Search

Traditional methods for transformer fault diagnosis are difficult to fit the complex nonlinear relationship between dissolved gas and fault type accurately, and their diagnosis accuracy is limited. To improve the accuracy of transformer fault diagnosis, a method of transformer fault diagnosis based on capsule network (CapsNet) is proposed. The convolutional layers with strong feature extraction ability are used to map the dissolved gas data into the feature space to realize the automatic extraction of key features. Furthermore, the mapping relationship between feature and fault type is established by primary capsule layers and digital capsule layers. Dynamic routing algorithm and back propagation algorithm are used to complete the training process of CapsNet. The simulation results show that the fault diagnosis performance of CapsNet is better than that of the traditional methods such as convolutional neural network, multi-layer perceptron, support vector machine, extreme gradient boosting tree, and light gradient lifting machine in different input characteristics and data scales, which can effectively meet the needs of diagnosis accuracy and provide guidance for the operation pattern of relay. The research can provide references for the fault diagnosis of the transformer., QC 20220118

Hand gesture recognition using surface electromyography (sEMG) has been one of the most efficient motion analysis techniques in human–computer interaction in the last few decades. In particular, multichannel sEMG techniques have achieved stable performance in hand gesture recognition. However, the general solution of collecting and labeling large data manually leads to time-consuming implementation. A novel learning method is therefore needed to facilitate efficient data collection and preprocessing. In this paper, a novel autonomous learning framework is proposed to integrate the benefits of both depth vision and EMG signals, which automatically label the class of collected EMG data using depth information. It then utilizes a multiple layer neural network (MNN) classifier to achieve real-time recognition of the hand gestures using only the sEMG. The overall framework is demonstrated in an augmented reality application by the recognition of 10 hand gestures using the Myo armband and an HTC VIVE PRO. The results show prominent performance by introducing depth information for real-time data labeling., QC 20211130

Large service providers use load balancers to dispatch millions of incoming connections per second towards thousands of servers. There are two basic yet critical requirements for a load balancer: uniform load distribution of the incoming connections across the servers, which requires to support advanced load balancing mechanisms, and per-connection-consistency (PCC), i.e, the ability to map packets belonging to the same connection to the same server even in the presence of changes in the number of active servers and load balancers. Yet, simultaneously meeting these requirements has been an elusive goal. Today’s load balancers minimize PCC violations at the price of non-uniform load distribution. This paper presents Cheetah, a load balancer that supports advanced load balancing mechanisms and PCC while being scalable, memory efficient, fast at processing packets, and offers comparable resilience to clogging attacks as with today’s load balancers. The Cheetah LB design guarantees PCC for any realizable server selection load balancing mechanism and can be deployed in both stateless and stateful manners, depending on operational needs. We implemented Cheetah on both a software and a Tofino-based hardware switch. Our evaluation shows that a stateless version of Cheetah guarantees PCC, has negligible packet processing overheads, and can support load balancing mechanisms that reduce the flow completion time by a factor of 2−3×.

Convolutional neural networks are widely used in artificial intelligence and Internet of Things area. As the scale of convolutional neural network expands, more and more processing units are provided for it. The systems are easy prone to error, and any computing problems in any layer of the network will lead to wrong output results. Traditional multimode redundancy methods make the systems more complex, and increase power consumption. This paper proposes an autonomous error-tolerant architecture for convolutional neural networks. Taking the LeNet-5 as an example, the network layers of CNN are mapped on the AET architecture, an error-tolerant synapse is designed to discover the errors, an active evolution scheme is designed to handle unrecoverable errors and implement network reconfiguration. This design is implemented on FPGA, and the experimental results show that this architecture can realize effective error tolerance for convolutional neural network and has fast error recovery ability under the premise of ensuring the same recognition accuracy., QC 20210915

Deep Learning (DL) has become popular due to its ease of use and accuracy, with Transfer Learning (TL) effectively reducing the number of images needed to solve environmental problems. However, this approach has some limitations which we set out to explore: Our goal is to detect the presence of an invasive blueberry species in aerial images of wetlands. This is a key problem in ecosystem protection which is also challenging in terms of DL due to the severe imbalance present in the data. Results for the ResNet50 network show a high classification accuracy while largely ignoring the blueberry class, rendering these results of limited practical interest to detect that specific class. Moreover, by using loss function weighting and data augmentation results more akin to our practical application, our goals can be obtained. Our experiments regarding TL show that ImageNet weights do not produce satisfactory results when only the final layer of the network is trained. Furthermore, only minor gains are obtained compared with random weights when the whole network is retrained. Finally, in a study of state-of-the-art DL architectures best results were obtained by the ResNeXt architecture with 93.75 True Positive Rate and 98.11 accuracy for the Blueberry class with ResNet50, Densenet, and wideResNet obtaining close results. © 2020 by the authors. Licensee MDPI, Basel, Switzerland.

