Person identification is a pivotal challenge in the security domain, with important and impactful applications such as identifying crime suspects and locating missing persons. One convenient person identification method is gait identification, where individuals are identified by their unique walking style. However, traditional methods of gait identification are often affected by variations in appearance and occlusion. This work introduces a novel and robust spatiotemporal kinematics‐informed non‐invasive gait identification (STONI‐GID) method that uses human pose estimation, occlusion state estimation and deep machine learning. Furthermore, unlike some existing methods, we demonstrate that our method remains unaffected by everyday appearance changes, environment, or viewing angle. Our approach achieved identification accuracy of up to 98.66% when evaluated using our primary dataset of 65 diverse participants in real‐world environments. Moreover, the model outperformed existing methods during cross‐dataset validation on the large Southampton dataset and the Gait Recognition Image and Depth Dataset (GRIDDS), achieving identification accuracies of 97.68% and 99.12%, respectively. Our findings will particularly advance the research frontiers of real‐world gait identification and impact interdisciplinary areas of security and healthcare applications.

Knowledge about infant language can go a long way in supporting parents, nurses, and care providers in improving babies' health conditions. Crying is the most effective tool through which babies convey their requirements. In this work, several studies dealing with infant cry detection and classification are contrasted. Research demonstrates that machine learning techniques can effectively categorize and classify infant needs and certain disorders. Several datasets, including Baby Chillanto, Donate A Cry Corpus and Dunstan Baby Language, are presented. After reviewing existing Datasets, preprocessing methodologies and audio feature extraction such as MFCC, RMS energy, etc., are discussed. For infant cry detection and classification, several algorithms, such as support vector machines (SVM), convolutional neural networks (CNN), k-nearest neighbors (KNN), Random Forest, etc., have been analyzed and utilized for such processes in general. Finally, the study explores various applications of infant cry analysis, highlighting its potential to improve infant care and facilitate early diagnosis. As a result of the findings, it has been observed that infant cry analysis can effectively identify different needs and potential health concerns with high accuracy. These machine learning models' classification outputs have the potential to (1) improve childcare practices, (2) detect medical issues earlier, and (3) monitor infants continuously. These features give medical professionals and caregivers useful information for prompt intervention. The implementation of these findings can be applied in hospitals, neonatal intensive care units (NICUs), smart baby monitoring systems, and research studies focused on early childhood development.

This paper presents a sign language translation system developed to enhance accessibility for lesser-supported sign languages, with a specific focus on Indian Sign Language. Given that India has the world's largest deaf population, the communication gap between sign language users and non-users poses a significant barrier. The objective of this research is to examine existing support systems for Indian Sign Language and to build a system capable of detecting, recognizing, and translating signed words through the use of connected cameras. To achieve this goal, image classification using transfer learning techniques was applied to a small dataset of Indian Sign Language gestures, resulting in an average recognition accuracy rate of 83%. Over 10,000 images across 12 different gestures were collected, and the training/validation process reached peak testing accuracy levels of 95%. The final model was deployed into a web-based application specifically designed for use in areas with limited access to advanced technology, aiming to reduce communication barriers and improve inclusivity for the deaf and hard-of-hearing community.

Teacher retention is a significant international problem. Up to half of current teachers suggest they may leave the profession in the near future, and half of newly qualified teachers leave within the first five years of joining the profession. This unsustainable retention problem threatens the stability of students' education worldwide. There are a variety of reasons that teachers state for influencing their departure from the profession. However, the most common, particularly for newly qualified teachers, is the feeling of burnout or being overworked. Artificial Intelligence (AI) presents opportunities to support teachers and to automate some of their work, thereby reducing their workload. However, such techniques use opaque methods, which often produce content that is not correct or credible. In this paper, we propose AskEFO, a chatbot that supports newly qualified teachers and answers queries based solely on input from approved qualified professionals. Evaluated by a cohort of 14 newly qualified English teachers, 79% stated they would use AskEFO instead of Google in the classroom, 93% stated that AskEFO would enhance their confidence when teaching and 100% indicated that it would positively impact their students.

