Training Signals Research Articles

Adversarial network integrating dual attention and sparse representation for semi-supervised semantic segmentation

Semantic segmentation is the important task of assigning a semantic label to each pixel. However, semantic segmentation based on the deep neural network usually requires massive annotations consumption to acquire better performance. To avoid the problem, some algorithms based on weakly-supervised and semi-supervised conditions have been proposed and achieved gradually improving performance in recent years. In this paper, we propose a novel semi-supervised adversarial network to alleviate the shortage of labeled data, which only requires a few labeled images to get competitive performance. The model is composed of two parts: the segmentation network and the discriminator network. The first part aims to semantically generate a segmented result that has the same size as the input color image. The discriminator network is designed in a fully convolutional manner to distinguish the predicted probability maps depending on the ground truth distribution. In particular, the probability maps are regarded as focal attention maps, which are fed back to the segmentation network to make the model converge faster, and the process can induce the model to focus on pixels that are hard to segment. To enhance the representation ability of image features, sparse representation and dual attention are adopted in the segmentation network. The sparse representation module aims to emphasize the object edges and locations by learning the convolutional sparse representation of the input color images, and the dual attention module can exploit the semantic interdependencies in two different dimensions. Moreover, the semi-supervised mechanism is introduced to the network, in which the adaptive parameter T that controls the sensitivity of the self-taught phase is proposed, and the training dataset is split into two parts for fully-supervised learning and semi-supervised learning. Specifically, the first part is unlabeled data, which is applied to provide supervised signals for semi-supervised training. The labeled data drawn from the other part is utilized for fully-supervised learning. Our semi-supervised adversarial framework can improve the learning ability and achieve higher performance, also, provide a novel approach to tackle the semantic segmentation task. Finally, comprehensive experiments on the PASCAL VOC 2012 and Cityscapes datasets are conducted to verify the effectiveness of the proposed model, which achieves the expected performance.

Efficient Graph Collaborative Filtering via Contrastive Learning.

Collaborative filtering (CF) aims to make recommendations for users by detecting user’s preference from the historical user–item interactions. Existing graph neural networks (GNN) based methods achieve satisfactory performance by exploiting the high-order connectivity between users and items, however they suffer from the poor training efficiency problem and easily introduce bias for information propagation. Moreover, the widely applied Bayesian personalized ranking (BPR) loss is insufficient to provide supervision signals for training due to the extremely sparse observed interactions. To deal with the above issues, we propose the Efficient Graph Collaborative Filtering (EGCF) method. Specifically, EGCF adopts merely one-layer graph convolution to model the collaborative signal for users and items from the first-order neighbors in the user–item interactions. Moreover, we introduce contrastive learning to enhance the representation learning of users and items by deriving the self-supervisions, which is jointly trained with the supervised learning. Extensive experiments are conducted on two benchmark datasets, i.e., Yelp2018 and Amazon-book, and the experimental results demonstrate that EGCF can achieve the state-of-the-art performance in terms of Recall and normalized discounted cumulative gain (NDCG), especially on ranking the target items at right positions. In addition, EGCF shows obvious advantages in the training efficiency compared with the competitive baselines, making it practicable for potential applications.

Artificial Noise Assisted In-Band Full-Duplex Secure Channel Estimation

This paper proposes a novel secure channel estimation technique to provide security against leakage of the channel estimates to any malicious user by utilizing artificial noise (AN) along with full-duplex (FD) transmissions. AN overcomes the drawback of FD transmission, where any strategically located eavesdropper can minimize the interference signal received from the FD receiver. The proposed secure channel estimation technique comprises three stages, where the first stage is responsible for the estimation of the residual self-interference (SI) channel. The second stage acquires rough channel estimates to design AN orthogonal to the channel between legitimate transmitter-receiver for the next training stage. In the third stage, both legitimate nodes transmit orthogonal AN signals along with the known training signals using FD transmissions. For power allocation, we have presented a novel local adaptive power allocation algorithm at each legitimate node to allocate the powers to the training signals, and AN signals while ensuring equivocation at the eavesdropper. We provide the mean square error (MSE) to indicate the performance achieved by the respective nodes. We have also provided the bit error rate (BER) simulation analysis to indicate the secure communication achieved by securing the channel estimation process. The presented simulation analysis indicates that the eavesdropper is unable to decode the transmitted information while the legitimate receiver has robustly decoded the transmitted information.

