A Multi-Feature Deep Learning Algorithm for Urban Traffic Classification with Limited Labeled Data

Acoustic sensors, if embedded in smart street lights, can help in capturing the activities (car honking, sirens, events, traffic, etc.) in cities. Needless to say, the acoustic data from such scenarios are complex due to multiple audio streams originating from different events, and when decomposed to independent signals, the amount of retrieved data volume is small in quantity which is inadequate to train deep neural networks. So, in this paper, we address the two challenges: a) separating the mixed signals, and b) developing an efficient acoustic classifier under data paucity. So, to address these challenges, we propose an architecture with supervised deep learning, where the initial captured mixed acoustics data are analyzed with Fast Fourier Transformation (FFT), followed by filtering the noise from the signal, and then decomposed to independent signals by fast independent component analysis (Fast ICA). To address the challenge of data paucity, we propose a multi feature-based deep neural network with high performance that is reflected in our experiments when compared to the conventional convolutional neural network (CNN) and multi-layer perceptron (MLP).

WebAppShield: An Approach Exploiting Machine Learning to Detect SQLi Attacks in an Application Layer in Run-Time

In recent years, SQL injection attacks have been identified as being prevalent against web applications. They affect network security and user data, which leads to a considerable loss of money and data every year. This paper presents the use of classification algorithms in machine learning using a method to classify the login data filtering inputs into "SQLi" or "Non-SQLi,” thus increasing the reliability and accuracy of results in terms of deciding whether an operation is an attack or a valid operation. A method as a Web-App is developed for auto-generated data replication to provide a twin of the targeted data structure. Shielding against SQLi attacks (WebAppShield) that verifies all users and prevents attackers (SQLi attacks) from entering and or accessing the database, which the machine learning module predicts as "Non-SQLi", has been developed. A special login form has been developed with a special instance of the data validation; this verification process secures the web application from its early stages. The system has been tested and validated, and up to 99% of SQLi attacks have been prevented.

Depth Camera Aided Dead-Reckoning Localization of Autonomous Mobile Robots in Unstructured Global Navigation Satellite System Denied Environments

In global navigation satellite system (GNSS) denied settings, such as indoor environments, autonomous mobile robots are often limited to dead-reckoning navigation techniques to determine their position, velocity, and attitude (PVA). Localization is typically accomplished by employing an inertial measurement unit (IMU), which, while precise in nature, accumulates errors rapidly and severely degrades the localization solution. Standard sensor fusion methods, such as Kalman filtering, aim to fuse precise IMU measurements with accurate aiding sensors to establish a precise and accurate solution. In indoor environments, where GNSS and no other a priori information is known about the environment, effective sensor fusion is difficult to achieve, as accurate aiding sensor choices are sparse. However, an opportunity arises by employing a depth camera in the indoor environment. A depth camera can capture point clouds of the surrounding floors and walls. Extracting attitude from these surfaces can serve as an accurate aiding source, which directly combats errors that arise due to gyroscope imperfections. This configuration for sensor fusion leads to a dramatic reduction of PVA error compared to traditional aiding sensor configurations. This paper provides the theoretical basis for the depth camera aiding sensor method, initial expectations of performance benefit via simulation, and hardware implementation thus verifying its veracity. Hardware implementation is performed on the Quanser Qbot 2™ mobile robot, with a Vector-Nav VN-200™ IMU and Kinect™ camera from Microsoft.

Adaptive Kalman Filter for Noise Estimation and Identification with Bayesian Approach

Bayesian approach can be used for parameter identification and extraction in state space models and its ability for analyzing sequence of data in dynamical system is proved in different literatures. In this paper, adaptive Kalman filter with Bayesian approach for identification of variances in measurement parameter noise is developed. Next, it is applied for estimation of the dynamical state and measurement data in discrete linear dynamical system. This algorithm at each step time estimates noise variance in measurement noise and state of system with Kalman filter. Next, approximation is designed at each step separately and consequently sufficient statistics of the state and noise variances are computed with a fixed-point iteration of an adaptive Kalman filter. Different simulations are applied for showing the influence of noise variance in measurement data on algorithm. Firstly, the effect of noise variance and its distribution on detection and identification performance is simulated in Kalman filter without Bayesian formulation. Then, simulation is applied to adaptive Kalman filter with the ability of noise variance tracking in measurement data. In these simulations, the influence of noise distribution of measurement data in each step is estimated, and true variance of data is obtained by algorithm and is compared in different scenarios. Afterwards, one typical modeling of nonlinear state space model with inducing noise measurement is simulated by this approach. Finally, the performance and the important limitations of this algorithm in these simulations are explained. 

