Pedestrian tracking, as a sub-task of object tracking, is widely applied in smart surveillance. However, the accuracy of object re-identification significantly decreases when objects encounter prolonged occlusion or undergo significant appearance changes. To address this issue, PTNet, a regression-based object tracking framework, is proposed in this paper. Firstly, a combination of DeepSORT and FastReid models is employed to resolve the problem of target re-identification after being occluded for a period of time. Secondly, a self-attention-based feature extraction model named BoTNet, incorporating the Global Multi-Head Self-Attention (MHSA), is integrated into the backbone network to capture global and long-distance features of targets in surveillance videos. Through ablation experiments, it was found that PTNet enhances accuracy by 4.8% compared to the benchmark. When compared to state-of-the-art object detection algorithms, PTNet demonstrates a significant improvement in performance.
The Abstract Alzheimer's Dementia (AD) presents significant diagnostic challenges, particularly in terms of early detection, where traditional methods often fall short due to their invasiveness and high costs. This study introduces a novel, non-invasive approach utilising emotional expressions captured from audio recordings to detect AD. Employing advanced digital signal processing techniques, including Facebook's Denoiser model, and deep learning methodologies through models such as Wav2Vec 2.0, this research aims to identify emotional disturbances that precede cognitive decline. Audio recordings were transformed into a tabular format, suitable for machine learning analysis. The LGBM Classifier and ensemble methods demonstrated superior performance, with the LGBM Classifier achieving the highest F1 score of 0.93 and an accuracy of 0.89 on a 3.5-second segment. These findings underscore the potential of combining emotional analysis with machine learning to enhance early AD detection, offering a simpler, more accessible diagnostic tool than currently available methods.
This study explores the potential of acoustic features extracted from speech recordings for detecting Alzheimer's Dementia (AD), employing a comprehensive approach that incorporates binary classification (healthy control vs. dementia), multiclass classification (healthy control, mild cognitive impairment, AD), and regression analyses (predicting MMSE scores). Additionally, demographic information of the participants was integrated to enhance the models' predictive accuracy. Our methodology involved processing each dataset version through a series of machine learning models tailored to each task, starting with a baseline version, followed by hyperparameter optimisation, and finally applying a combination of preprocessing steps (scaling, outlier removal, dimensionality reduction, and skewness correction) to identify the optimal setup for each model. The findings indicate that preprocessing steps significantly improve model performance across all tasks, underscoring the importance of data preparation in machine learning workflows for healthcare applications. Notably, the use of acoustic data alone for AD detection shows promising results, suggesting a pathway toward more generalised approaches that could incorporate recordings in various languages without linguistic dependency. This opens up the possibility for scalable, non-invasive screening tools for AD, leveraging the universal nature of acoustic markers in speech for early detection and monitoring of this condition.
Learning from Demonstration (LfD) techniques are invaluable for capturing complex human behaviors for robotic arm manipulations, yet they frequently encounter challenges such as avoiding singularities and respecting joint limits when directly applied to robotic systems. These challenges often lead to mechanical non-compliance, inaccuracies, and increased computational demands during control method adjustments. To address these issues, this study introduces a robust approach that integrates Damped Least Squares Inverse Kinematics (DLS-IK) with Probabilistic Movement Primitives (ProMPs). By leveraging DLS-IK to generate kinematically feasible trajectories, and embedding these within the ProMPs framework, our method not only ensures mechanical compliance but also capitalizes on the probabilistic modeling capabilities of ProMPs. This synergy addresses a significant gap in traditional LfD applications—aligning human demonstrations with the mechanical constraints of robotics, independent of the demonstrator’s expertise. Our integrated approach refines the LfD process, enabling the generation of precise, reliable, and mechanically compliant movements in robotic arms, thereby reducing the typical inaccuracies and computational burdens associated with conventional LfD methods.
This paper reviews the impacts of user behavior, dwelling characteristics, and social networks on energy savings within the UK housing sector. The study highlights the role of strategic energy planning, emphasising stakeholder engagement and the potential of social networks to influence energy-saving practices. It shows that integrating behavioral insights with architectural design is beneficial to enhancing energy use efficiency and reducing energy cost. By examining the interplay between human behavior and technology, a holistic approach to energy conservation is advocated integrating technological innovations and social impacts that drive sustainable energy usage.