Graph Based Training Resource Allocation Scheme for CoMP Transmission in CRAN: A Low Complexity Solution

Coordinated Multi Point (CoMP) transmission in Cloud Radio Area Network (CRAN) strives to provide the required Quality of Service (QoS), irrespective of user's location. Specifically, CoMP enables intra-cluster Access Points (APs) to maximize user's performance utilizing channel information, which is obtained by exchanging prior-known pilot signals among the intended users and intra-cluster APs. These signals are often referred as training signals, and are shared over Training Resources (TRs). However, CoMP is pillared on user-centric network, thereby the resulting clusters are inevitably overlapped. As a result, the TR allocation schemes for these networks require considerably large computational complexity. In this paper, we introduce the notion of Time Slot Tagging (TST), in which each AP is tagged with an appropriate training slot. Further, clusters are formed on the basis of this tagged information. Specifically, TST eliminates the requirement of resource allocation for individual users which saves a significant time in each clustering cycle, and thus enhances the link utilization factor. Also, TST is designed in a manner that it assures the presence of preferred APs in each intended cluster. Further, it is shown that the proposed approach performs significantly close to the most optimistic scenario.

Autonomous Driving using CNN

Human beings are currently addicted to automation and robotics technologies. The state-of-the-art in deep learning technologies and AI is the subject of this autonomous driving. Driving with automated driving systems promises to be safe, enjoyable, and efficient.. It is preferable to train in a virtual environment first and then move to a real-world one. Its goal is to enable a vehicle to recognise its surroundings and navigate without the need for human intervention. The raw pixels from a single front-facing camera were directly transferred to driving commands using a convolution neural network (CNN). This end-to-end strategy proved to be remarkably effective, The system automatically learns internal representations of the essential processing stages such as detecting useful road components using only the human steering angle as the training signal. We never expressly taught it to recognise the contour of roadways, for example. In comparison to explicit issue decomposition, such as lane marking detection, Our end-to-end solution optimises all processing processes at the same time, including path planning and control. We believe that this will lead to improved performance and smaller systems in the long run. Internal components will self-optimize to maximise overall system performance, resulting in improved performance.

Emitter Identification of Digital Modulation Transmitter Based on Nonlinearity and Modulation Distortion of Power Amplifier.

Specific transmitter identification (SEI) is a technology that uses a received signal to identify to which individual radiation source the transmitted signal belongs. It can complete the identification of the signal transmitter in a non-cooperative scenario. Therefore, there are broad application prospects in the field of wireless-communication-network security, spectral resource management, and military battlefield-target communication countermeasures. This article demodulates and reconstructs a digital modulation signal to obtain a signal without modulator distortion and power-amplifier nonlinearity. Comparing the reconstructed signal with the actual received signal, the coefficient representation of the nonlinearity of the power amplifier and the distortion of the modulator can be obtained, and these coefficients can be used as the fingerprint characteristics of different transmitters through a convolutional neural network (CNN) to complete the identification of specific transmitters. The existing SEI strategy for changing the modulation parameters of a test signal is to mix part of the test signal with the training signal so that the classifier can learn the signal of which the modulation parameter was changed. This method is still data-oriented and cannot process signals for which the classifier has not been trained. It has certain limitations in practical applications. We compared the fingerprint features extracted by the method in this study with the fingerprint features extracted by the bispectral method. When SNR < 20 dB, the recognition accuracy of the bispectral method dropped rapidly. The method in this paper still achieved 86% recognition accuracy when SNR = 0 dB. When the carrier frequency of the test signal was changed, the bispectral feature failed, and the proposed method could still achieve a recognition accuracy of about 70%. When changing the test-signal baud rate, the proposed method could still achieve a classification accuracy rate of more than 70% for four different individual radiation sources when SNR = 0 dB.