Flamingos are known for their vibrant pink and reddish hues, which are not merely aesthetic but indicative of their overall health and diet. These colors are derived from carotenoid pigments in their food sources, making coloration a vital marker for monitoring their well-being and environmental conditions. This study introduces a two-stage classification methodology designed to safeguard flamingo populations by leveraging deep learning techniques. Convolutional Neural Networks (CNNs) are used for both shape and color detection, ensuring accurate identification of flamingos and insights into their health status. Simulation results demonstrated the CNNs model’s effectiveness, making it a valuable resource for wildlife conservation efforts aimed at preserving flamingo habitats. The first stage employs digital classification filters to distinguish flamingo images from other species, achieving an accuracy of 97.52%, while the second stage refines these detections through color analysis with an accuracy of 86.27%. This approach promises to mark a significant advancement in wildlife conservation, offering reliable methods for assessing and managing flamingo populations in their natural environments.

Water issues related to Iraqi rivers and reservoirs have devastating consequences on the citizens’ health and country’s economy. This has been seen in the increased number of non-communicable diseases and jobs’ shortage in the country as-a-whole. This paper discusses the factors contributing to water scarcity in Iraq and proposes intelligent solutions. Factors contributing to water issues are not limited to extreme weather conditions, and include water scarcity, evaporation, water contamination and water salinity. What makes the situation more complicated is the outdated water systems that have been used for monitoring, analysis, irrigation, and control of water. In addition, human factors play important role in decision-making in terms of use of these systems. Uses of these systems is not an easy task as it requires well-established standard operating procedures and comprehensive/integrated databases; where, both are lacking in Iraq. The paper is envisaged to implement an Intelligent Water System that utilises the latter technologies for trying to solve all the challenges within the existing water management system. Implementing such intelligent water system and intelligent barriers will benefit Iraq on the short and long-term with specific aids to the environment, economy, citizens and the country as-a-whole. Thus, the proposed intelligent water system will result in cleaner air and water, better quality crop and land, and less pollution. Citizens will have better health and quality of life and more job opportunities. In addition, the country will have new jobs and industries, smarter and more sustainable agricultures, self-sustainable smart cities, self-sufficient economy, and healthier.

This study explores the application of deep learning techniques, specifically Convolutional Neural Networks, for the classification of dog breeds from images. By employing varying input image resolutions, the research evaluates the impact of resolution on the accuracy and efficiency of the model. Five experiments were conducted using resolutions of 64, 128, 224, 256, and 512 pixels to assess model performance. The results indicate that an input resolution of 256x256 pixels yields the highest accuracy, achieving 94.74% with an optimal balance between detail and processing complexity. However, certain breeds, such as the American Foxhound and Anatolian Shepherd Dog, exhibited lower classification performance, highlighting the importance of considering breed-specific characteristics in model development. The findings emphasize the critical role of image resolution in training deep learning models and suggest that a 256x256 resolution offers the best trade-off between accuracy and computational efficiency for dog breed classification.

Lower back pain (LBP) is a global medical condition that affects more than $\mathbf{8 0 \%}$ of people at least once in their lifetime. Various abnormal conditions could cause LBP. One of these conditions is Lumbar Spinal Stenosis (LSS), which is usually a serious condition, hence the need for its prompt diagnosis. The visual assessment process of magnetic resonance imaging (MRI) is expensive in terms of time and effort and prone to delays. The integration of an AI-enabled system to help clinicians in diagnosing patients with LBP is expected to alleviate the burden on radiologists and foster a less time-consuming and cost-efficient diagnostic process. In this paper, a convolutional neural network (CNN) model has been built to diagnose T2-weighted (T2W) axial MRI scans. The model detects LSS in these scans and has achieved remarkable accuracy and recall of $\mathbf{9 1 \%}$ and $\mathbf{9 6 \%}$, respectively. Explainable AI (XAI) using LIME’s ImageExplainer was also implemented to ensure the model offers explainable insights. This is to ensure the model is not only accurate but also reliable.