This paper presents a methodology of design and implementation details of a solution based on Internet of things (IoT) applied to home automation. The solution is implemented using the system on a chip ESP8266. The proposed architecture allows the interaction between sensors and actuators (client nodes) linked through a home wireless network. The protocol was implemented in the application layer of the TCP/IP model, while subtle changes on the remaining layers of the TCP/IP stack are applied, as follows: 1) The IEEE 802.11 standard is used in the physical layer to define basic aspects such as transmitter power and receiver sensitivity; 2) CSMA/CA was used as a media access method, allowing multiple stations to use the same transmission medium; 3) in the network layer, the interaction between the nodes is configured in mesh mode, where each node uses a 32-bit identifier based on its MAC address; 4) the transport layer is configured using TCP, to guarantee reliability in data delivery; and, 5) the MQTT protocol is used in the application layer. To evaluate the performance of the proposed protocol, there were manufactured prototypes of the client nodes. An application for Android-based mobile devices was also developed, so that the user can control and manage the network from their own device.

In this paper, we investigate the secrecy performance of a Cooperative Cognitive Radio Network (CCRN) in the presence of an eavesdropper (EAV). The secondary users (SUs) are subject to three constraints which include peak transmit power level and interference limitation with respect to the primary user (PU) as well as secrecy outage constraints due to the EAV. Secrecy outage is achieved when the EAV cannot decode the targeted signal, but communications in the secondary network is still possible (non-zero capacity exists). Approximation expressions of the secrecy outage probability and the probability of non-zero secrecy capacity are derived to evaluate the secrecy performance. Monte Carlo simulations are provided to examine the accuracy of the derived approximation expressions. Based on this, power allocation policies for the SUs are derived, satisfying all the constraints while maximizing the secrecy performance as well as the quality of service performance of the secondary network. It can be concluded that with knowledge of the channel state information (CSI) of the EAV it is possible to calculate the optimal value for the secrecy outage threshold of the secondary user (SU) which in turn allows maximizing the secrecy performance. Most interestingly, our numerical results illustrate that the secrecy performance of the system is much improved when the parameters obtained using the CSI of the EAV are calculated optimally. Thence, the system can adjust the power allocation so that no eavesdropping occurs even without reducing quality of service (QoS) performance compared to a network without any EAV.

This work presents InfiniWolf, a novel multi-sensor smartwatch that can achieve self-sustainability exploiting thermal and solar energy harvesting, performing computationally high demanding tasks. The smartwatch embeds both a System-on-Chip (SoC) with an ARM Cortex-M processor and Bluetooth Low Energy (BLE) and Mr. Wolf, an open-hardware RISC-V based parallel ultra-low-power processor that boosts the processing capabilities on board by more than one order of magnitude, while also increasing energy efficiency. We demonstrate its functionality based on a sample application scenario performing stress detection with multi-layer artificial neural networks on a wearable multi-sensor bracelet. Experimental results show the benefits in terms of energy efficiency and latency of Mr. Wolf over an ARM Cortex-M4F micro-controllers and the possibility, under specific assumptions, to be self-sustainable using thermal and solar energy harvesting while performing up to 24 stress classifications per minute in indoor conditions. © 2020 EDAA.

In this paper, we aim to investigate the superiority and practicability of many-parameter wavelet and Golay transforms (MPWT and MPGT) from the physical layer security (PHY-LS) perspective. We propose novel Intelligent OFDM-telecommunication system (Intelligent-OFDM-TCS), based on many-parameter transforms (MPTs). New system uses inverse MPT for modulation at the transmitter and direct MPT for demodulation at the receiver. The purpose of employing the MPTs is to improve the PHY-LS of wireless transmissions against to the wide-band anti-jamming communication. Each MPT depends on finite set of independent Jacobi parameters (angles), which could be changed independently one of another. When parameters are changed, multi-parametric transform is changed too taking form of a set known (and unknown) orthogonal (or unitary) wavelet transforms. We implement the following performances as bit error rate (BER), symbol error rate (SER), peak to average power ratio (PAPR), the Shannon-Wyner secrecy capacity (SWSC) for novel Intelligent-MPWT-OFDM-TCS. Previous research has shown that the conventional OFDM TCS based on discrete Fourier transform (DFT) has unsatisfactory characteristics in BER, PARP, SWSC and in anti-eavesdropping communications. We study Intelligent-MPT-OFDM-TCS to find out optimal values of angle parameters of MPT optimized BER, PAPR, SWSC, anti-eavesdropping effects. Simulation results show that the proposed Intelligent OFDM-TCS have better performances than the conventional OFDM system based on DFT against eavesdropping. © 2019 IOP Publishing Ltd. All rights reserved.