Optimizing Reservoir Computers for Signal Classification.

Reservoir computers are a type of recurrent neural network for which the network connections are not changed. To train the reservoir computer, a set of output signals from the network are fit to a training signal by a linear fit. As a result, training of a reservoir computer is fast, and reservoir computers may be built from analog hardware, resulting in high speed and low power consumption. To get the best performance from a reservoir computer, the hyperparameters of the reservoir computer must be optimized. In signal classification problems, parameter optimization may be computationally difficult; it is necessary to compare many realizations of the test signals to get good statistics on the classification probability. In this work, it is shown in both a spiking reservoir computer and a reservoir computer using continuous variables that the optimum classification performance occurs for the hyperparameters that maximize the entropy of the reservoir computer. Optimizing for entropy only requires a single realization of each signal to be classified, making the process much faster to compute.

Neural network learning of improved compressive sensing sampling and receptive field structure

While the theory of compressive sensing (CS) in modern signal processing typically indicates that uniformly random sampling facilitates the efficient recovery of sparse signals, such measurements are infeasible in many engineering applications and are not well reflected by the constraints of natural systems, including neuronal networks in the brain. Uniformly random sampling also does not leverage the underlying structure of many classes of signals, and may therefore be suboptimal in these cases. We address these issues by formulating a novel neural network framework for learning improved CS sampling based on the intrinsic structure present in classes of training signals. Beyond sparsity in an appropriate domain, this approach does not assume knowledge of any specific signal statistics and is purely data-driven. The learning methodology is biologically realistic in that it utilizes (1) asymmetric feedback and feedforward connections in the neural network and (2) only information from adjacent layers in training the CS measurement matrix. Observing a broad spectrum of learned sampling paradigms that improve CS signal reconstructions relative to uniformly random sampling, our learned sampling is widely applicable across logistical constraints. Motivated by the receptive field structure of sensory systems, we specifically analyze natural scene inputs and demonstrate improved CS reconstruction as a result of training across several choices of penalization schemes on the sampling weights. Considering this learning is effective even under sparse and spatially localized constraints, as commonly observed in the brain, we hypothesize that neuronal connectivity may have manifested with the aim of providing a compressive encoding of data by leveraging its sparse structure, thereby achieving efficient signal transmission.

Learning an Effective Context-Response Matching Model with Self-Supervised Tasks for Retrieval-based Dialogues

Building an intelligent dialogue system with the ability to select a proper response according to a multi-turn context is a great challenging task. Existing studies focus on building a context-response matching model with various neural architectures or pretrained language models (PLMs) and typically learning with a single response prediction task. These approaches overlook many potential training signals contained in dialogue data, which might be beneficial for context understanding and produce better features for response prediction. Besides, the response retrieved from existing dialogue systems supervised by the conventional way still faces some critical challenges, including incoherence and inconsistency. To address these issues, in this paper, we propose learning a context-response matching model with auxiliary self-supervised tasks designed for the dialogue data based on pre-trained language models. Specifically, we introduce four self-supervised tasks including next session prediction, utterance restoration, incoherence detection and consistency discrimination, and jointly train the PLM-based response selection model with these auxiliary tasks in a multi-task manner. By this means, the auxiliary tasks can guide the learning of the matching model to achieve a better local optimum and select a more proper response. Experiment results on two benchmarks indicate that the proposed auxiliary self-supervised tasks bring significant improvement for multi-turn response selection in retrieval-based dialogues, and our model achieves new state-of-the-art results on both datasets.