Dimensionality Reduction in Modal Analysis for Structural Health Monitoring

Autonomous structural health monitoring (SHM) of many structures and bridges became a topic of paramount importance for maintenance purposes and safety reasons. This paper proposes a set of machine learning (ML) tools to perform automatic feature selection and detection of anomalies in a bridge from vibrational data and compare different feature extraction schemes to increase the accuracy and reduce the amount of data collected. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric data in both standard and damaged conditions. The proposed framework starts from the first four fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by time-domain filtering (tracking). The fundamental frequencies extracted are then fed to a dimensionality reduction block implemented through two different approaches: feature selection (intelligent multiplexer) that tries to estimate the most reliable frequencies based on the evaluation of some statistical features (i.e., entropy, variance, kurtosis), and feature extraction (auto-associative neural network (ANN)) that combine the fundamental frequencies to extract new damage sensitive features in a low dimensional feature space. Finally, one-class classification (OCC) algorithms perform anomaly detection, trained with standard condition points, and tested with normal and anomaly ones. In particular, principal component analysis (PCA), kernel principal component analysis (KPCA), and autoassociative neural network (ANN) are presented and their performance are compared. It is also shown that, by evaluating the correct features, the anomaly can be detected with accuracy and an F1 score greater than 95%.

Towards End-To-End Disease Prediction from Raw Metagenomic Data

Analysis of the human microbiome using metagenomic sequencing data has demonstrated high ability in discriminating various human diseases. Raw metagenomic sequencing data require multiple complex and computationally heavy bioinformatics steps prior to data analysis. Such data contain millions of short sequences read from the fragmented DNA sequences and stored as fastq files. Conventional processing pipelines consist in multiple steps including quality control, filtering, alignment of sequences against genomic catalogs (genes, species, taxonomic levels, functional pathways, etc.). These pipelines are complex to use, time consuming and rely on a large number of parameters that often provide variability and impact the estimation of the microbiome elements. Training Deep Neural Networks directly from raw sequencing data is a promising approach to bypass some of the challenges associated with mainstream bioinformatics pipelines. Most of these methods use the concept of word and sentence embeddings that create a meaningful and numerical representation of DNA sequences, while extracting features and reducing the dimensionality of the data. In this paper we present an end-to-end approach that classifies patients into disease groups directly from raw metagenomic reads: metagenome2vec. This approach is composed of four steps (i) generating a vocabulary of k-mers and learning their numerical embeddings; (ii) learning DNA sequence (read) embeddings; (iii) identifying the genome from which the sequence is most likely to come and (iv) training a multiple instance learning classifier which predicts the phenotype based on the vector representation of the raw data. An attention mechanism is applied in the network so that the model can be interpreted, assigning a weight to the influence of the prediction for each genome. Using two public real-life data-sets as well a simulated one, we demonstrated that this original approach reaches high performance, comparable with the state-of-the-art methods applied directly on processed data though mainstream bioinformatics workflows. These results are encouraging for this proof of concept work. We believe that with further dedication, the DNN models have the potential to surpass mainstream bioinformatics workflows in disease classification tasks.

A Structured Mechanism for Identifying Political Influencers on Social Media Platforms: Top 10 Saudi Political Twitter Users

Social media networks, such as Twitter, offer the perfect opportunity to either positively or negatively affect political attitudes on large audiences. The existence of influential users who have developed a reputation for their knowledge and experience of specific topics is a major factor contributing to this impact. Therefore, knowledge of the mechanisms to identify influential users on social media is vital for understanding their effect on their audience. The concept of the influential user is related to the concept of opinion leaders' to indicate that ideas first flow from mass media to opinion leaders and then to the rest of the population. Hence, the objective of this research was to provide reliable and accurate structural mechanisms to identify influential users, which could be applied to different platforms, places, and subjects. Twitter was selected as the platform of interest, and Saudi Arabia as the context for the investigation. These were selected because Saudi Arabia has a large number of Twitter users, some of whom are considerably active in setting agendas and disseminating ideas. The study considered the scientific methods that have been used to identify public opinion leaders before, utilizing metrics software on Twitter. The key findings propose multiple novel metrics to compare Twitter influencers, including the number of followers, social authority and the use of political hashtags, and four secondary filtering measures. Thus, using ratio and percentage calculations to classify the most influential users, Twitter accounts were filtered, analyzed and included. The structured approach is used as a mechanism to explore the top ten influencers on Twitter from the political domain in Saudi Arabia.

