Time Delay Neural Network Research Articles

Crime Spatiotemporal Prediction With Fused Objective Function in Time Delay Neural Network

In the criminology area, to detain the serial criminal, the forthcoming serial crime time, distance, and criminal's biography are essential keys. The main concern of this study is on the upcoming serial crime distance, time, and suspect biographies such as age and nationality. In conjunction with having time delays, the dynamic classifier, like Time Delay Neural Network (TDNN) utilized to perform nonlinear techniques-based predictions. The TDNN classifier system, like Back Propagation Through Time (BPTT) and Nonlinear Autoregressive with Exogenous Input (NARX) are two prominent examples. However, BPTT and NARX techniques are unable to identify the dynamic system by using single-activation functions due to producing lower accuracy. Hence, during the training phase, the direct minimization of the TDNN error can further enhance the single activation function. Thus, this work introduces an enhanced NARX (eNARX) model based on the proposed activation functions of SiRBF via fusion of two functions of the hyperbolic tangent (Tansig) and Radial Basis Function (RBF), in the same hidden layer. If a fusion of activation functions can affect the TDNN error minimization, then fusing of the Tansig and RBF functions can produce a precise prediction for crime spatiotemporal. To evaluate the proposed technique and compared it with existing NARX and BPTT, we utilized five time-series datasets, namely, Dow Jones Index, Monthly River flow in cubic meters per second, Daily temperature, and UKM-PDRM datasets namely, “Suspect & Capture” and “Crime Plotting.” The analysis of the results demonstrated that the proposed eNARX produce higher accuracy in comparison to other techniques of NARX and BPTT. Consequently, the proposed technique provides more effective results for the prediction of commercial serial crime

Human-Robots Interaction by Facial Expression Recognition

This paper presents a facial expressions recognition system to command both mobile and arm robot. The proposed system mainly consists of two modules: facial expressions recognition and robots command. The first module aims to extract the ROI (Region Of Interest like: mouth, eyes, eyebrow) using Gradient Vector Flow (GVF) snake segmentation and the Euclidian distance calculation (compatible with the MPEG-4 description of the six universal emotions). To preserve the temporal aspect of the processing from FEEDTUM database (video file), Time Delay Neural Network (TDNN) is used as classifier of the universal facial expressions such as happiness, sadness, surprise, anger, fear, disgust and neutral. While the second module, analyzes recognized facial expressions and translates them into a language to communicate with robots by establishing command law.

Machine Learning Adaptive Computational Capacity Prediction for Dynamic Resource Management in C-RAN

Efficient computational resource management in 5G Cloud Radio Access Network (C-RAN) environments is a challenging problem because it has to account simultaneously for throughput, latency, power efficiency, and optimization tradeoffs. The assumption of a fixed computational capacity at the baseband unit (BBU) pools may result in underutilized or oversubscribed resources, thus affecting the overall Quality of Service (QoS). As resources are virtualized at the BBU pools, they could be dynamically instantiated according to the required computational capacity (RCC). In this paper, a new strategy for Dynamic Resource Management with Adaptive Computational capacity (DRM-AC) using machine learning (ML) techniques is proposed. Three ML algorithms have been tested to select the best predicting approach: support vector machine (SVM), time-delay neural network (TDNN), and long short-term memory (LSTM). DRM-AC reduces the average of unused resources by 96 %, but there is still QoS degradation when RCC is higher than the predicted computational capacity (PCC). To further improve, two new strategies are proposed and tested in a realistic scenario: DRM-AC with pre-filtering (DRM-AC-PF) and DRM-AC with error shifting (DRM-AC-ES), reducing the average of unsatisfied resources by 98 % and 99.9 % compared to the DRM-AC, respectively.