Edinburgh accountants in public practice pre-collective organisation: 1757–1834

ABSTRACT This study is intended to increase knowledge and improve understanding of early public accountancy professionalisation in Scotland by applying the prosopographical research method to a community of practitioners in the capital city of Edinburgh in the early-nineteenth century. Using archival data, the study identifies the collective professional and social characteristics of 124 Edinburgh practitioners in 1834 by means of career-related analyses of their origin, education, training, and service-related signals of movement to occupational ascendency prior to the community’s later collective organisation. The study makes visible a structured and mature community operating in several occupational jurisdictions involving multi-disciplinary knowledge; maintaining a subordinate but mutually-dependent relationship with the legal profession; having a primary role in emerging insurance services; and achieving individual practitioner status recognition in a class-conscious city. Evidence of signals of movement to occupational ascendency adds to existing knowledge and understanding of the pre-collective organisation phase of public accountancy professionalisation in Scotland.

Modeling of high-speed laser photography system for field projectile testing

Field projectile testing plays a crucial role in national defense. Laser high-speed photography records transient processes with high temporal and spatial resolution and is an effective diagnostic method for range testing. In this study, a laser high-speed photography model is established from detailed aspects, including laser pulse energy for reflective imaging, accurate synchronization of pulse train and shutter signal, minimum exposure time and shooting frequency for targets with different velocities, and magnification correction used in the image measurement. Shooting experiments conducted in Nanshan Test center verify the correctness of the theoretical analysis and parameter selection. The models and methods described in this article can be a guidance for field testing.

Direction‐based jamming detection and suppression in mmWave massive MIMO networks

In this paper, we study the problem of physical layer security in the uplink of millimeter-wave massive multiple-input multiple-output (MIMO) networks and propose a jamming detection and suppression method. The proposed method is based on directional information of the received signals at the base station antenna array. The proposed jamming detection method can accurately detect both the existence and direction of the jammer using the received pilot signals in the training phase. The obtained information is then exploited to develop a channel estimator that excludes the jammer's angular subspace from received training signals. The estimated channel information is then used for designing a combiner at the base station that is able to effectively cancel out the deliberate interference of the jammer. By numerical simulations, we evaluate the performance of the proposed jamming detection method in terms of correct detection probability and false alarm probability and show its effectiveness when the jammer's power is substantially lower than the user's power. Also, our results show that the proposed jamming suppression method can achieve a very close spectral efficiency as the case of no jamming in the network

Joint Message-Passing-Based Bidirectional Channel Estimation and Equalization With Superimposed Training for Underwater Acoustic Communications

Acquiring accurate channel state information and mitigating severe intersymbol interference are challenging for underwater acoustic communications with moving transceivers due to the rapid changes of the underwater acoustic channels. In this work, we address the issue using a superimposed training (ST) scheme with a powerful channel estimation method. Different from the conventional time-multiplexed training, training sequences with a small power are superimposed with symbol sequences. The training signals are transmitted over all time, leading to enhanced tracking capability to deal with time-varying channels at the cost of only a small power loss. To realize this, based on the belief propagation, we develop a message-passing-based bidirectional channel estimation (BCE) algorithm, where all messages are Gaussian, enabling efficient implementation. In particular, the channel correlations are fully exploited through a forward recursion and a backward recursion, thereby achieving accurate channel estimation. Moreover, the ST-based BCE is combined with channel equalization (in the frequency domain) and decoding, and they are performed jointly in an iterative manner to significantly enhance the overall system performance. Field experiments were carried out in Jiaozhou Bay in 2019, and the results verify the effectiveness of the proposed scheme and algorithm.