Eye tracking and head pose estimation (HPE) have previously lacked reliability, interpretability, and comprehensibility. For instance, many works rely on traditional computer vision methods, which may not perform well in dynamic and realistic environments. Recently, a widespread trend has emerged, leveraging deep learning for HPE specifically framed as a regression task; however, considering the real-time applications, the problem could be better formulated as classification (e.g., left, centre, right head pose and gaze) using a hybrid approach. For the first time, we present a complete facial profiling approach to extract micro and macro facial movement, gaze, and eye state features, which can be used for various applications related to comprehension analysis. The multi-model approach provides discrete human-understandable head pose estimations utilising deep transfer learning, a newly introduced method of head roll calculation, gaze estimation via iris detection, and eye state estimation (i.e. , open or closed). Unlike existing works, this approach can automatically analyse the input image or video frame to produce human-understandable binary codes (e.g., eye open or close, looking left or right, etc.) for each facial component ( aka face channels). The proposed approach is validated on multiple standard datasets, indicating outperformance compared to existing methods in several aspects, including reliability, generalisation, completeness, and interpretability. This work will significantly impact several diverse domains, including psychological and cognitive tasks with a broad scope of applications, such as in police interrogations and investigations, animal behaviour, and smart applications, including driver behaviour analysis, student attention measurement, and automated camera flashes.

The number of livestock farms and their sizes (particularly the sheep farms) are on the rise, in response to the growing demands of food supply chain for increasing population. The detection and monitoring of sheep activities particularly in huge farms is tedious and challenging task. Therefore, a reliable and cost-effective sheep activity detection system which can be utilized for the virtual fencing, is of high demand. Existing data-driven approaches use accelerometer data for sheep monitoring and activity detection however, there are several limitations with these methods such as generating high volume of data with noise, relatively expensive, and not very reliable. This study presents a non-invasive computer vision-based approach along with deep transfer learning for sheep detection and determining whether the corresponding state is ‘active’ or ‘inactive’. We complied a primary dataset comprising sheep in diverse poses and activities in a realistic outdoor environment. A custom YOLOV5s model is trained over new dataset and validated on purely unseen sheep instances for the model evaluation. The statistical outcomes demonstrate the robustness of proposed approach for various sheep activity detection. Our method has diverse implications and uses in the development of reliable and economical systems for monitoring sheep, particularly in extensive farms and virtual fencing applications.

ChatGPT is an extensive language model under the umbrella of generative artificial intelligence that produces answers from data and images curated from online resources. Despite the capability to produce accurate responses, but requires verification; the responses are based on statistical patterns rather than true comprehension, i.e., it does not have consciousness and does not understand the questions from the perspective of human comprehension. The ability of ChatGPT to understand and react to questions in a humanistic way has garnered a lot of public and scientific interest over the past year. This study analyzes responses of ChatGPT to 100 questions on epilepsy in order to assess the validity of the tool in this field. Besides, this work sheds light on the advantages and disadvantages of the approach in this particular topic by analyzing responses of ChatGPT to queries on epilepsy. The study evaluates the model performance by looking at the completeness, correctness, and relevancy of responses. The findings in this paper indicate that ChatGPT has limits because of its training data and design structure, even though it could give insightful and appropriate answers to inquiries about epilepsy. It is concluded that ChatGPT can be an advantageous tool for medical professionals working on the subject of epilepsy. Nonetheless, it should be noted that ChatGPT should be utilized cautiously and in conjunction with various information sources, like clinical practice guidelines and peer-reviewed studies.

Where traditional motorways contain hard shoulders to provide refuge for broken-down vehicles, smart motorways instead use live outer lane to ease congestion. The live lanes can be closed due to accidents or breakdowns which is communicated to other road users through overhead gantry signs. This can only occur if the traffic management control center is made aware of the stationary vehicle(s), through either notification via phone call or Motorway Incident Detection and Automatic Signaling (MIDAS) induction loop technology. Alternatively, radar-based stopped vehicle detection is used to identify non-moving objects on the highway. However, this technology is unable to recognize objects or distinguish between congestion and break down etc., which leads to generate false alarms. For the first time, we propose a fully autonomous computer vision and deep learning-based solution to detect stationary vehicles on highways and local roads. We employ deep transfer learning to build a custom-trained vehicle detection model using a newly prepared dataset comprising over 105,000 annotated vehicle instances. DeepSort algorithm is employed for real-time vehicle tracking through associating instances between time series frames, followed by a rule-based algorithm to identify the current state of detected vehicles. Experimental outcomes show our approach as outperforming the state-of-the-art methods in terms of efficient and reliable detection of stationary vehicles (with 98.3% accuracy) as well as distinguish them from congestions when evaluated over video streams captured in realistic dynamic and diverse conditions.