Analysis of the Effects of Supraglottal Tract Surgical Procedures in Automatic Speaker Recognition Performance

This article evaluates the impact in the performance of state-of-the-art automatic speaker recognition schemes of three surgical procedures modifying the supraglottal tract structures of speakers. To do so, a new corpus (Cuco) was recorded, containing the speech of 107 speakers before and after surgery. Speakers were divided into four groups depending on the type of surgery: tonsillectomy, functional endoscopy sinus surgery (FESS), septoplasty, and controls. The analyzed speaker recognition schemes were i-vectors, i-vectors with supervised Universal Background Model, i-vectors employing Time-delay Deep Neural Networks and x-vectors. In all cases, probabilistic linear discriminant analysis was employed in the back-end. Results show changes in the speech of patients who underwent tonsillectomy or FESS after surgery in contrast to controls or patients who had a septoplasty, where not significant variations are observed. These changes increase the Equal Error Rate (EER) of the analyzed speaker recognition schemes for the septoplasty and FESS groups when employing enrollment data recorded before the surgery. Moreover, surgery has a similar influence in the speech of female and male speakers with respect to the analyzed schemes. In consequence, results suggest that it is advisable to update the speaker's enrollment speech after three months following supraglottal tract surgery to ensure that the effects of the operation and post-operative recovery period do not influence the performance of the automatic speaker recognition systems.

Rider model identification using dynamic neural networks

Abstract Car driver modeling is a well-known research topic, with significant existing contributions. In contrast, important questions related to motorcyclist modeling remain unanswered. This study focuses on identifying a motorcyclist model that can predict the steering angle and the rider roll angle. A black box rider model in the form of a time delay neural network is presented. This model was developed using experimental data recorded with an instrumented motorcycle from the VIROLO++ research project. It is used for three main issues. First, the selection of input signals and their impact on prediction performance is discussed. Next, the model’s ability to predict the behavior of a variety of motorcyclists is demonstrated. Finally, the nonlinearity of the model is analyzed. These results pave the way to the development of a cybernetic rider model.

A Modified Maximal Divergence Sequential Auto-Encoder and Time Delay Neural Network Models for Vulnerable Binary Codes Detection

Since the risks associated with software vulnerabilities are rapidly increasing, the detection of vulnerabilities in binary code has become an important area of concern for the software community. However, research studies associated with the detection of vulnerabilities in binary code remain limited to the handcrafted features referenced by a specific group of experts in the field. This paper considers other possibilities to add on the subject of detecting vulnerabilities in binary code. Herein, we utilize recent studies conducted on the topic of deep learning and specifically study a maximal divergence sequential auto-encoder (MDSAE) model to propose a modified version (MDSAE-NR). We also propose an altered interpretation of time-delay neural network (TDNN-NR) by incorporating a new regularization technique that produced optimized results. Finally, both models achieved good predictive performance using different evaluation metrics such as accuracy, recall, precision and F1 score compared to the baseline results. Based on the results of our experiments, we observed a 2 to 2.5% average improvement in each performance measure of interest.

Efficient Digital Predistortion Using Sparse Neural Network

This paper proposes an efficient neural-network-based digital predistortion (DPD), named as envelope time-delay neural network (ETDNN) DPD. The method complies with the physical characteristics of radio-frequency (RF) power amplifiers (PAs) and uses a more compact DPD model than the conventional neural-network-based DPD. Additionally, a structured pruning technique is presented and used to reduce the computational complexity. It is shown that the resulting ETDNN obtained after applying pruning becomes so sparse that its complexity is comparable to that of conventional DPDs such as memory polynomial(MP) and generalized memory polynomial (GMP), while the degradation in performance due to the pruning is negligible. In an experiment on a 3.5-GHz GaN Doherty power amplifier (PA), our method with the proposed pruning had only one-thirtieth the computational complexity of the conventional neural-network-based DPD for the same adjacent channel leakage ratio (ACLR). Moreover, compared with memory-polynomial-based digital predistortion, our method with the proposed pruning achieved a 7-dB improvement in ACLR, despite it having lower computational complexity.

Distributed Synchronization of Coupled Time-Delay Neural Networks Based on Randomly Occurring Control

In this paper, the distributed synchronization of stochastic coupled neural networks with time-varying delay is concerned via randomly occurring control. We use two Bernoulli stochastic variables to describe the occurrence of distributed adaptive control and updating law according to certain probabilities. The distributed adaptive control and updating law for each vertex in the network depend on the state information on each vertex’s neighborhood. Based on Lyapunov stability theory, It&#244 differential equations, etc., by constructing the appropriate Lyapunov functional, we study and obtain sufficient conditions for the distributed synchronization of such networks in mean square.