Fast Approximate Bayesian Contextual Cold Start Learning (FAB-COST)

Cold-start is a notoriously difficult problem which can occur in recommendation systems, and arises when there is insufficient information to draw inferences for users or items. To address this challenge, a contextual bandit algorithm – the Fast Approximate Bayesian Contextual Cold Start Learning algorithm (FAB-COST) – is proposed, which is designed to provide improved accuracy compared to the traditionally used Laplace approximation in the logistic contextual bandit, while controlling both algorithmic complexity and computational cost. To this end, FAB-COST uses a combination of two moment projection variational methods: Expectation Propagation (EP), which performs well at the cold start, but becomes slow as the amount of data increases; and Assumed Density Filtering (ADF), which has slower growth of computational cost with data size but requires more data to obtain an acceptable level of accuracy. By switching from EP to ADF when the dataset becomes large, it is able to exploit their complementary strengths. The empirical justification for FAB-COST is presented, and systematically compared to other approaches on simulated data. In a benchmark against the Laplace approximation on real data consisting of over 670, 000 impressions from autotrader.co.uk, FAB-COST demonstrates at one point increase of over 16% in user clicks. On the basis of these results, it is argued that FAB-COST is likely to be an attractive approach to cold-start recommendation systems in a variety of contexts.

Automatic Product Identification Based on Deep-Learning Theory in an Assembly Line

Automated object recognition and identification systems are widely used throughout the world, particularly in assembly lines, where they perform quality control and automatic part selection tasks. This article presents the design and implementation of an object recognition system in an assembly line. The proposed shapes-color recognition system is based on deep learning theory in a specially designed convolutional network architecture. The used methodology involve stages such as: image capturing, color filtering, location of object mass centers, horizontal and vertical object boundaries, and object clipping. Once the objects are cut out, they are sent to a convolutional neural network, which automatically identifies the type of figure. The identification system works in real-time. The implementation was done on a Raspberry Pi 3 system and on a Jetson-Nano device. The proposal is used in an assembly course of bachelor’s degree in industrial engineering. The results presented include studying the efficiency of the recognition and processing time.

Extracting Terrain Points from Airborne Laser Scanning Data in Densely Forested Areas

Airborne Laser Scanning (ALS) is one of the main technologies for generating high-resolution digital terrain models (DTMs). DTMs are crucial to several applications, such as topographic mapping, flood zone delineation, geographic information systems (GIS), hydrological modelling, spatial analysis, etc. Laser scanning system generates irregularly spaced three-dimensional cloud of points. Raw ALS data are mainly ground points (that represent the bare earth) and non-ground points (that represent buildings, trees, cars, etc.). Removing all the non-ground points from the raw data is referred to as filtering. Filtering heavily forested areas is considered a difficult and challenging task as the canopy stops laser pulses from reaching the terrain surface. This research presents an approach for removing non-ground points from raw ALS data in densely forested areas. Smoothing splines are exploited to interpolate and fit the noisy ALS data. The presented filter utilizes a weight function to allocate weights for each point of the data. Furthermore, unlike most of the methods, the presented filtering algorithm is designed to be automatic. Three different forested areas in the United Kingdom are used to assess the performance of the algorithm. The results show that the generated DTMs from the filtered data are accurate (when compared against reference terrain data) and the performance of the method is stable for all the heavily forested data samples. The average root mean square error (RMSE) value is 0.35 m.

Building and Tree Detection Using Multiscale Matched Filtering

In this study, an automated building and tree detection method is proposed using DSM data and true orthophoto image. A multiscale matched filtering is used on DSM data. Therefore, first watershed transform is applied. Then, Otsu’s thresholding method is used as an adaptive threshold to segment each watershed region. Detected objects are masked with NDVI to separate buildings and trees. The proposed method is able to detect buildings and trees without entering any elevation threshold. We tested our method on ISPRS semantic labeling dataset and obtained promising results.