Animal-Vehicle Accidents have shown deep increase in the middle east regions over the last decades. These collisions resulting from camels fleeing the wildlife and crossing the roads and hence endangering drivers and camel's lives and leading to habitat degradation. Additionality, the size, strength, and the unpredictable behavior of camels play a key role in high mortality rates in the camel-vehicle collisions. Various solutions and countermeasures such as warning signs and fences have been adopted in the past. However, several drawbacks are associated to them, and their effectiveness are reducing with time. Therefore, this study proposes a framework for the use of machine learning approaches and computer vision for the detection and recognition of camels. This can help to provide warning to drivers about potential animal crossings in an effort to mitigate camel-vehicle accidents.

Liver segmentation and lesion detection in biomedical applications has gained noticeable attention due to the labor-intensive efforts required to manually segment regions of interest (e.g., liver, tumor) from CT scans for efficient and accurate diagnosis. Segmentation in this context has been a comprehensive area of development to assess liver disease, enabling physicians to evaluate liver volume, morphology, and the presence of lesions. Recent advancements in biomedical imaging and deep learning approaches have led to various efforts for automated segmentation at higher level of confidence proposing significant advantages with respect to efficiency and accuracy. Accordingly, an approach for automated liver segmentation and lesion detection was developed in this study utilizing a deep learning-based U-Net model. A novel approach to automatically segment the liver and detect hepatic lesions in medical images was proposed using a U-Net model, a popular deep learning architecture for biomedical image segmentation tasks. Model training and validation was performed based on a small image set of 20 IRCAD tomography bases, compiled from the IRCAD repository, hosted by the Institute for Research Against Digestive Cancer. Various preprocessing steps on each image slice were first performed (e.g., resizing, normalization) to handle potential model over/under-fitting and enhance its robustness against dynamic image modalities. Model performance with respect to the six selected performance metrics (i.e., Dice coefficient, sensitivity, specificity, Jaccard index, precision, F1-score) demonstrated excellent model performance of >95% accuracy when tested on a subset of purely unseen images from the test set. The analysis showed that such automated segmentation via deep learning can be achieved for customized datasets while providing significant advantages over traditional labor-intensive manual segmentation approaches.

This comprehensive review explores the landscape of assistive technology designed to support physically disabled adults. The paper encompasses a thorough examination of technological interventions aimed at enhancing the independence and quality of life for individuals facing physical disabilities. Focusing on a diverse range of assistive devices and systems, including mobility aids, communication tools, and adaptive interfaces, the review assesses their effectiveness, usability, and impact on daily living. It synthesizes current research findings, technological advancements, and user experiences, offering insights into the evolving field of assistive technology. Additionally, the review addresses challenges, potential future developments, and the role of emerging technologies in furthering the integration of assistive solutions. This synthesis contributes to a holistic understanding of the state-of-the-art in assistive technology for physically disabled adults, providing a valuable resource for researchers, practitioners, and policymakers working towards enhancing inclusivity and empowerment.

Predicting the criminals' behaviour is a difficult task to accomplish. It is unexpected in most cases and can possibly transpire at any time, which is challenging for police agencies and victims being affected by the offences. The proposed work presents a crime prediction model using the stop & search dataset and the demographic of those charged with possession of a weapon. The study is first of its kind using multiple publicly available datasets to predict the effectiveness of ‘stop & search’ interventions by the police. We employ multiple machine learning algorithms to predict whether a ‘further action’ is required following the stop & search by the police. We utilise several data science techniques mainly including pre-processing, feature engineering and appropriate use of model selection. The proposed model produced 93.20% accuracy using Random Forest classifier. The outcomes of this research can be useful by relevant authorities to anticipate the crime at a specific time and location through the analysis of patterns that will support decision-making and help on deterrent effective strategies to lower offences being committed.

