Neural Networks For Signal Processing Research Articles

Mathematical Analysis of Neural Networks for Signal Processing

In flag preparing, neural systems have gotten to be exceptionally valuable since they give adaptable ways to see at and get it complicated information patterns. A full scientific ponder of neural systems utilized for flag preparing errands is given in this paper. The most objective of the consider is to figure out the fundamental thoughts, structures, and numerical equations that make neural networks work well in preparing information from distinctive regions. The fundamental thoughts of neural systems are looked at, beginning with neurons, layers, and actuation capacities, which are the network's building squares. Scientific models appear how these parts work together within the arrange and how they offer assistance turn input signals into valuable yields through rehashed learning forms. The numerical modeling of feedforward and repetitive neural systems is emphasized, appearing how well they work with both consecutive and time-series information. The think about moreover talks approximately more progressed points, such as optimization strategies like slope plummet and its varieties, which are exceptionally critical for educating neural systems to create forecasts with less botch. The science system incorporates looking into regularization strategies such as L1 and L2 regularization, which are vital for maintaining a strategic distance from overfitting and making generalization way better. Neural systems are utilized in flag handling in a wide run of real-world circumstances, counting discourse and voice acknowledgment, picture handling, organic flag examination, and broadcast communications. Case considers appear how neural systems offer assistance illuminate issues within the genuine world by finding highlights, lowering noise, and making strides flag acknowledgment in situations that alter.

Flexible Updating of Internet of Things Computing Functions through Optimizing Dynamic Partial Reconfiguration

With applications to become increasingly compute- and data-intensive, requiring more processing power, many Internet of Things (IoT) platforms in robots, drones, and autonomous vehicles that implement neural network inference, cryptographic functions or signal processing (e.g., multimedia, communication), employ field programmable gate arrays (FPGAs). At the same time, dynamic partial reconfiguration (DPR) in modern FPGAs enable changing the function of a part of the FPGA by dynamically loading new bitstreams to the logic regions without affecting the function of other parts of the FPGA. This is especially useful to update functions of IoT devices while in operation for bug fixing or functionality adjustments and, more importantly, when these IoT devices integrate low-cost FPGAs that can hardly realize many hard accelerators. To deal with one of the major limitations of using partial reconfiguration in IoT devices, this work introduces techniques to flexibly use DPR, namely, FLEXDPR, by sharing reconfigurable partitions among different accelerator functions and by supporting virtual relocation of these functions. Experimental results on the Xilinx Zynq-7000 platform reveal energy and latency efficiency improvements of about 20%, on average. Overall, the suggested approach can reduce partial reconfiguration overhead while easing the scheduler’s decisions for the deployment of hardware functions throughout time and space in a performance-conscious manner.

Neural Network Signal Processing of a Quasi-Distributed Fiber-Optic Measuring Network with Amplitude Sensors

Neural Network Signal Processing of a Quasi-Distributed Fiber-Optic Measuring Network with Amplitude Sensors

VHXLA: A post-earthquake damage prediction method for high-speed railway track-bridge system using VMD and hybrid neural network

The timely and accurate prediction of post-earthquake damage is critically important for ensuring the safety of high-speed railway track-bridge systems. The study introduces a novel method known as the VHXLA model. This innovative model blends variational mode decomposition (VMD) with a hybrid neural network for predicting non-linear multi-component post-earthquake damage in high-speed railways. The VHXLA model comprises a signal processing module and a deep learning module. The signal processing module utilises VMD decomposition, and the Hilbert transform (HT) to transform two-dimensional seismic signals into four-dimensional complex time–frequency signals, thus simplifying the task of identifying seismic time–frequency characteristics via the neural network. The deep learning module integrates diverse types of neural network components, such as the Xception CNN feature extraction submodule, the LSTM RNN temporal learning submodule, and the multi-head attention mechanism submodule. A Bayesian self-optimisation method is implemented to determine the number of decomposition layers in VMD and select essential hyperparameters. The model's effectiveness is evaluated by predicting the damage of an experimentally validated finite element model, subjecting it to seismic loads, and thus gauging its performance. The results indicated that the signal processing module, based on VMD decomposition, significantly improves the neural network's signal processing capability. In addition, the synergistic integration of different modules in the VHXLA model provides superior prediction accuracy compared to pre-existing damage prediction models. Notably, the prediction accuracy is consistent across different positions of the same predicted component in the high-speed railway track-bridge system.