In this paper, we aim to investigate the superiority and practicability of many-parameter transforms (MPTs) from the physical layer security (PHY-LS) perspective. We propose novel Intelligent OFDM-telecommunication systems based on complex and quaternion MPTs. The new systems use inverse MPT (IMPT) for modulation at the transmitter and MPT for demodulation at the receiver. The purpose of employing the MPT is to improve: 1) the PHY-LS of wireless transmissions against to the wide-band anti-jamming and anti-eavesdropping communication; 2) the bit error rate (BER) performance with respect to the conventional OFDM-TCS; 3) the peak to average power ratio (PAPR). Each MPT depends on finite set of independent parameters (angles). When parameters are changed, many-parametric transform is also changed taking form of a set known (and unknown) orthogonal (or unitary) transforms. For this reason, the concrete values of parameters are specific "key" for entry into OFDM-TCS. Vector of parameters belong to multi-dimension torus space. Scanning of this space for find out the "key" (the concrete values of parameters) is hard problem. MPT has the form of the product of the Jacobi rotation matrixes and it describes a fast algorithm for MPT. The main advantage of using MPT in OFDM TCS is that it is a very flexible anti-eavesdropping and anti-jamming Intelligent OFDM TCS. To the best of our knowledge, this is the first work that utilizes the MPT theory to facilitate the PHY-LS through parameterization of unitary transforms. © 2019 IOP Publishing Ltd. All rights reserved.

In this paper, we aim to investigate the superiority and practicability of many-parameter Fourier transforms (MPFT) from the physical layer security (PHY-LS) perspective. We propose novel Intelligent OFDM-telecommunication system (Intelligent-OFDM-TCS), based on MPFT. New system uses inverse MPFT for modulation at the transmitter and direct MPFT for demodulation at the receiver. The purpose of employing the MPFTs is to improve the PHY-LS of wireless transmissions against to the wide-band anti-jamming communication. Each MPFT depends on finite set of independent parameters (angles), which could be changed independently one from another. When parameters are changed, multi-parametric transform is also changed taking form of a set known (and unknown) orthogonal (or unitary) transforms. We implement the following performances as bit error rate (BER), symbol error rate (SER), the Shannon-Wyner secrecy capacity (SWSC) for novel Intelligent-MPWT-OFDM-TCS. Previous research has shown that the conventional OFDM TCS based on discrete Fourier transform (DFT) has unsatisfactory characteristics in BER, SWSC and in anti-eavesdropping communications. We study Intelligent-MPWT-OFDM-TCS to find out optimal values of angle parameters of MPFT optimized BER, SWSC, anti-eavesdropping effects. Simulation results show that the proposed Intelligent OFDM-TCS have better performances than the conventional OFDM system based on DFT against eavesdropping. © 2019 IOP Publishing Ltd. All rights reserved.

The agricultural and farming industries have been widely influenced by the disruption of the Internet of Things. The impact of the IoT is more limited in countries with less penetration of mobile internet such as sub-Saharan countries, where agriculture commonly accounts for 10 to 50% of their GPD. The boom of low-power wide-area networks (LPWAN) in the last decade, with technologies such as LoRa or NB-IoT, has mitigated this providing a relatively cheap infrastructure that enables low-power and long-range transmissions. Nonetheless, the benefits that LPWAN technologies enable have the disadvantage of low-bandwidth transmissions. Therefore, the integration of Edge and Fog computing, moving data analytics and compression near end devices, is key in order to extend functionality. By integrating artificial intelligence at the local network layer, or Edge AI, we present a system architecture and implementation that expands the possibilities of smart agriculture and farming applications with Edge and Fog computing and LPWAN technology for large area coverage. We propose and implement a system consisting on a sensor node, an Edge gateway, LoRa repeaters, Fog gateway, cloud servers and end-user terminal application. At the Edge layer, we propose the implementation of a CNN-based image compression method in order to send in a single message information about hundreds or thousands of sensor nodes within the gateway's range. We use advanced compression techniques to reduce the size of data up to 67% with a decompression error below 5%, within a novel scheme for IoT data., Part of ISBN 9781728132891QC 20210920