A Character Detection Method for Ancient Yi Books Based on Connected Components and Regressive Character Segmentation

Character detection is an important issue for character recognition of ancient Yi books. The accuracy of detection directly affects the recognition effect of ancient Yi books. Considering the complex layout, the lack of standard typesetting and the mixed arrangement between images and texts, we propose a character detection method for ancient Yi books based on connected components and regressive character segmentation. First, the scanned images of ancient Yi books are preprocessed with nonlocal mean filtering, and then a modified local adaptive threshold binarization algorithm is used to obtain the binary images to segment the foreground and background for the images. Second, the non-text areas are removed by the method based on connected components. Finally, the single character in the ancient Yi books is segmented by our method. The experimental results show that the method can effectively separate the text areas and non-text areas for ancient Yi books and achieve higher accuracy and recall rate in the experiment of character detection, and effectively solve the problem of character detection and segmentation in character recognition of ancient books.

Linear Quadratic Gaussian/Loop Transfer Recover Control Flight Control on a Nonlinear Model

As part of the development of a 4D autopilot system for unmanned aerial vehicles (UAVs), i.e. a time-dependent robust trajectory generation and control algorithm, this work addresses the problem of optimal path control based on the flight sensors data output that may be unreliable due to noise on data acquisition and/or transmission under certain circumstances. Although several filtering methods, such as the Kalman-Bucy filter or the Linear Quadratic Gaussian/Loop Transfer Recover Control (LQG/LTR), are available, the utter complexity of the control system, together with the robustness and reliability required of such a system on a UAV for airworthiness certifiable autonomous flight, required the development of a proper robust filter for a nonlinear system, as a way of further mitigate errors propagation to the control system and improve its ,performance. As such, a nonlinear algorithm based upon the LQG/LTR, is validated through computational simulation testing, is proposed on this paper.

Make Up Flash: Web Application for the Improvement of Physical Appearance in Images Based on Recognition Methods

This paper presents a web application for the improvement of images through recognition. The web application is based on the analysis of picture-based recognition methods that allow an improvement on the physical appearance of people posting in social networks. The basis relies on the study of tools that can correct or improve some features of the face, with the help of a wide collection of user images taken as reference to build a facial profile. Automatic facial profiling can be achieved with a deeper study of the Object Detection Library. It was possible to improve the initial images with the help of MATLAB and its filtering functions. The user can have a direct interaction with the program and manually adjust his preferences.

All-Optical Function Based on Self-Similar Spectral Broadening for 2R Regeneration in High-Bit-Rate Optical Transmission Systems

In this paper, we demonstrate basic all-optical functions for 2R regeneration (Re-amplification and Re-shaping) based on self-similar spectral broadening in low normal dispersion and highly nonlinear fiber (ND-HNLF) to regenerate the signal through optical filtering including the transfer function characteristics, and output extinction ratio. Our approach of all-optical 2R regeneration is based on those of Mamyshev. The numerical study reveals the self-similar spectral broadening very effective for 2R all-optical regeneration; the proposed design presents high stability compared to a conventional regenerator using SPM broadening with reduction of the intensity fluctuations and improvement of the extinction ratio.

Optical Flow Based System for Cross Traffic Alert

This document describes an advanced system and methodology for Cross Traffic Alert (CTA), able to detect vehicles that move into the vehicle driving path from the left or right side. The camera is supposed to be not only on a vehicle still, e.g. at a traffic light or at an intersection, but also moving slowly, e.g. in a car park. In all of the aforementioned conditions, a driver’s short loss of concentration or distraction can easily lead to a serious accident. A valid support to avoid these kinds of car crashes is represented by the proposed system. It is an extension of our previous work, related to a clustering system, which only works on fixed cameras. Just a vanish point calculation and simple optical flow filtering, to eliminate motion vectors due to the car relative movement, is performed to let the system achieve high performances with different scenarios, cameras and resolutions. The proposed system just uses as input the optical flow, which is hardware implemented in the proposed platform and since the elaboration of the whole system is really speed and power consumption, it is inserted directly in the camera framework, allowing to execute all the processing in real-time.