Damage localization and quantification in offshore jacket structures using signal processing and intelligent system

Offshore platforms operate in harsh environmental conditions, and a significant portion of their structure is submerged underwater. Consequently, inspection and repair operations are costly and complex, making it crucial to have a precise damage detection system to increase safety and reduce maintenance expenses. This paper presents an intelligent signal processing-based system for damage detection in offshore jacket platforms. First, two damage indicators based on the signal's entropy and power are introduced to localize damage in jacket-type offshore structures. Then, an intelligent system that combines an artificial neural network and several signal processing techniques is used to quantify damage in various parts of these structures. To this end, various signal processing techniques extract the response signal features. Next, the proposed neural network model is trained using a data set, including the extracted feature vectors and the severity of the known damage scenarios. As a case study, an offshore jacket platform installed in the Persian Gulf was examined to evaluate the proposed method's effectiveness. The results reveal that the provided method detects the damage locations and severities with high processing speed and accuracy. Further, selecting suitable features and performing initial processing on training feature vectors can increase the accuracy of detecting damage intensity.

Neuromorphic computing hardware and neural architectures for robotics.

Neuromorphic hardware enables fast and power-efficient neural network-based artificial intelligence that is well suited to solving robotic tasks. Neuromorphic algorithms can be further developed following neural computing principles and neural network architectures inspired by biological neural systems. In this Viewpoint, we provide an overview of recent insights from neuroscience that could enhance signal processing in artificial neural networks on chip and unlock innovative applications in robotics and autonomous intelligent systems. These insights uncover computing principles, primitives, and algorithms on different levels of abstraction and call for more research into the basis of neural computation and neuronally inspired computing hardware.

Effect of DNA Aptamer Concentration on the Conductivity of a Water-Gated Al:ZnO Thin-Film Transistor-Based Biosensor.

Field-effect transistor-based biosensors (bio-FETs) are promising candidates for the rapid high-sensitivity and high-selectivity sensing of various analytes in healthcare, clinical diagnostics, and the food industry. However, bio-FETs still have several unresolved problems that hinder their technological transfer, such as electrical stability. Therefore, it is important to develop reliable, efficient devices and establish facile electrochemical characterization methods. In this work, we have fabricated a flexible biosensor based on an Al:ZnO thin-film transistor (TFT) gated through an aqueous electrolyte on a polyimide substrate. In addition, we demonstrated techniques for establishing the operating range of such devices. The Al:ZnO-based devices with a channel length/width ratio of 12.35 and a channel thickness of 50 nm were produced at room temperature via magnetron sputtering. These Al:ZnO-based devices exhibited high field-effect mobility (μ = 6.85 cm2/Vs) and threshold voltage (Vth = 654 mV), thus showing promise for application on temperature-sensitive substrates. X-ray photoelectron spectroscopy was used to verify the chemical composition of the deposited films, while the morphological aspects of the films were assessed using scanning electron and atomic force microscopies. The gate–channel electric capacitance of 40 nF/cm2 was determined using electrochemical impedance spectroscopy, while the electrochemical window of the gate–channel system was determined as 1.8 V (from −0.6 V to +1.2 V) using cyclic voltammetry. A deionized water solution of 10 mer (CCC AAG GTC C) DNA aptamer (molar weight −2972.9 g/mol) in a concentration ranging from 1–1000 pM/μL was used as an analyte. An increase in aptamer concentration caused a proportional decrease in the TFT channel conductivity. The techniques demonstrated in this work can be applied to optimize the operating parameters of various semiconductor materials in order to create a universal detection platform for biosensing applications, such as multi-element FET sensor arrays based on various composition nanostructured films, which use advanced neural network signal processing.

Environmental Audio Tagging Using Deep Convolution Neural Network and Digital Signal Processing

Machine learning has experienced a strong growth in recent years, due to increased dataset sizes and computational power, and to advances in deep learning methods that can learn to make predictions in extremely non-linear problem settings. The intense problem of automatic environmental sound classification has received alarming attention from the research community in recent years. In this paper the audio dataset is converted into mass spectrogram using Digital Signal Processing (DSP). The spectrogram thus obtained is fed to the Convolutional Neural Network (CNN) for the classification of the audio signal. In this we present a deep convolutional neural network architecture with localized kernels for environmental sound. By training the network on another additional deformed data, the hope is that the network becomes invariant to all deformations and generalizes better to all unseen data. We show that the proposed DSP in combination with CNN architecture, yields state-of-the-art performance for environmental sound classification.

