, facial recognition, and iris recognition. However, these methods are costly for installation and later maintenance. In this paper, we introduce a cloud access control system that employs the ability of sensing acceleration of Wireless Identification and Sensing Platform (WISP) tags combined with customized motions. Users are allowed to pass the access control system only if they operate user‐defined motions correctly. Authorized users can define the authentication motions by themselves, which not only facilitates the daily use but also improves the security of the cloud access control system. We conduct intensive experimental measurements. The experimental results verify the effectiveness of our proposed system.

In complex landform the satellite signals are blocked so seriously that BeiDou satellitenavigation system (BDS) cannot achieve real‐time high‐precision dynamicalpositioning, which seriously restrict its widespread in precise mapping anddisaster relief. This study proposes global positioning system (GPS)/BDS vectortracking loop (VTL) assisted by the neural network (NN) to improve theperformance of receivers in complex environments. In GPS/BDS VTL, pre‐filter andnavigation extended Kalman filter are both used, and discriminators are replacedby pre‐filter structure; then, the output information of pre‐filter is used asinput of the navigation filter, at the same time the pseudorange rate ofpre‐filter is updated by pseudorange rate of navigation filter; finally, the NNis used in the GPS/BDS vector tracking structure, in the training stage, theoutput from navigation filter is adopted as the input of the NN. This method canprevent the error growth due to weak signal from deteriorating the entiretracking loop performance. Moreover, the NN is employed to provide a betterprediction of residuals such as the Doppler frequency and code phase. Theresults show that the proposed adaptive GPS/BDS VTL shows a better trackingperformance compared with the conventional VTL.

Summary In order to improve the positioning and navigation performance of Global Navigation Satellite System (GNSS) receivers, a novel method to extract auxiliary information for GNSS receiver is proposed in this paper, which obtains the area GNSS auxiliary information (AGAI) with enough credibility and area attribute. Firstly, a mass of historical GNSS intermediate frequency data is divided into blocks to be acquired and tracked parallel getting massive pseudorange and navigation message (MPD). Then, the massive MPD are weighted and fused parallel by an appropriate weight matrix, which is determined by a priori weighting based on altitude angle and posterior weighting based on M‐residuals variable components estimation. Lastly, the fused information is corresponded to the corresponding location coordinate, completing the parallel extraction of AGAI. The method implemented by parallel programming models MapReduce of Hadoop to guarantee a high efficiency. Experimental results show that the positioning and velocity error of GNSS receivers are reduced by 18.24% and 20.48% using AGAI instead of traditional auxiliary information, and the execution time of the method using MapReduce is reduced by 46.72%, so the proposed method is reliable and effective. Copyright © 2016 John Wiley & Sons, Ltd.

We propose and implement a novel positioning system, WalkieLokie, which directly calculates the relative position from a smart device to a target. The requirement of the target is simple: it is attached with a “dummy” acoustic speaker, which does not have any other rich capabilities, such as audio recording, communication, or computation. Hence, the proliferation of smart devices, together with the cheap accessory (e.g., dummy speaker) embedded in daily used items (e.g., smart clothes), paves the way for WalkieLokie applications. WalkieLokie leverages the walking motion for locating an acoustic speaker. The key insight is that the distance between the user and the speaker varies in real time when the user walks, and the pattern of the variance implies the relative position. We design a novel algorithm to estimate the position and signal processing methods to support accurate positioning. The experiment results show that the mean errors of ranging and direction estimation are 0.63 m and 2.46°, respectively. Extensive experiments conducted in noisy environments validate the robustness of WalkieLokie.

Radio Frequency Identification (RFID) is an emerging technology for electronic labeling of objects for the purpose of automatically identifying, categorizing, locating, and tracking the objects. But in their current form RFID systems are susceptible to cloning attacks that seriously threaten RFID applications but are hard to prevent. Existing protocols aimed at detecting whether there are cloning attacks in single-reader RFID systems. In this paper, we investigate the cloning attacks identification in the multireader scenario and first propose a time-efficient protocol, called the time-efficient Cloning Attacks Identification Protocol (CAIP) to identify all cloned tags in multireaders RFID systems. We evaluate the performance of CAIP through extensive simulations. The results show that CAIP can identify all the cloned tags in large-scale RFID systems fairly fast with required accuracy.

In this paper, we propose and implement WalkieLokie, a novel acoustic-based relative positioning system. WalkieLokie facilitates a multitude of Augmented Reality (AR) applications: users with smart devices can passively acquire surrounding information in real time, similar to the commercial AR system Wikitude; the surrounding presenters, who want to share information or introduce themselves, can actively launch the function on demand. The key rational of WalkieLokie is that a user can perceive a series of spatial-related acoustic signals emitted from a presenter, which depicts the relation position between the user and the presenter. The proliferation of smart devices, together with the cheap accessory (e.g., dummy speaker) embedded in daily used items (e.g., smart clothes), paves the way for WalkieLokie applications. We design a novel algorithm to estimate the position and signal processing methods to support accurate positioning. The experiment results show that the mean error of ranging and direction estimation is 0.63m and 2.46 degrees respectively. Extensive experiments conducted in noisy environments validate the robustness of WalkieLokie.

Structural Health Monitoring (SHM) plays an important rule to improve life safety and achieve economic benefits. The use of conventional wireless sensor networks for SHM incurs a huge maintenance cost for frequent battery changes. In this paper, we adopt passive RFID tags, which are power-free devices, to build fine-grained SHM systems. To get timely report of the monitoring data, we propose a power-free monitoring scheme with compressive sensing (PFM-CS) to collect aggregated SHM events. It takes the advantage of compressive sensing to reconstruct original events with high accuracy. We evaluate the performance of PFM-CS using the real world data trace collected from a nineteen-storey building in Singapore. The results validate the effectiveness and efficiency of PFM-CS.

The Chinese Bei Dou Navigation Satellite System Signal (BDS) [1] and the modernized GNSS take along a feature of the new modulations is the presence of two channels which is consist of the data and pilot components. The specific structure promoted by secondary codes and the presence of data/pilot channels. However, thanks to the new tiered code, it limits the coherent integration time for acquisition, in order to extend the coherent integration time, relative signs between subsequent portions of the incoming signal have to be estimated and used to remove the effect of sign transitions. To solve the problem the traditional algorithm like tree-based acquire technique suffers a high computational complexity. The paper demonstrates a new algorithm to acquire the BDS signal with a low complexity. The proposed algorithms are simulated in terms of the performance of a software defined receiver.

For mobile devices, communication via cellular networks consumes more energy than via WiFi networks, and suffers an expensive limited data plan. On the other hand, as the coverage and the density of WiFI networks are smaller than those of the cellular networks, users cannot purely rely on WiFi to access the Internet. In this work we present a behavior-aware and preference-based approach to prefetch news webpages for the user to visit in the near future, by exploiting the WiFi network connections to reduce the energy and monetary cost. We first design an efficient preference learning algorithm to keep track of the user's changing interests, and then by predicting the appearance and durations of the WiFi network connections, our prefetch approach optimizes when to prefetch to maximize the user experience while lowing the prefetch cost. Our prefetch approach also exploits the idle period of WiFi connections to reduce the tail-energy consumption. We implement our approach in iPhone and our extensive evaluations show that our system achieves about 60% hit ratio, saves about 50% cellular data usage, and reduces the energy cost by 7%.