Optimal Duty-Cycle Modulation Scheme for Analog-To-Digital Conversion Systems

This paper presents an optimal duty-cycle modulation (ODCM) scheme for analog-to-digital conversion (ADC) systems. The overall ODCM-Based ADC problem is decoupled into optimal DCM and digital filtering sub-problems, while taking into account constraints of mutual design parameters between the two. Using a set of three lemmas and four morphological theorems, the ODCM sub-problem is modelled as a nonlinear cost function with nonlinear constraints. Then, a weighted least pth norm of the error between ideal and predicted frequency responses is used as a cost function for the digital filtering sub-problem. In addition, MATLAB fmincon and MATLAB iirlnorm tools are used as optimal DCM and least pth norm solvers respectively. Furthermore, the virtual simulation scheme of an overall prototyping ODCM-based ADC system is implemented and well tested with the help of Simulink tool according to relevant set of design data, i.e., 3 KHz of modulating bandwidth, 172 KHz of maximum modulation frequency and 25 MHZ of sampling frequency. Finally, the results obtained and presented show that the ODCM-based ADC achieves under 3 KHz of modulating bandwidth: 57 dBc of SINAD (signal-to-noise and distorsion), 58 dB of SFDR (Surpious free dynamic range) -80 dBc of THD (total harmonic distorsion), and 10 bits of minimum resolution. These performance levels appear to be a great challenge within the class of oversampling ADC topologies, with 2nd order IIR (infinite impulse response) decimation filter.

Modeling of a UAV Longitudinal Dynamics through System Identification Technique

System identification of an Unmanned Aerial Vehicle (UAV), to acquire its mathematical model, is a significant step in the process of aircraft flight automation. The need for reliable mathematical model is an established requirement for autopilot design, flight simulator development, aircraft performance appraisal, analysis of aircraft modifications, preflight testing of prototype aircraft and investigation of fatigue life and stress distribution etc.  This research is aimed at system identification of a fixed wing UAV by means of specifically designed flight experiment. The purposely designed flight maneuvers were performed on the UAV and aircraft states were recorded during these flights. Acquired data were preprocessed for noise filtering and bias removal followed by parameter estimation of longitudinal dynamics transfer functions using MATLAB system identification toolbox. Black box identification based transfer function models, in response to elevator and throttle inputs, were estimated using least square error   technique. The identification results show a high confidence level and goodness of fit between the estimated model and actual aircraft response.

Object Detection in Digital Images under Non-Standardized Conditions Using Illumination and Shadow Filtering

In recent years, object detection has gained much attention and very encouraging research area in the field of computer vision. The robust object boundaries detection in an image is demanded in numerous applications of human computer interaction and automated surveillance systems. Many methods and approaches have been developed for automatic object detection in various fields, such as automotive, quality control management and environmental services. Inappropriately, to the best of our knowledge, object detection under illumination with shadow consideration has not been well solved yet. Furthermore, this problem is also one of the major hurdles to keeping an object detection method from the practical applications. This paper presents an approach to automatic object detection in images under non-standardized environmental conditions. A key challenge is how to detect the object, particularly under uneven illumination conditions. Image capturing conditions the algorithms need to consider a variety of possible environmental factors as the colour information, lightening and shadows varies from image to image. Existing methods mostly failed to produce the appropriate result due to variation in colour information, lightening effects, threshold specifications, histogram dependencies and colour ranges. To overcome these limitations we propose an object detection algorithm, with pre-processing methods, to reduce the interference caused by shadow and illumination effects without fixed parameters. We use the Y CrCb colour model without any specific colour ranges and predefined threshold values. The segmented object regions are further classified using morphological operations (Erosion and Dilation) and contours. Proposed approach applied on a large image data set acquired under various environmental conditions for wood stack detection. Experiments show the promising result of the proposed approach in comparison with existing methods.

Adaptive Filtering in Subbands for Supervised Source Separation

This paper investigates MIMO (Multiple-Input Multiple-Output) adaptive filtering techniques for the application of supervised source separation in the context of convolutive mixtures. From the observation that there is correlation among the signals of the different mixtures, an improvement in the NSAF (Normalized Subband Adaptive Filter) algorithm is proposed in order to accelerate its convergence rate. Simulation results with mixtures of speech signals in reverberant environments show the superior performance of the proposed algorithm with respect to the performances of the NLMS (Normalized Least-Mean-Square) and conventional NSAF, considering both the convergence speed and SIR (Signal-to-Interference Ratio) after convergence.