In robot control with physical interaction, like robot-assisted surgery and bilateral teleoperation, the availability of reliable interaction force information has proved to be capable of increasing the control precision and of dealing with the surrounding complex environments. Usually, force sensors are mounted between the end effector of the robot manipulator and the tool for measuring the interaction forces on the tooltip. In this case, the force acquired from the force sensor includes not only the interaction force but also the gravity force of the tool. Hence the tool dynamic identification is required for accurate dynamic simulation and model-based control. Although model-based techniques have already been widely used in traditional robotic arms control, their accuracy is limited due to the lack of specific dynamic models. This work proposes a model-free technique for dynamic identification using multi-layer neural networks (MNN). It utilizes two types of MNN architectures based on both feed-forward networks (FF-MNN) and cascade-forward networks (CF-MNN) to model the tool dynamics. Compared with the model-based technique, i.e., curve fitting (CF), the accuracy of the tool identification is improved. After the identification and calibration, a further demonstration of bilateral teleoperation is presented using a serial robot (LWR4+, KUKA, Germany) and a haptic manipulator (SIGMA 7, Force Dimension, Switzerland). Results demonstrate the promising performance of the model-free tool identification technique using MNN, improving the results provided by model-based methods., QC 20220615

Physical layer authentication based on channel state information is an effective solution to preventing spoofing attacks in wireless communications by comparing the channel impulse responses. Existing theoretical analyses and experiments have proved the feasibility and efficiency in labs or offices. However, the environment of industrial wireless communication is significantly different. This paper applies physical layer authentication based on channel state information to measurements from four different industrial wireless communication scenarios, including indoor, outdoor, moving, and stationary scenarios. The analysis of the results allows to derive meaningful insights on the applicability of such a method to industrial wireless communications., QC 20200624

Given the pivotal role of optical networks in supporting critical societal services, their robustness to deliberate attacks targeting disruption at the physical layer requires advanced approaches for security assurance, diagnostics and response. This paper analyzes the necessary advancements in optical network security needed to achieve secure and self-diagnosable systems., QC 20191104

Exploiting highly directional antenna arrays to compensate for severe propagation loss is one of the defining features in millimeter-wave (mmWave) communications. With efficient beamforming techniques, mmWave transceivers can form steerable narrow beams. Therefore, different from convenient microwave networks, the signal or interference power in mmWave communications is highly directional and closely related to critical parameters such as interference distance and angles of departure/arrival. The high directivity implies that the co-channel interference among simultaneously active mmWave links can be expected to be significantly smaller than that for omnidirectional links and, meanwhile, the security of mmWave communications may also be enhanced. In this context, will the traditional interference mitigation and physical-layer security techniques still be efficient or necessary in mmWave networks? The answer may be negative in certain conditions. However, there is no detailed analysis on the conditions for this issue. In this paper, we jointly analyze the inter-beam interference and secrecy performance of mmWave communications for their close relation to signal-to-interference-plus-noise ratio. We also derive various performance limits (e.g., interference/eavesdropping distance limit, transmission power limit, offset angle limit, and beamwidth limit) of interference and eavesdropping immunity in mmWave networks. The theoretical and numerical results verify our analysis., QC 20220405

We construct a single layer feed forward network and analyze the constructed system using information theoretic tools, such as mutual information and data processing inequality. We derive a threshold on the number of hidden nodes required to achieve a good classification performance. Classification performance is expected to saturate as we increase the number of hidden nodes more than the threshold. The threshold is further verified by experimental studies on benchmark datasets. Index Terms-Neural networks, mutual information, extreme learning machine, invertible function., QC 20190403