CHOOSING A NEURAL NETWORK STRUCTURE FOR SIGNAL PROCESSING AS AN EXPERIMENT PLANNING

The paper is devoted to the use of artificial neural networks for signal processing in electrical engineering and electric power industry. Direct propagation neural network (perceptron) is considered as an object in the theory of experiment planning. The variants of the neural network structure empirical choice, the quality criteria of its training and testing are analyzed. It is shown that the perceptron structure choice, the training sample, and the training algorithms require planning. Variables and parameters of neuro algorithm that can act as factors, state parameters, and disturbing influences in the framework of the experimental planning theory are discussed. The proposed approach is demonstrated by the example of neural network analysis of the industrial frequency signal of 50 Hz nonlinear distortions. The possibility of using an elementary perceptron with one hidden layer and a minimum number of neurons to correct the transformer saturation current is analyzed. The conditions under which the neuro algorithm allows one to restore the values of the main harmonic amplitude, frequency and phase with an error of no more than one percent are revealed. The signal processing in a «sliding window» with a duration of a fraction of the fundamental frequency period is proposed, and the neuro algorithm accuracy characteristics are estimated. The possibility to automate the neural network structure choosing for signal processing is discussed.

Neural Networks Application on Human Skin Biophysical Impedance Characterizations

Artificial neural networks (ANNs) are basically the structures that perform input–output mapping. This mapping mimics the signal processing in biological neural networks. The basic element of biological neural network is a neuron. Neurons receive input signals from other neurons or the environment, process them, and generate their output which represents the input to another neuron of the network. Neurons can change their sensitivity to input signals. Each neuron has a simple rule to process an input signal. Biological neural networks have the property that signals are processed through many parallel connections (massively parallel processing). The activity of all neurons in these parallel connections is summed and represents the output of the whole network. The main feature of biological neural networks is that changes in the sensitivity of the neurons lead to changes in the operation of the entire network. This is called adaptation and is correlated with the learning process of living organisms. In this paper, a set of artificial neural networks are used for classifying the human skin biophysical impedance data.

Novel Arithmetics in Deep Neural Networks Signal Processing for Autonomous Driving: Challenges and Opportunities

This article focuses on the trends, opportunities, and challenges of novel arithmetic for deep neural network (DNN) signal processing, with particular reference to assisted- and autonomous driving applications. Due to strict constraints in terms of the latency, dependability, and security of autonomous driving, machine perception (i.e., detection and decision tasks) based on DNNs cannot be implemented by relying on remote cloud access. These tasks must be performed in real time in embedded systems on board the vehicle, particularly for the inference phase (considering the use of DNNs pretrained during an offline step). When developing a DNN computing platform, the choice of the computing arithmetic matters. Moreover, functional safe applications, such as autonomous driving, impose severe constraints on the effect that signal processing accuracy has on the final rate of wrong detection/decisions. Hence, after reviewing the different choices and tradeoffs concerning arithmetic, both in academia and industry, we highlight the issues in implementing DNN accelerators to achieve accurate and lowcomplexity processing of automotive sensor signals (the latter coming from diverse sources, such as cameras, radar, lidar, and ultrasonics). The focus is on both general-purpose operations massively used in DNNs, such as multiplying, accumulating, and comparing, and on specific functions, including, for example, sigmoid or hyperbolic tangents used for neuron activation.