Fractional Order PID Control Law for Trajectory Tracking Using Fractional Order Time-Delay Recurrent Neural Networks for Fractional Order Complex Dynamical Systems

In this paper presents an extension to the theory on the analysisof stability of complex systems with time delay of fractionalorder, the previous study is based on the theory of functionsof Lyapunov-Krasovski and the Fractional Order PID controllaw. The analytical results are illustrated by the simulation offractional nonlinear systems interconnected with time delaywhich are forced to follow complex trajectories as chaoticsystems.Fractional complex dynamical systems, trajectory track-ing, Fractional Order Lyapunov-Krasovskii theory, FractionalOrder PID control law.

Trajectory tracking error using fractional order time-delay recurrent neural networks using Krasovskii-Lur’e functional for Chua’s circuit via inverse optimal control

This paper presents an application of a Fractional Order Time Delay Neural Networks to chaos synchronization. The two main methodologies, on which the approach is based, are fractional order time-delay recurrent neural networks and the fractional order inverse optimal control for nonlinear systems. The problem of trajectory tracking is studied, based on the fractional order Lyapunov-Krasovskii and Lur’e theory, that achieves the global asymptotic stability of the tracking error between a delayed recurrent neural network and a reference function is obtained. The method is illustrated for the synchronization, the analytic results we present a trajectory tracking simulation of a fractional order time-delay dynamical network and the Fractional Order Chua’s circuits

Continuous and recurrent pattern dynamic neural networks recognition of electrophysiological signals

In the last few years, recurrent and continuous algorithms have became key factors in the solution of diverse pattern recognition problems. The main goal of this study is to introduce four classes of recurrent and continuous artificial neural networks (ANN) that can be implemented for pattern recognition of electrophysiological signals. Such networks are generally known as dynamic neural networks (DNN). The proposed DNN based pattern recognizer uses biosignals raw data as input. This processing method allows capturing the signal time dynamics, which is considered as an intrinsic characteristic of physiology signals. Therefore, recurrent and differential ANN structures were developed to construct different versions of dynamic automatic pattern recognizer. The first one describes the application of Recurrent Neural Networks (RNN) to enforce the biosignal analysis which evolves over time with a fixed sampling period. Three different DNNs with continuous dynamics are introduced. Differential neural network (DifNN) with the capability of learning the evolution of the signal in continuous time, a time-delay neural network (TDNN) for classification is implemented to consider the time-delayed characteristics of the electrophysiological signals and a complex valued neural network (CVNN) which considered the signals to be classified may be pre-processed with a frequency analysis technique. Two different databases of diverse physiological signals are used in this study to validate the application of dynamic neural networks. A first database considers electromiographic (EMG) signals which are tested using the DifNN, TDNN and CVNN. The second database includes gait in Parkinson's disease database signals which are used in the evaluation procedure of RNN. Two validation methods are used to justify the application of dynamic ANNs as pattern recognizer for the EMG activities and the health level classification of patients suffering from Parkinson's: generalization-regularization and the k-fold cross validation. The accuracy estimation and the confusion matrix evaluation confirm the superiority of the proposed approach compared to classical feed-forward ANN pattern recognizer. The particular case of the RNN is also implemented in a 32-bits micro-controller embedded device.

Estimation of tire–road friction for road vehicles: a time delay neural network approach

The performance of vehicle active safety systems is dependent on the friction force arising from the contact of tires and the road surface. Therefore, an adequate knowledge of the tire–road friction coefficient is of great importance to achieve a good performance of different vehicle control systems. This paper deals with the tire–road friction coefficient estimation problem through the knowledge of lateral tire force. A time delay neural network (TDNN) is adopted for the proposed estimation design. The TDNN aims at detecting road friction coefficient under lateral force excitations avoiding the use of standard mathematical tire models, which may provide a more efficient method with robust results. Moreover, the approach is able to estimate the road friction at each wheel independently, instead of using lumped axle models simplifications. Simulations based on a realistic vehicle model are carried out on different road surfaces and driving maneuvers to verify the effectiveness of the proposed estimation method. The results are compared with a classical approach, a model-based method modeled as a nonlinear regression.