Deception detection plays a vital role in various domains, from security and law enforcement to human behavior analysis. In this paper, we propose a comprehensive system for deception detection that leverages S&A smart sensing device, deep transfer learning, deep learning techniques, and explainable artificial intelligence. Our approach combines visual, auditory, thermal, cardiovascular, and respiratory cues, offering enhanced accuracy and resistance to countermeasures. Deep Transfer Learning is employed to adapt pre-trained models to the deception detection task, overcoming data limitations. Incorporating Explainable AI techniques enhances transparency and interpretability, fostering trust and collaboration in human-machine interactions. Our research lays the groundwork for future advancements in deception detection technology, addressing challenges and providing promising opportunities in the realm of deception detection.

Speech modality has recently been explored at a large scale with the advancement of computational capabilities, forensics applications, and the availability of big data. Automation of speaker identification via advanced machine learning approaches is gaining momentum to improve various applications, particularly forensics and information security. Due to the lack of large Arabic datasets, existing speaker identification models are trained on small amounts of speech inputs which is inefficient when applied in real-life. In the present study, an unsupervised Arabic speech embedding model is created, leveraging large amounts of unlabeled speech data to help streamline the identification process for various security areas. The proposed framework is divided into two sections: unsupervised speech feature extraction and supervised speaker identification. An unlabeled Arabic speech dataset was constructed with approximately 1 million samples to develop the deep feature embedding model. Various machine learning classifiers and a deep neural network were used for speaker identification on the labeled Emirati speech database, which contains utterances of eight traditional Emirati phrases. The K-nearest neighbor classifier achieved the best result with the speech embeddings for speaker identification with an average accuracy of 93.58% using five-fold cross-validation.

Outdoor mobility of individuals with visual impairment is challenging particularly where collision with obstacles can have significant impact on both physical and mental health. A variety of technological mobility aids for visually impaired people (VIP) have been studied and proposed in the literature which mainly utilise machine intelligence and deep learning (DL) approaches for object detection. However, object detection via the existing approaches suffers from reliability challenge due to real-time dynamics or the lack of available domain knowledge for specific obstacles identified by the VIP as potential hazards. In the present study, an object detection model (ObDtM) based on deep transfer learning techniques was developed for a custom-built dataset comprising of specific obstacles identified by the VIP as potential hazards. A custom dataset was compiled and manually annotated from various publicly available sources to train the ObDtM. Experiments were conducted to evaluate the proposed ObDtM for unseen obstacles kept as the test set. Results showed that ObDtM outperformed the state-of-the-art with 97% mean Average Precision (mAP), indicating a robust and generalizable DL approach. The compiled dataset and the ObDtM is useful for several potential use cases, particularly highlighting the use of DL in IoT and smart city applications. Additionally, a smart synergetic outdoor mobility framework was proposed for VIP (SOMAVIP) allowing comprehensive and accurate semantic representation of the surroundings by utilising the proposed ObDtM, cloud services, internet of things (IoT), and digital environment in the context of emerging smart city infrastructure. The proposed SOMAVIP can be highly impactful for improving VIPs’ quality of life mainly for safer, cost-effective, and reliable independent outdoor mobility enriched with real-time perception and interpretations of the surroundings.

Alzheimer's disease (AD) is a type of brain disorder that is regarded as a degenerative disease because the corresponding symptoms aggravate with the time progression. Single nucleotide polymorphisms (SNPs) have been identified as relevant biomarkers for this condition. This study aims to identify SNPs biomarkers associated with the AD in order to perform a reliable classification of AD. In contrast to existing related works, we utilize deep transfer learning with varying experimental analysis for reliable classification of AD. For this purpose, the convolutional neural networks (CNN) are firstly trained over the genome-wide association studies (GWAS) dataset requested from the AD neuroimaging initiative. We then employ the deep transfer learning for further training of our CNN (as base model) over a different AD GWAS dataset, to extract the final set of features. The extracted features are then fed into Support Vector Machine for classification of AD. Detailed experiments are performed using multiple datasets and varying experimental configurations. The statistical outcomes indicate an accuracy of 89% which is a significant improvement when benchmarked with existing related works.