Downlink Channel Reconstruction for Spatial Multiplexing in Massive MIMO Systems

To get channel state information (CSI) at a base station (BS), most of researches on massive multiple-input multiple-output (MIMO) systems consider time division duplexing (TDD) to get benefit from the uplink and downlink channel reciprocity. Even in TDD, however, the BS still needs to transmit downlink training signals, which are referred to as channel state information reference signals (CSI-RSs) in the 3GPP standard, to support spatial multiplexing in practice. This is because there are many cases that the number of transmit antennas is less than the number of receive antennas at a user equipment (UE) due to power consumption and circuit complexity issues. Because of this mismatch, uplink sounding reference signals (SRSs) from the UE are not enough for the BS to obtain full downlink MIMO CSI. Therefore, after receiving the downlink CSI-RSs, the UE needs to feed back quantized CSI to the BS using a pre-defined codebook to support spatial multiplexing. In this paper, possible approaches to reconstruct full downlink MIMO CSI at the BS are proposed by exploiting both the SRS and quantized downlink CSI considering practical antenna structures with reduced downlink CSI-RS overhead. Numerical results show that the spectral efficiencies by spatial multiplexing based on the proposed downlink MIMO CSI reconstruction techniques outperform the conventional methods solely based on the quantized CSI.

Learned Camera Gain and Exposure Control for Improved Visual Feature Detection and Matching

Successful visual navigation depends upon capturing images that contain sufficient useful information. In this letter, we explore a data-driven approach to account for environmental lighting changes, improving the quality of images for use in visual odometry (VO) or visual simultaneous localization and mapping (SLAM). We train a deep convolutional neural network model to predictively adjust camera gain and exposure time parameters such that consecutive images contain a maximal number of matchable features. The training process is fully self-supervised: our training signal is derived from an underlying VO or SLAM pipeline and, as a result, the model is optimized to perform well with that specific pipeline. We demonstrate through extensive real-world experiments that our network can anticipate and compensate for dramatic lighting changes (e.g., transitions into and out of road tunnels), maintaining a substantially higher number of inlier feature matches than competing camera parameter control algorithms.

AoA Estimation for OAM Communication Systems With Mode-Frequency Multi-Time ESPRIT Method

Radio orbital angular momentum (OAM) communications require accurate alignment between the transmit and receive beam directions. Accordingly, a key feature of OAM receivers is the ability to reliably estimate the angle of arrival (AoA) of multi-mode OAM beams. Considering the limitations of existing AoA estimation techniques, in this paper, we propose an easier-to-implement AoA estimation method based on applying multiple times the estimating signal parameters via rotational invariance techniques (ESPRIT) algorithm to the received training signals in OAM mode and frequency domains, which is denoted as the mode-frequency (M-F) multi-time (MT)-ESPRIT algorithm. With this method, the misalignment error of real OAM channels can be greatly reduced and the performance approaches that of ideally aligned OAM channels.

Learning Spring Mass Locomotion: Guiding Policies With a Reduced-Order Model

In this letter, we describe an approach to achieve dynamic legged locomotion on physical robots which combines existing methods for control with reinforcement learning. Specifically, our goal is a control hierarchy in which highest-level behaviors are planned through reduced-order models, which describe the fundamental physics of legged locomotion, and lower level controllers utilize a learned policy that can bridge the gap between the idealized, simple model and the complex, full order robot. The high-level planner can use a model of the environment and be task specific, while the low-level learned controller can execute a wide range of motions so that it applies to many different tasks. In this letter, we describe this learned dynamic walking controller and show that a range of walking motions from reduced-order models can be used as the command and primary training signal for learned policies. The resulting policies do not attempt to naively track the motion (as a traditional trajectory tracking controller would) but instead balance immediate motion tracking with long term stability. The resulting controller is demonstrated on a human scale, unconstrained, untethered bipedal robot at speeds up to 1.2 m/s. This letter builds the foundation of a generic, dynamic learned walking controller that can be applied to many different tasks.