Electroencephalogram based communication system for locked in state person using mentally spelled tasks with optimized network model

Due to growth in population, Individual persons with disabilities are increasing daily. To overcome the disability especially in Locked in State (LIS) due to Spinal Cord Injury (SCI), we planned to design four states moving robot from four imagery tasks signals acquired from three electrode systems by placing the electrodes in three positions namely T1, T3 and FP1. At the time of the study we extract the features from Continuous Wavelet Transform (CWT) and trained with Optimized Neural Network model to analyze the features. The proposed network model showed the highest performances with an accuracy of 93.86 % then that of conventional network model. To confirm the performances we conduct offline test. The offline test also proved that new network model recognizing accuracy was higher than the conventional network model with recognizing accuracy of 97.50 %. To verify our result we conducted Information Transfer Rate (ITR), from this analysis we concluded that optimized network model outperforms the other network models like conventional ordinary Feed Forward Neural Network, Time Delay Neural Network and Elman Neural Networks with an accuracy of 21.67 bits per sec. By analyzing classification performances, recognizing accuracy and Information Transformation Rate (ITR), we concluded that CWT features with optimized neural network model performances were comparably greater than that of normal or conventional neural network model and also the study proved that performances of male subjects was appreciated compared to female subjects.

Periodicity and stabilization control of the delayed Filippov system with perturbation

By employing Leray-Schauder alternative theorem of set-valued maps and non-Lyapunov method (non-smooth analysis, inequality analysis, matrix theory), this paper investigates the problems of periodicity and stabilization for time-delayed Filippov system with perturbation. Several criteria are obtained to ensure the existence of periodic solution of time-delayed Filippov system by using differential inclusion. By designing appropriate switching state-feedback controller, the asymptotic stabilization and exponential stabilization control of Filippov system are realized. Applying these criteria and control design method to a class of time-delayed neural networks with perturbation and discontinuous activation functions under a periodic environment. The developed theorems improve the existing results and their effectiveness are demonstrated by numerical example.

Signal identification system for developing rehabilitative device using deep learning algorithms.

Paralyzed patients were increasing day by day. Some of the neurodegenerative diseases like amyotrophic lateral sclerosis, Brainstem Leison, Stupor and Muscular dystrophy affect the muscle movements in the body. The affected persons were unable to migrate. To overcome from their problem they need some assistive technology with the help of bio signals. Electrooculogram (EOG) based Human Computer Interaction (HCI) is one of the technique used in recent days to overcome such problem. In this paper we clearly check the possibilities of creating nine states HCI by our proposed method. Signals were captured through five electrodes placed on the subjects face around the eyes. These signals were amplified with ADT26 bio amplifier, filtered with notch filter, and processed with reference power and band power techniques to extract features to detect the eye movements and mapped with Time Delay Neural Network to classify the eye movements to generate control signal to control external hardware devices. Our experimental study reports that maximum average classification of 91.09% for reference power feature and 91.55%-for band power feature respectively. The obtained result confirms that band power features with TDNN network models shows better performance than reference features for all subjects. From this outcome we conclude that band power features with TDNN network models was more suitable for classifying the eleven difference eye movements for individual subjects. To validate the result obtained from this method we categorize the subjects in age wise to check the accuracy of the system. Single trail analysis was conducted in offline to identify the recognizing accuracy of the proposed system. The result summarize that band power features with TDNN network models exceed the reference power with TDNN network model used in this study. Through the outcome we conclude that that band power features with TDNN network was more suitable for designing EOG based HCI in offline mode.