Feasibility study of range verification based on proton-induced acoustic signals and recurrent neural network

Range verification in proton therapy is a critical quality assurance task. We studied the feasibility of online range verification based on proton-induced acoustic signals, using a bidirectional long-short-term-memory recurrent neural network and various signal processing techniques. Dose distribution of 1D pencil proton beams inside a CT image-based phantom was analytically calculated. The propagation of acoustic signal inside the phantom was modeled using the k-Wave toolbox. For signal processing, five methods were investigated: down-sampling (DS), DS + HT (Hilbert transform), Wavelet decomposition (Wavedec db1, db4 and db20). The performances were quantitatively evaluated in terms of mean absolute error, mean relative error (MRE) and the Bragg peak localization error (). In addition, the study analyzed the impact of noise levels, the number of sensors, as well as the location of sensors. For the noiseless case (32 sensors), the Wavedec db1 method demonstrates the best performance: is less than one pixel and the dose accuracy over the region adjacent to the Bragg peak (MRE50) is ∼3.04%. With the presence of noise, the Wavedec db1 method demonstrates the best noise immunity, achieving less than 1 mm and an MRE50 of ∼12%. The proposed machine learning framework may become a useful tool allowing for online range verification in proton therapy.

CMUT-based biosensor with convolutional neural network signal processing

The improvement of the micromachined ultrasound transducer based (CMUT) biosensor fabrication technology and signal processing, which led to higher signal to noise ratio is reported. The biosensor contains interdigitally arranged CMUT structure with gold-coated analytical area. It is assembled with the plexiglass microchannels. CMUTs were fabricated with the wafer bonding technology for 5 MHz operation in immersion. For signal processing the convolutional neural network (CNN) was developed and trained to classify the sensor data to different propagation delay values. For training of the network 750 thousand signals representing different properties of the bioanalyte and different noise conditions was simulated by the finite time difference domain (FDTD) model. The capability of the CNN algorithm to classify the propagation delay data was compared with the adaptive passband filter signal processing algorithm used in our previous version of the senor. Both sensing channels were run simultaneously with the reference liquids in the microchannel: deionized water switching to 0.9% saline. It was found that CNN channel is capable to improve the signal to noise ratio for this experiment to 75 dB, when the same property for the passband filter channel was only 60 dB. This led to the generalization about the advantage of CNN channel to provide 15 dB less of instrumental noise. Finally, the real-time detection ability of the bovine serum albumin (BSA) deposition on the analytical area of improved sensor was demonstrated.

Synaptic inhibition tunes contrast computation in the retina.

Inhibition shapes activity and signal processing in neural networks through numerous mechanisms mediated by many different cell types. Here, we examined how one type of GABAergic interneuron in the retina, the A17 amacrine cell, influences visual information processing. Our results suggest that A17s, which make reciprocal feedback inhibitory synapses onto rod bipolar cell (RBC) synaptic terminals, extend the luminance range over which RBC synapses compute temporal contrast and enhance the reliability of contrast signals over this range. Inhibition from other amacrine cells does not influence these computational features. Although A17-mediated feedback is mediated by both GABAA and GABAC receptors, the latter plays the primary role in extending the range of contrast computation. These results identify specific functions for an inhibitory interneuron subtype, as well as specific synaptic receptors, in a behaviorally relevant neural computation.

GMR Sensors and Neural Networks Applied to the Contactless Measurement of Direct Electrical Currents

Clamp ammeters based on coils are restricted to the measurement of alternating currents, through the associated magnetic field. There are also commercial versions able of measuring direct currents, based on Hall effect sensors. This manuscript presents improvements of a previously presented prototype of a contactless ammeter based on commercial giant magnetoresistance magnetometers, associated with neural networks for signal processing, able to measure direct electrical currents and to infer the distance between the sensor and the electrical current conductor.

Conduction Delay Learning Model for Unsupervised and Supervised Classification of Spatio-Temporal Spike Patterns.

Precise spike timing is considered to play a fundamental role in communications and signal processing in biological neural networks. Understanding the mechanism of spike timing adjustment would deepen our understanding of biological systems and enable advanced engineering applications such as efficient computational architectures. However, the biological mechanisms that adjust and maintain spike timing remain unclear. Existing algorithms adopt a supervised approach, which adjusts the axonal conduction delay and synaptic efficacy until the spike timings approximate the desired timings. This study proposes a spike timing-dependent learning model that adjusts the axonal conduction delay and synaptic efficacy in both unsupervised and supervised manners. The proposed learning algorithm approximates the Expectation-Maximization algorithm, and classifies the input data encoded into spatio-temporal spike patterns. Even in the supervised classification, the algorithm requires no external spikes indicating the desired spike timings unlike existing algorithms. Furthermore, because the algorithm is consistent with biological models and hypotheses found in existing biological studies, it could capture the mechanism underlying biological delay learning.

