Search

The main limitation to obstacle avoidance nowadays has been negative road anomalies which is the term we used to refer to potholes and cracks due to their negative drop from the surface of the road. This has for long been a limitation because of the fact that they exist in different, random and stochastic shapes. Today’s technology lacks the presence of sensors capable of detecting negative road anomalies efficiently as the latter surpasses the sensor’s limitations rendering the sensing technique inaccurate. A significant amount of research has been focused on the detection of negative road anomalies due to the fact that this topic is becoming a hot research topic. In this paper, the existing techniques will be reviewed. Their limitations will be highlighted and they will be assessed via certain performance indicators and via some chosen criteria which will be introduced.

This paper investigates the use of system memory features as the characteristics of simpler embedded devices. The ICmetrics technology is based on the hardware and software features of the systems in order to generate the unique identifiers. Usually memory features are considered very volatile as they are considered very unstable usually. But, in this case the memory features are very crucial as they tell us more about the system function. Hence, these features were used for analysis and a multivariate gaussian model was created and validated by comparing testing sample against the training data with 98% accuracy amongst 6 devices. This confirms the usefulness of the system memory features in adding a layer of security to the devices.

In this paper, we present a novel methodology based on machine learning for identifying the most appropriate from a set of available state-of-the-art object detectors for a given application. Our particular interest is to develop a road map for identifying verifiably optimal selections, especially for challenging applications such as detecting small objects in a mixed-size object dataset. State-of-the-art object detection systems often find the localisation of small-size objects challenging since most are usually trained on large-size objects. These contain abundant information as they occupy a large number of pixels relative to the total image size. This fact is normally exploited by the model during training and inference processes. To dissect and understand this process, our approach systematically examines detectors’ performances using two very distinct deep convolutional networks. The first is the single-stage YOLO V3 and the second is the double-stage Faster R-CNN. Specifically, our proposed method explores and visually illustrates the impact of feature extraction layers, number of anchor boxes, data augmentation, etc., utilising ideas from the field of explainable Artificial Intelligence (XAI). Our results, for example, show that multi-head YOLO V3 detectors trained using augmented data produce better performance even with a fewer number of anchor boxes. Moreover, robustness regarding the detector’s ability to explain how a specific decision was reached is investigated using different explanation techniques. Finally, two new visualisation techniques are proposed, WS-Grad and Concat-Grad, for identifying explanation cues of different detectors. These are applied to specific object detection tasks to illustrate their reliability and transparency with respect to the decision process. It is shown that the proposed techniques can result in high resolution and comprehensive heatmaps of the image areas, significantly affecting detector decisions as compared to the state-of-the-art techniques tested.

Presentation attack artefacts can be used to subvert the operation of biometric systems by being presented to the sensors of such systems. In this work, we propose the use of visual stimuli with randomised trajectories to stimulate eye movements for the detection of such spoofing attacks. The presentation of a moving visual challenge is used to ensure that some pupillary motion is stimulated and then captured with a camera. Various types of challenge trajectories are explored on different planar geometries representing prospective devices where the challenge could be presented to users. To evaluate the system, photo, 2D mask and 3D mask attack artefacts were used and pupillary movement data were captured from 80 volunteers performing genuine and spoofing attempts. The results support the potential of the proposed features for the detection of biometric presentation attacks.

Deep learning plays a vital role in classifying different arrhythmias using electrocardiography (ECG) data. Nevertheless, training deep learning models normally requires a large amount of data and can lead to privacy concerns. Unfortunately, a large amount of healthcare data cannot be easily collected from a single silo. Additionally, deep learning models are like black-box, with no explainability of the predicted results, which is often required in clinical healthcare. This limits the application of deep learning in real-world health systems. In this paper, to address the above-mentioned challenges, we design a novel end-to-end framework in a federated setting for ECG-based healthcare using explainable artificial intelligence (XAI) and deep convolutional neural networks (CNN). The federated setting is used to solve challenges such as data availability and privacy concerns. Furthermore, the proposed framework effectively classifies different arrhythmias using an autoencoder and a classifier, both based on a CNN. Additionally, we propose an XAI-based module on top of the proposed classifier for interpretability of the classification results, which helps clinical practitioners to interpret the predictions of the classifier and to make quick and reliable decisions. The proposed framework was trained and tested using the baseline Massachusetts Institute of Technology - Boston’s Beth Israel Hospital (MIT-BIH) Arrhythmia database. The trained classifier outperformed existing work by achieving accuracy up to 94.5% and 98.9% for arrhythmia detection using noisy and clean data, respectively, with five-fold cross-validation. We also propose a new communication cost reduction method to reduce the communication costs and to enhance the privacy of users’ data in the federated setting. While the proposed framework was tested and validated for ECG classification, it is general enough to be extended to many other healthcare applications.