For the first-generation missiles that are still widely applied on modern battlefields, the point-source infrared decoy is an economical and effective jamming alternative. To improve the survival probability of target, it is essential to analyze the jamming effect deeply. In this paper, based on the target acquisition principle of spin-scan seeker, the mathematical expression of error signal is derived. By constructing a complete missile guidance loop, the jamming effect of the point-source infrared decoy for the first-generation missile is analyzed with different jamming parameters, including radiation power and separation velocity. The simulation results indicate that a decoy with higher radiation power intensity than that of target will leads to successful jamming. Moreover, increasing the separation velocity can enlarge the miss distance. The jamming effect analysis will give some suggestions for the target to deploy the point-source infrared decoy for successful evasion.
Integrating intelligent robots into healthcare environments promises to revolutionise patient care, diagnosis, treatment, and rehabilitation. This paper reviews the current state of Human-Robot Collaboration (HRC) in healthcare, highlighting significant advancements and challenges. It provides a comprehensive overview of critical terminologies and concepts, reviews the latest robotics technologies, and explores their potential to enhance healthcare delivery. Our review reveals substantial opportunities for robots to augment healthcare professionals, improve patient monitoring, and manage repetitive or physically demanding tasks. We also discuss the critical challenges of ensuring patient safety, maintaining privacy and data security, addressing ethical considerations, and overcoming technical limitations such as interoperability and adaptability in diverse healthcare settings. By examining recent studies and innovations, this paper identifies emerging trends and suggests future directions for research and development in HRC within the healthcare sector.
With the development of robotics, the new flexible robotic arm has slowly entered people’s daily lives, helping or even replacing humans in various environments to carry out tasks with unique advantages on safety and flexibility. To increase the flexibility of robotic arm in grasping and achieve a dynamic and wider grasp range, a flexible gripper with inflatable fingers and inflatable palm is designed, simulated and tested. To support the operation of this flexible gripper, the corresponding pneumatic drive control through the solenoid valve has been designed. Simulation and experimental results show that the proposed flexible gripper is capable of grasp objects under diverse geometries.
Recommendation systems rely on an accurate user model to understand users’ needs to make a personal recommendation. Traditional user modeling uses users’ past behaviors during a "supply-meets-demand" interaction. This approach failed to capture the dynamic and emergence of new items and the shifting of user interests. The recommendation systems, built based on this user model trap users in their previous interests and make recommendations without counting their interest shift. We propose a new approach that integrates a non-stationary transformer into a recommendation system to capture the temporal dynamics of supplies and shifting user interests. Our experiments demonstrate the framework’s superiority over benchmark models. The empirical results confirm the efficacy of our proposed framework and significant performance enhancements for recommendations.
This paper presents an effective online process fault diagnosis method by integrating recurrence plots (RP) with convolutional neural networks (CNN). To cope with the high dimension of process data, principal component analysis (PCA) is applied to the original process data. RPs are then produced using the major principal components (PCs). As RPs are symmetric, this paper proposes to merge the two RPs for the first and second PCs into one to represent more information. The merged RPs serve as the inputs to a CNN, which is trained for fault diagnosis. The proposed fault diagnosis method is demonstrated on a simulated continuous stirred tank reactor (CSTR) system. It is shown that the proposed fault diagnosis system gives enhanced diagnosis performance.
Offshore wind turbines, with their larger rotors designed to boost power capacity, contend with increased aerodynamic and operational loads. These heightened loads often result in mechanical damage and decreased efficiency. Furthermore, because of variations in wind speed across the rotor, each blade experiences unique loads. These loads can cause fatigue and vibration in the blades, which can impair the blades' performance. The rotor blades' angle of attack with respect to the wind is adjusted by a well-designed pitch controller to minimise these problems and maximise power production. This paper addresses the differences between onshore and offshore wind turbines, nonlinear characteristics of wind turbines and lowers blade loads by presenting an innovative pitch angle controlling approach using fuzzy logic. Three controllers are used to create and assess a mathematical model of the pitch controlling system: PID, Fuzzy and Fuzzy- PID. Simulation results indicate that the Fuzzy-PID controller surpasses other strategies in mitigating uncertainties and disturbances in offshore wind turbine systems. Compared to conventional PID control, the Fuzzy-PID hybrid system demonstrates notable improvements, including a reduced rise time of 0.489 seconds, settling time of 4.327 seconds, overshoot of 6.989%, undershoot of 1.59%, and zero steady-state error. This innovative approach holds promise for enhancing offshore wind turbine efficiency, ensuring sustainable energy production in the face of escalating demands.