In the energy resource-constrained wireless applications, turbo codes are frequently employed to guarantee reliable data communication. To both reduce the power dissipation of the turbo decoder and the probability of data frame retransmission in the physical layer, memory capacity reduced near optimal turbo decoder is of special importance from the perspective of practical implementation. In this regard, a state metrics compressed decoding technique is proposed. By inserting two modules in the conventional turbo decoding architecture, a smaller quantization scheme can be applied to the compressed state metrics. Furthermore, structure of the inserted modules is described in detail. We demonstrate that one or two rounds of compression/decompression are performed in most cases during the iterative decoding process. At the cost of limited dummy decoding complexity, the state metrics cache (SMC) capacity is reduced by 53.75%. Although the proposed technique is a lossy compression strategy, the introduced errors only have tiny negative influence on the decoding performance as compared with the optimal Log-MAP algorithm., Funding details: 2015M570776, China Postdoctoral Science Foundation; Funding details: SWU113044, SWU, Southwest University; Funding details: 61671390, NSFC, National Natural Science Foundation of China; Funding text: This work was supported in part by the National Natural Science Foundation of China under Grant 61671390, in part by the China Postdoctoral Science Foundation under Grant 2015M570776, and in part by the Fundamental Research Fund for Central University (Southwest University) under Grant SWU113044. QC 20181127

Existing wireless communication systems are not able to meet the stringent requirements for critical machine-to-machine communications regarding ultra-reliability and low-latency. Since increasing the communication reliability often comes at the price of increasing the latency as well, new mechanisms must be proposed that consider both challenges together. A promising approach, according to analytical work, is to increase the reliability by using cooperative diversity, where all stations within range help each other in the transmission process. Theoretical analyses, however, only provide a limited insight regarding the actual performance due to the strong assumptions they make to model such complex systems. In this paper, we thus evaluate the practical feasibility of ultra-reliable low-latency communication through cooperation by designing a data link protocol that incorporates a best relay selection mechanism. We implement our protocol in a real-world testbed, consisting of software-defined radios, to gain a better understanding of how future ultra-reliable low-latency systems should be designed and implemented. Our measurement campaigns show that at a given low target latency of 1 ms, we achieve a packet error rate between 10-5 and 10-7 with a standard 802.11a physical layer., Conference code: 139092; Export Date: 22 October 2018; Conference Paper; Funding details: DFG, Deutsche Forschungsgemeinschaft; Funding text: ACKNOWLEDGMENTS We would like to thank the DFG for the support within the Cluster of Excellence “Integrative Production Technology for High-Wage Countries”. REFERENCES.QC 20181022

The experience of using an artificial neural network for approximating the average height and average diameter of 187 pine stand of various ages (from 7 to 120 years) and density (from 0.4 to 10.7 thousand pieces / ha) is described in the article. As an object of research, there are pure pine stands growing in the ribbon burs of the Altai Krai territory and the Republic of Kazakhstan. All considered stands grow in dry forest growing conditions and have a different origin. Approximation of the data was carried out using the Neural Network Toolbox, which is part of the MATLAB software package. A two-layer neural network with a direct connection, a hidden layer of sigmoid-type neurons and linear output neurons was used in the course of the work. The number of neurons in the hidden layer of the network was chosen experimentally and was chosen equal to five. The aim of the work was to create a mathematical model that allows to determine the average height and average diameter of pine stand of a certain age and density. The article provides a table of the approximated values of the above taxation indices. A comparison of the approximating ability of an artificial neural network and the Mitcherlich function is made, based on the data of absolute and average approximation errors. The conclusion is drawn that the artificial neural network coped with the approximation of the taxation indices better than it was possible to do with the help of the Mitcherlich function. However, the model obtained does not describe the initial data, since the allowable limit of the mean error of approximation was exceeded. © 2018 CEUR-WS. All rights reserved.

In recent years, Deep Neural Networks (DNNs) have achieved tremendous success for diverse problems such as classification and decision making. Efficient support for DNNs on CPUs, GPUs and accelerators has become a prolific area of research, resulting in a plethora of techniques for energy-efficient DNN inference. However, previous proposals focus on a single execution of a DNN. Popular applications, such as speech recognition or video classification, require multiple back-to-back executions of a DNN to process a sequence of inputs (e.g., audio frames, images). In this paper, we show that consecutive inputs exhibit a high degree of similarity, causing the inputs/outputs of the different layers to be extremely similar for successive frames of speech or images of a video. Based on this observation, we propose a technique to reuse some results of the previous execution, instead of computing the entire DNN. Computations related to inputs with negligible changes can be avoided with minor impact on accuracy, saving a large percentage of computations and memory accesses. We propose an implementation of our reuse-based inference scheme on top of a state-of-the-art DNN accelerator. Results show that, on average, more than 60% of the inputs of any neural network layer tested exhibit negligible changes with respect to the previous execution. Avoiding the memory accesses and computations for these inputs results in 63% energy savings on average., Peer Reviewed, Postprint (published version)