As the adoption of digital technologies expands, it becomes vital to build trust and confidence in the integrity of such technology. The SPIRIT project investigates the proof of concept of employing novel secure and privacy-ensuring techniques in services set-up in the Internet of Things (IoT) environment, aiming to increase the trust of users in IoTbased systems. The proposed system integrates three highly novel technology concepts developed by the consortium partners. Specifically, a technology, termed ICMetrics, for deriving encryption keys directly from the operating characteristics of digital devices; secondly, a technology based on a contentbased signature of user data in order to ensure the integrity of sent data upon arrival; a third technology, termed semantic firewall, which is able to allow or deny the transmission of data derived from an IoT device according to the information contained within the data and the information gathered about the requester.

Individuals relying on wheelchairs for mobility are subject to the risk of injury due to their exposure to whole-body vibrations for long periods of time as per ISO 2631-1. Our study evaluates the feasibility of integrating ride dynamics measurements (i.e. vertical accelerations) as expressions of user travel comfort assessment to smart robotic wheelchairs. This will also help to mitigate the injury risk caused by the continuous exposure to vibrations using real time electronic measurement systems in order to ensure that the wheelchair’s movement dynamics (acceleration and speed) and the user’s comfort is adapted to the surrounding environment, specifically the type of ground surface type, as per the ISO standard mentioned previously.

CAPTCHAs are security mechanisms that try to prevent automated abuse of computer services. Many CAPTCHAs have been proposed but most have known security flaws against advanced attacks. In order to avoid a kind of oracle attacks in which the attacker learns about ground truth labels via active interactions with the CAPTCHA service as an oracle, Kwon and Cha proposed a new CAPTCHA scheme that employ uncertainties and trap images to generate adaptive CAPTCHA challenges, which we call “Uncertainty and Trap Strengthened CAPTCHA” (UTS-CAPTCHA) in this paper. Adaptive CAPTCHA challenges are used widely (either explicitly or implicitly) but the role of such adaptive mechanisms in the security of CAPTCHAs has received little attention from researchers. In this paper we present a statistical fundamental design flaw of UTS-CAPTCHA. This flaw leaks information regarding ground truth labels of images used. Exploiting this flaw, an attacker can use the UTS-CAPTCHA service as an oracle, and perform several different statistical learning-based attacks against UTS-CAPTCHA, increasing any reasonable initial success rate up to 100% according to our theoretical estimation and experimental simulations. Based on our proposed attacks, we discuss how the fundamental idea behind our attacks may be generalized to attack other CAPTCHA schemes and propose a new principle and a number of concrete guidelines for designing new CAPTCHA schemes in the future.

Research on iris recognition have observed that iris texture has inherent radial correlation. However, currently, there lacks a deeper insight into iris textural correlation. Few research focus on a quantitative and comprehensive analysis on this correlation. In this paper, we perform a quantitative analysis on iris textural correlation. We employ steering kernels to model the textural correlation in images. We conduct experiments on three benchmark datasets covering iris captures with varying quality. We find that the local textural correlation varies due to local characteristics in iris images, while the general trend of textural correlation goes along the radial direction. Moreover, we demonstrate that the information on iris textural correlation can be utilized to improve iris recognition. We employ this information to produce iris codes. We show that the iris code with the information on textural correlation achieves an improved performance compared to traditional iris codes.

INTRODUCTION: Many online services use data-sharing nudges to solicit personal data from their customers for personalized services.\ud \ud OBJECTIVES: This study aims to study people’s privacy preferences in sharing different types of personal data under different nudging conditions, how digital nudging can change their data sharing willingness, and if people’s data sharing preferences can be predicted using their responses to a questionnaire.\ud \ud METHODS: This paper reports a machine learning-based analysis on people’s privacy preference patterns under four different data-sharing nudging conditions (without nudging, monetary incentives, non-monetary incentives, and privacy assurance). The analysis is based on data collected from 685 UK residents who participated in a panel survey. Their self-reported willingness levels towards sharing 23 different types of personal data were analyzed by using both unsupervised (clustering) and supervised (classification) machine learning algorithms.\ud \ud RESULTS: The results led to a better understanding of people’s privacy preference patterns across different data-sharing nudging conditions, e.g., our participants’ preferences are distributed in a space of 48 possible profiles more sparsely than we expected, and the unexpected observation that all the three data-sharing nudging strategies led to an overall negative effect: they led to a reduced level of self-reported willingness for more participants, comparing with the case of no nudging at all. Our experiments with supervised machine learning models also showed that people’s privacy (data-sharing) preference profiles can be automatically predicted with a good accuracy, even when a small questionnaire with just seven questions is used.\ud \ud CONCLUSION: Our work revealed a more complicated structure of people’s privacy preference profiles, which have some dependencies on the type of data nudging and the type of personal data shared. Such complicated privacy preference profiles can be effectively analyzed using machine learning methods, including automatic prediction based on a small questionnaire. The negative results on the overall effect of different data-sharing nudges imply that service providers should consider if and how to use such mechanisms to incentivise their consumers to share personal data. We believe that more consumer-centric and transparent methods and tools should be used to help improve trust between consumers and service providers.