High Rate CV-QKD Secured Mobile WDM Fronthaul for Dense 5G Radio Networks

A coherent transmission methodology for a continuous-variable quantum key distribution (CV-QKD) system based on quantum-heterodyne measurement through a coherent intradyne receiver is experimentally demonstrated in the framework of 5G mobile fronthaul links. Continuous optical carrier synchronization is obtained through training information, which is multiplexing to the quantum signal as pilot tone in both, frequency and polarization. Spectral tailoring by means of optical carrier suppression and single-sideband modulation is adopted to simultaneously mitigate crosstalk into the quantum channel and self-interference for the pilot tone, thus allowing for a high signal-to-noise ratio for this training signal. Frequency offset correction and optical phase estimation for the free-running local oscillator of the receiver is accurately performed and guarantees low-noise quantum signal reception at high symbol rates of 250 MHz and 500 MHz with additional Nyquist pulse shaping. A low excess noise in the order of 0.1% to 0.5% of shot-noise units is obtained for fiber-based transmission over a fronthaul link reach of 13.2 km. Moreover, co-existence with 11 carrier-grade classical signals is experimentally investigated. Joint signal transmission in the C-band of both, quantum signal and classical signals, is successfully demonstrated. Secure-key rates of 18 and 10 Mb/s are obtained under strict security assumptions, where Eve has control of the receiver noise, for a dark and a lit fiber link, respectively. Moreover, rates of 85 and 72 Mb/s are resulting for a trusted receiver scenario. These secure-key rates are well addressing the requirements for time-shared CV-QKD system in densified 5G radio access networks with cloud-based processing.

Evaluation of an Information Flow Gain Algorithm for Microsensor Information Flow in Limber Motor Rehabilitation

This paper conducts an evaluative study on the rehabilitation of limb motor function by using a microsensor information flow gain algorithm and investigates the surface electromyography (EMG) signals of the upper limb during rehabilitation training. The surface EMG signals contain a large amount of limb movement information. By analysing and processing the surface EMG signals, we can grasp the human muscle movement state and identify the human upper limb movement intention. The EMG signals were processed by the trap and filter combination denoising method and wavelet denoising method, respectively, the signal-to-noise ratio was used to evaluate the noise reduction effect, and finally, the wavelet denoising method with a better noise reduction effect was selected to process all the EMG signals. After the noise is removed, the signal is extracted in the time domain and frequency domain, and the root mean square (RMS), absolute mean, median frequency in the time domain, and average power frequency in the frequency domain are selected and input to the classifier for pattern recognition. The support vector machine is used to classify the myoelectric signals and optimize the parameters in the support vector machine using the grid search method and particle swarm optimization algorithm and classify the test samples using the trained support vector machine. Compared with the classification results of the grid search optimized support vector machine, the optimized vector machine has a 7% higher recognition rate, reaching 85%. The action recognition classification method of myoelectric signals is combined with an upper limb rehabilitation training platform to verify the feasibility of using myoelectric signals for rehabilitation training. After the classifier recognizes the upper limb movements, the upper computer sends movement commands to the controller to make the rehabilitation platform move according to the recognition results, and finally, the movement execution accuracy of the rehabilitation platform reaches 80% on average.

Visual Navigation With Multiple Goals Based on Deep Reinforcement Learning.

Learning to adapt to a series of different goals in visual navigation is challenging. In this work, we present a model-embedded actor-critic architecture for the multigoal visual navigation task. To enhance the task cooperation in multigoal learning, we introduce two new designs to the reinforcement learning scheme: inverse dynamics model (InvDM) and multigoal colearning (MgCl). Specifically, InvDM is proposed to capture the navigation-relevant association between state and goal and provide additional training signals to relieve the sparse reward issue. MgCl aims at improving the sample efficiency and supports the agent to learn from unintentional positive experiences. Besides, to further improve the scene generalization capability of the agent, we present an enhanced navigation model that consists of two self-supervised auxiliary task modules. The first module, which is named path closed-loop detection, helps to understand whether the state has been experienced. The second one, namely the state-target matching module, tries to figure out the difference between state and goal. Extensive results on the interactive platform AI2-THOR demonstrate that the agent trained with the proposed method converges faster than state-of-the-art methods while owning good generalization capability. The video demonstration is available at https://vsislab.github.io/mgvn.