Spike-Based Synaptic Plasticity in Silicon: Design, Implementation, Application, and Challenges

The ability to carry out signal processing, classification, recognition, and computation in artificial spiking neural networks (SNNs) is mediated by their synapses. In particular, through activity-dependent alteration of their efficacies, synapses play a fundamental role in learning. The mathematical prescriptions under which synapses modify their weights are termed synaptic plasticity rules. These learning rules can be based on abstract computational neuroscience models or on detailed biophysical ones. As these rules are being proposed and developed by experimental and computational neuroscientists, engineers strive to design and implement them in silicon and en masse in order to employ them in complex real-world applications. In this paper, we describe analog very large-scale integration (VLSI) circuit implementations of multiple synaptic plasticity rules, ranging from phenomenological ones (e.g., based on spike timing, mean firing rates, or both) to biophysically realistic ones (e.g., calcium-dependent models). We discuss the application domains, weaknesses, and strengths of various representative approaches proposed in the literature, and provide insight into the challenges that engineers face when designing and implementing synaptic plasticity rules in VLSI technology for utilizing them in real-world applications.

From evolving artificial gene regulatory networks to evolving spiking neural networks for signal processing

From evolving artificial gene regulatory networks to evolving spiking neural networks for signal processing

Phasic and Tonic Patterns of Locus Coeruleus Output Differentially Modulate Sensory Network Function in the Awake Rat

Neurons of the nucleus locus coeruleus (LC) discharge with phasic bursts of activity superimposed on highly regular tonic discharge rates. Phasic bursts are elicited by bottom-up input mechanisms involving novel/salient sensory stimuli and top-down decision making processes; whereas tonic rates largely fluctuate according to arousal levels and behavioral states. Although it is generally believed that these two modes of activity differentially modulate information processing in LC targets, the unique role of phasic versus tonic LC output on signal processing in cells, circuits, and neural networks of waking animals is not well understood. In the current study, simultaneous recordings of individual neurons within ventral posterior medial thalamus and barrel field cortex of conscious rats provided evidence that each mode of LC output produces a unique modulatory impact on single neuron responsiveness to sensory-driven synaptic input and representations of sensory information across ensembles of simultaneously recorded cells. Each mode of LC activation specifically modulated the relationship between sensory-stimulus intensity and the subsequent responses of individual neurons and neural ensembles. Overall these results indicate that phasic versus tonic modes of LC discharge exert fundamentally different modulatory effects on target neuronal circuits within the rodent trigeminal somatosensory system. As such, each mode of LC output may differentially influence signal processing as a means of optimizing behaviorally relevant neural computations within this sensory network. Likely the ability of the LC system to differentially regulate neural responses and local circuit operations according to behavioral demands extends to other brain regions including those involved in higher cognitive functions.

Optimally regularised kernel Fisher discriminant classification

Mika, Rätsch, Weston, Schölkopf and Müller [Mika, S., Rätsch, G., Weston, J., Schölkopf, B., & Müller, K.-R. (1999). Fisher discriminant analysis with kernels. In Neural networks for signal processing: Vol. IX (pp. 41-48). New York: IEEE Press] introduce a non-linear formulation of Fisher's linear discriminant, based on the now familiar "kernel trick", demonstrating state-of-the-art performance on a wide range of real-world benchmark datasets. In this paper, we extend an existing analytical expression for the leave-one-out cross-validation error [Cawley, G. C., & Talbot, N. L. C. (2003b). Efficient leave-one-out cross-validation of kernel Fisher discriminant classifiers. Pattern Recognition, 36(11), 2585-2592] such that the leave-one-out error can be re-estimated following a change in the value of the regularisation parameter with a computational complexity of only O(l(2)) operations, which is substantially less than the O(l(3)) operations required for the basic training algorithm. This allows the regularisation parameter to be tuned at an essentially negligible computational cost. This is achieved by performing the discriminant analysis in canonical form. The proposed method is therefore a useful component of a model selection strategy for this class of kernel machines that alternates between updates of the kernel and regularisation parameters. Results obtained on real-world and synthetic benchmark datasets indicate that the proposed method is competitive with model selection based on k-fold cross-validation in terms of generalisation, whilst being considerably faster.