Deep Scattering Spectra with Deep Neural Networks for Acoustic Scene Classification Tasks

As one of the most commonly used features, Mel-frequency cepstral coefficients (MFCCs) are less discriminative at high frequency. A novel technique, known as Deep scattering spectrum (DSS), addresses this issue and looks to preserve greater details. DSS feature has shown promise both on classification and recognition tasks. In this paper, we extend the use of DSS feature for acoustic scene classification task. Results on Detection and classification of acoustic scenes and events (DCASE) 2016 and 2017 show that DSS provided 4:8% and 17:4% relative improvements in accuracy over MFCC features, within a state-of-the-art time delay neural network framework.

Using a Time Delay Neural Network Approach to Diagnose the Out-of-Control Signals for a Multivariate Normal Process with Variance Shifts

With the rapid development of advanced sensor technologies, it has become popular to monitor multiple quality variables for a manufacturing process. Consequently, multivariate statistical process control (MSPC) charts have been commonly used for monitoring multivariate processes. The primary function of MSPC charts is to trigger an out-of-control signal when faults occur in a process. However, because two or more quality variables are involved in a multivariate process, it is very difficult to diagnose which one or which combination of quality variables is responsible for the MSPC signal. Though some statistical decomposition methods may provide possible solutions, the mathematical difficulty could confine the applications. This study presents a time delay neural network (TDNN) classifier to diagnose the quality variables that cause out-of-control signals for a multivariate normal process (MNP) with variance shifts. To demonstrate the effectiveness of our proposed approach, a series of simulated experiments were conducted. The results were compared with artificial neural network (ANN), support vector machine (SVM) and multivariate adaptive regression splines (MARS) classifiers. It was found that the proposed TDNN classifier was able to accurately recognize the contributors of out-of-control signal for MNPs.

Finite-time boundedness for time-delay neural networks with external disturbances via weight learning approaches

This paper addresses the issue of finite-time boundedness for time-delay neural networks with external disturbances via weight learning. With the help of a group of inequalities and combining with the Lyapunov theory, weight learning rules are devised to ensure the neural networks to be finite-time bounded for the fixed connection weight matrix case and the fixed delayed connection weight matrix case, respectively. Sufficient conditions on the existence of the desired learning rules are presented in the form of linear matrix inequalities, which are easily verified by MATLAB software. It is shown that the proposed learning rules also guarantee the finite-time stability of the time-delay neural networks. Finally, a numerical example is employed to show the applicability of the devised weight learning rules.

ANFIS for Tamil Phoneme Classification

Phoneme recognition is an intricate problem lying under non-linear systems. Most research in this area revolve around try to model the pattern of features observed in the speech spectra with the use of Hidden Markov Models (HMM), various types of neural networks like deep recurrent neural networks, time delay neural networks, etc. for efficient phoneme recognition. In this paper, we study the effectiveness of the hybrid architecture, the Adaptive Neuro-Fuzzy Inference System (ANFIS) for capturing the spectral features of the speech signal to handle the problem of Phoneme Recognition. In spite of a wide range of research in this field, here we examine the power of ANFIS for least explored Tamil phoneme recognition problem. The experimental results have shown the ability of the model to learn the patterns associated with various phonetic classes, indicated with recognition improvement in terms of accuracy to its counterparts.

The Usefulness of Imperfect Speech Data for ASR Development in Low-Resource Languages

When the National Centre for Human Language Technology (NCHLT) Speech corpus was released, it created various opportunities for speech technology development in the 11 official, but critically under-resourced, languages of South Africa. Since then, the substantial improvements in acoustic modeling that deep architectures achieved for well-resourced languages ushered in a new data requirement: their development requires hundreds of hours of speech. A suitable strategy for the enlargement of speech resources for the South African languages is therefore required. The first possibility was to look for data that has already been collected but has not been included in an existing corpus. Additional data was collected during the NCHLT project that was not included in the official corpus: it only contains a curated, but limited subset of the data. In this paper, we first analyze the additional resources that could be harvested from the auxiliary NCHLT data. We also measure the effect of this data on acoustic modeling. The analysis incorporates recent factorized time-delay neural networks (TDNN-F). These models significantly reduce phone error rates for all languages. In addition, data augmentation and cross-corpus validation experiments for a number of the datasets illustrate the utility of the auxiliary NCHLT data.