Covid-19 vaccine hesitancy and acceptance delay is an unprecedented challenge for concerned authorities. Existing studies lack the investigation about public vaccination acceptance, specifically for children. In this study, we surveyed the adult population in the UK to determine the diversity in public perception and acceptance of Covid-19 vaccination specifically for the children, among different sociodemographic groups. Statistical results and intelligent clustering outcomes indicate significant relationships between sociodemographic diversity and vaccination acceptance for children and their families. Acceptability for children is significantly dependent on ethnicity$(\mathrm{p}=3.7\mathrm{e}-05)$, age group, and gender, where only 47% of participants show willingness towards children's vaccination. Primary dataset in this study, along with the experimental outcomes, might be useful for public awareness and policy makers towards better preparation for future epidemics as well as working globally to combat the ongoing Covid-19 variations while running effective vaccination campaigns in the identified sociodemographic groups.

To control and manage traffics as well as guide the driver on roads, the lines on the roads are used. In addition, these lines serve as barriers and to ensure the safe, smooth and harmonious flow of traffic. However, in some countries, these lines are missed and causes the driving chaos; and thus, car accidents happen. A car accident is one of the most causes of death and the majority of the accident are due to human error. This research works aims to help driver to provide safety roads and reduces or eliminate care accidents. Object detection is a technique used to find and locate objects in images. In this works, YOLO Version 3 is the network used to detect the object in the frame because of its speed, simplicity, and ability to predict as well as classify objects. In addition, a steering angle circuit is designed and implemented to measure the direction of the car. The steering angle measurements is used with object detection (vehicles and pedestrians) to issue a warning when these objects are close to the driving car (10 meters). After the objects are detected using YOLO V3, the distance of the detected objects is measured using the height of the object. While being affordable and low-cost, the system achieved positive and competitive results, this system can be used at night and in dark environments.

There is growing interest in the use of automated psychological profiling systems, specifically applying machine learning to the field of deception detection. Several psychological studies and machine-based models have been reporting the use of eye interaction, gaze and facial movements as important clues to deception detection. However, the identification of very specific and distinctive features is still required. For the first time, we investigate the fine-grained level eyes and facial micro-movements to identify the distinctive features that provide significant clues for the automated deception detection. A real-time deception detection approach was developed utilizing advanced computer vision and machine learning approaches to model the non-verbal deceptive behavior. Artificial neural networks, random forests and support vector machines were selected as base models for the data on the total of 262,000 discrete measurements with 1,26,291 and 128,735 of deceptive and truthful instances, respectively. The data set used in this study is part of an ongoing programme to collect a larger dataset on the effects of gender and ethnicity on deception detection. Some observations are made based on this data which should not be interpreted as scientific conclusions, but pointers for future work. Analysis of the above models revealed that eye movements carry relatively important clues to distinguish truthful and deceptive behaviours. The research outcomes align with the findings from forensic psychologists who also reported the eye movements as distinctive for the truthful and deceptive behavior. The research outcomes and proposed approach are beneficial for human experts and has many applications within interdisciplinary domains.

Advances in Information and Communications Technology (ICT) have increased the growth of Massive open online courses (MOOCs) applied in distance learning environments. Various tools have been utilized to deliver interactive content including pictures, figures, and videos that can motivate the learners to build new cognitive skills. High ranking universities have adopted MOOCs as an efficient dashboard platform where learners from around the world can participate in such courses. The students learning progress is evaluated by using set computer-marked assessments. In particular, the computer gives immediate feedback to the student once he or she completes the online assessments. The researchers claim that student success rate in an online course can be related to their performance at the previous session in addition to the level of engagement. Insufficient attention has been paid by literature to evaluate whether student performance and engagement in the prior assessments could affect student achievement in the next assessments. In this paper, two predictive models have been designed namely students' assessments grades and final students' performance. The models can be used to detect the factors that influence students' learning achievement in MOOCs. The result shows that both models gain feasible and accurate results. The lowest RSME gain by RF acquire a value of 8.131 for students assessments grades model while GBM yields the highest accuracy in final students' performance, an average value of 0.086 was achieved.