Assistive robotic applications require systems capable of interaction in the human world, a workspace which is highly dynamic and not always predictable. Mobile assistive devices face the additional and complex problem of when and if intervention should occur, therefore before any trajectory assistance is given, the robotic device must know where it is in real-time, without unnecessary disruption or delay to the user requirements. In this paper, we demonstrate a novel robust method for determining room identification from floor features in a real-time computational frame for autonomous and assistive robotics in the human environment. We utilize two inexpensive sensors: an optical mouse sensor for straightforward and rapid, texture or pattern sampling, and a four color photodiode light sensor for fast color determination. We show how data relating floor texture and color obtained from typical dynamic human environments, using these two sensors, compares favorably with data obtained from a standard webcam. We show that suitable data can be extracted from these two sensors at a rate 16 times faster than a standard webcam, and that these data are in a form which can be rapidly processed using readily available classification techniques, suitable for real-time system application. We achieved a 95% correct classification accuracy identifying 133 rooms’ flooring from 35 classes, suitable for fast coarse global room localization application, boundary crossing detection, and additionally some degree of surface type identification.

Powered wheelchair users may find operation in enclosed environments such as buildings difficult, a fundamental problem exists: wheelchairs are not much narrower than the doorway they wish to pass through. The ability to detect and pass through doorways represents a major current challenge for automated guided wheelchairs. We utilize a simple doorway pattern recognition technique for fast processing in a real-time system for robotic wheelchair users. We are able to show a 96% detection and identification of 5 individual doorways and an 86% recognition rate of 22 separate approach angles and translations. We conclude that pattern recognition using features obtained from simple constrained infrared ranging sensor data binning can be utilized for fast identification of doorways, and important coarse position and approach angle determination, suitable for real-time trajectory adjustment, representing a significant enhancement in this area.

Mobile robotic localisation obtained from simple sensor data potentially offers real-time real-world integration. Computationally highly efficient Weightless Neural Networks, when used for location determination, further enhances performance potential. This paper introduces techniques for the identification of rooms or locations in the absence of complex and succinct information. Using simple floor colour and texture, and room geometrics from ranging data, although inherent uncertainties exist, these limited simple fused real-time sensor data can be easily resolved into a room identification criterion using architectures generated by a Genetic Algorithm technique applied to a Weightless Neural Network Architecture.

Human assistive devices need to be effective with real-time assistance in real world situations: powered wheelchair users require reassuring robust support, especially in the area of collision avoidance. However, it is important that the intelligent system does not take away control from the user. The patient must be allowed to provide the intelligence in the system and the assistive technology must be engineered to be sufficiently smart to recognize and accommodate this. Robotic assistance employed in the healthcare arena must therefore emphasize positive support rather than adopting an intrusive role. Weightless Neural Networks are an excellent pattern recognition tool for real-time applications. This paper introduces a technique for look-ahead identification of open doorways and junctions. Simple sensor data in real-time is used to detect open doors with inherent data uncertainties using a technique applied to a Weightless Neural Network Architecture.

Among numerous pattern recognition methods the neural network approach has been the subject of much research due to its ability to learn from a given collection of representative examples. This paper is concerned with the design of a Weightless Neural Network, which decomposes a given pattern into several sets of n points, termed n-tuples. Considerable research has shown that by optimising the input connection mapping of such n-tuple networks classification performance can be improved significantly. This paper investigates the hybridisation of Genetic Algorithm (GA) and Particle Swarm Optimisation (PSO) techniques in search of better connection maps to the N-tuples. Experiments were conducted to evaluate the proposed method by applying the trained classifier to recognise hand-printed digits from a widely used database compiled by U.S. National Institute of Standards and Technology (NIST).

With the increasing incidence and prevalence of chronic brain injury patients and the current financial constraints in healthcare budgets, there is a need for a more intelligent way to realise the current practice of neuro-rehabilitation service provision. Brain-computer Interface (BCI) systems have the potential to address this issue to a certain extent only if carefully designed research can demonstrate that these systems are accurate, safe, cost-effective, are able to increase patient/carer satisfaction and enhance their quality of life. Therefore, one of the objectives of the proposed study was to examine whether participants (patients with brain injury and a sample of reference population) were able to use a low cost BCI system (Emotiv EPOC) to interact with a computer and to communicate via spelling words. Patients participated in the study did not have prior experience in using BCI headsets so as to measure the user experience in the first-exposure to BCI training. To measure emotional arousal of participants we used an ElectroDermal Activity Sensor (Qsensor by Affectiva). For the signal processing and feature extraction of imagery controls the Cognitive Suite of Emotiv's Control Panel was used. Our study reports the key findings based on data obtained from a group of patients and a sample reference population and presents the implications for the design and development of a BCI system for communication and control. The study also evaluates the performance of the system when used practically in context of an acute clinical environment.

Efficient localisation is a highly desirable property for an autonomous navigation system. Weightless neural networks offer a real-time approach to robotics applications by reducing hardware and software requirements for pattern recognition techniques. Such networks offer the potential for objects, structures, routes and locations to be easily identified and maps constructed from fused limited sensor data as information becomes available. We show that in the absence of concise and complex information, localisation can be obtained using simple algorithms from data with inherent uncertainties using a combination of Genetic Algorithm techniques applied to a Weightless Neural Architecture.