Gender bias within Artificial intelligence driven systems is currently a hot topic and is one of a number of areas where the data used to train, validate and test machine learning algorithms is under more scrutiny than ever before. In this paper we investigate if there is a difference between the nonverbal cues to deception generated by males and females through the use of an automated deception detection system. The system uses hierarchical neural networks to extract 36 channels of non-verbal head and facial behaviors whilst male and female participants are engaged in either a deceptive or truthful roleplaying task. An Image Vector dataset, comprising of 86584 vectors, is collated which uses a fixed sliding window slot of 1 second to record deceptive or truthful slots. Experiments were conducted on three variants of the dataset, all males, all females and mixed in order to examine if the differences in cues generated by males and females lead to differences in the accuracies of machine learning algorithms which classify their behavior. Results showed differences in nonverbal cues between males and females, with both genders at a disadvantage when treated by classifiers trained on both genders rather than classifiers specifically trained for each gender. However, there was no striking disadvantageous effect beyond the influence of their relative frequency of occurrence in the dataset.

Freezing of Gait (FoG) is a motor symptom of Parkinson's disease (PD) that frequently occurs in the long-term sufferers of the disease. FoG may result to nursing home admission as it can lead to falls, and therefore, it impacts negatively on the quality of life. The focus of this study is the systematic evaluation of machine learning techniques in conjunction with varying size time windows and time/frequency domain feature sets in predicting a FoG event before its onset. In the experiments, the Daphnet FoG dataset is used to benchmark performance. This consists of accelerometer signals obtained from sensors mounted on the ankle, thigh and trunk of the PD patients. The dataset is annotated with instances of normal activity events, and FoG events. To predict the onset of FoG, the dataset is augmented with an additional class, termed ‘transition’, which relates to a manually defined period prior to the occurrence of a FoG episode. In this research, five machine learning models are used, namely, Random Forest, Extreme Gradient Boosting, Gradient Boosting, Support Vector Machines using Radial Basis Functions, and Neural Networks. Support Vector Machines with Radial Basis kernels provided the best performance achieving sensitivity values of 72.34%, 91.49%, 75.00%, and specificity values of 87.36%, 88.51% and 93.62%, for the FoG, transition and normal activity classes, respectively.

Outdoor mobility of Visually Impaired People (VIPs) has always been challenging due to the dynamically varying scenes and environmental states. Variety of systems have been introduced to assist VIPs' mobility that include sensor mounted canes and use of machine intelligence. However, these systems are not reliable when used to navigate the VIPs in dynamically changing environments. The associated challenges are the robust sensing and avoiding diverse types of obstacles, dynamically modelling the changing environmental states (e.g. moving objects, road-works), and effective communication to interpret the environmental states and hazards. In this paper, we propose an intelligent wearable auditory display framework that will process real-time video and multi-sensor data streams to: a) identify the type of obstacles, b) recognize the surrounding scene/objects and corresponding attributes (e.g. geometry, size, shape, distance from user), c) automatically generate the descriptive information about the recognized obstacle/objects and attributes, d) produce accurate, precise and reliable spatial information and corresponding instructions in audio-visual form to assist and navigate VIPs safely with or without the assistance of traditional means.

In this paper an automatic deception detection system, which analyses participant deception risk scores from non-verbal behaviour captured during an interview conducted by an Avatar, is demonstrated. The system is built on a configuration of artificial neural networks, which are used to detect facial objects and extract non-verbal behaviour in the form of micro gestures over short periods of time. A set of empirical experiments was conducted based a typical airport security scenario of packing a suitcase. Data was collected through 30 participants participating in either a truthful or deceptive scenarios being interviewed by a machine based border guard Avatar. Promising results were achieved using raw unprocessed data on un-optimized classifier neural networks. These indicate that a machine based interviewing technique can elicit non-verbal interviewee behavior, which allows an automatic system to detect deception.

The Artificial Neural Network is generally considered to be an effective classifier, but also a “Black Box” component whose internal behavior cannot be understood by human users. This lack of transparency forms a barrier to acceptance in high-stakes applications by the general public. This paper investigates the use of a hybrid model comprising multiple artificial neural networks with a final C4.5 decision tree classifier to investigate the potential of explaining the classification decision through production rules. Two large datasets collected from comprehension studies are used to investigate the value of the C4.5 decision tree as the overall comprehension classifier in terms of accuracy and decision transparency. Empirical trials show that higher accuracies are achieved through using a decision tree classifier, but the significant tree size questions the rule transparency to a human.