Extended Reality systems, such as Augmented Reality (AR), are expected to play a crucial role in future industries, particularly by empowering users into Industry 5.0. In this context, the adoption of Head-Mounted AR displays (HMD) based on new developments in video see-through (VST) technology indicates significant potential in the manufacturing sector. However, the literature on this technology’s usage in maintenance operations is still limited. This research proposes an original user study to investigate the cognitive workload and usability issues of an AR VST HMD in assisting the maintenance of a Programmable Logic Controller (PLC) system. The results suggest that participants did not perceive a high mental workload during task execution. Although the new technology still has some visualisation and interaction issues, most users consider these minor problems. This study highlights the potential of video see-through HMDs for empowering engineers and operators in advanced manufacturing and provides significant new insights regarding technical challenges to be addressed, considerations for effective implementation and suggestions for future research.
Collaborative robotic configurations for monitoring and tracking human beings for safety and efficiency have attracted interest in industrial revolution. The fusion of different types of sensors embedded in collaborative robotic systems significantly improve robotic perception. However, current methods have not deeply explored the capabilities of multi-sensory configurations including visible and thermal sensors. In this paper, we propose a contactless multi-sensor fusion including visible and thermal dual camera for collaborative robots to improve the robotic perception for human safety. Remote photoplethysmography detection and infrared thermal camera were used to measure the heart rate and body temperature.
The global shift towards energy-efficient electric vehicles (EVs) has prompted the research community to investigate efficiency improvements, enabling the increasing success of the EV industry. It is vital to utilise advancements in control to implement software updates for existing hardware, to improve the efficiency of the Interior Permanent Magnet Synchronous Motor (IPMSM) powertrain. Maximum Torque Per Ampere (MTPA) control is currently the benchmark for the EV powertrain providing the minimum copper loss per ampere. However, Maximum Efficiency Per Ampere (MEPA) considers minimisation of both copper loss and core loss of the IPMSM. Current MEPA methods are particularly parameter dependent; therefore, the implementation of parameter estimators is essential to guarantee optimal tracking of operating points to provide optimal efficiency. The proposed approach implements a gradient descent core loss resistance estimator in order to accurately determine the optimal operational point. The methodology is verified using simulations that present the results throughout the operational range.
Considering the face as a vital and most informative portion of the human body, it reflects different high-level information about an individual. This high-level information includes Age, Gender, and Emotion. Facial muscles’ shape and movement can be the best descriptors for the automatic extraction of these high-level facial features. Detection of these high-level features has applications in different areas including entertainment, surveillance, multimedia, and educational training. However, with the varying nature of these features, it becomes difficult to capture one class with the variability of other classes. This article presents a lightweight heterogeneous neural network with one shared backbone and three network heads to predict multiple face features: landmarks, age, and gender. The proposed system (MultiHeadCNN) captures these high-level facial features in the wild with extreme face pose, occlusions, and lightening conditions. The system is capable of predicting one type of feature with different variability of other types: predicting gender for different age groups and vice versa. The system is tested on comprehensive (UTKFace) and complex (Adience) datasets with varying age, gender, pose, and lightening conditions. The experiment shows promising results in terms of accuracy, with results for age and gender detection on the UTKFace and Adience datasets being 99.9%, 99.7%, 90.3%, and 61.7%, respectively. Furthermore, the parallel inference speed is 20 frames per second.
This paper proposes a new bearing fault diagnosis model (PCIDCNN) based on multi-source information fusion with principal component analysis and improved 1DCNN model to achieve the diagnosis of bearing faults under the operating conditions of alternating loads. The proposed model improves the multi-sensor data fusion ability and feature learning ability to solve the problem of information overload and noise interference of multi-source information during bearing operation. The bearings are the key component of rotating machinery. It is crucial to make timely and accurate fault diagnosis on bearings for the reliability and safety. PCA is employed to fuse signals from multiple sensors to obtain the fused data. We propose an improved 1DCNN model combining the attention mechanism and fused pooling layer to capture important fault features adequately. Experimental results based on real datasets show that the proposed method is able to analyze and diagnose the bearing fault signals, accurately identify different fault types, with obvious advantages over traditional machine learning models, achieving the diagnosis precision rate of 96.33%.
Freeform surfaces are widely used in advanced manufacturing due to their versatility, yet their complex geometries make characterization challenging. To do a surface texture analysis, it is often necessary to first remove the form of the surface. When the nominal form is unknown, the reference form can be approximated by filtration. If a triangular mesh represents the freeform surface, mesh smoothing is a type of filtration. Among various methods, Laplacian smoothing is the most employed for mesh smoothing due to its linear simplicity. Traditional Laplacian smoothing techniques use uniform, distance, or cotangent weights to approximate the discrete Laplace-Beltrami operator for each vertex. However, Laplacian smoothing has limitations, such as accuracy, shrinkage effects, and edge blurring. This study first investigates and compares Gaussian convolution and Laplacian smoothing techniques, aiming to integrate their advantages. Based on these findings, a novel approach is proposed to enhance Laplacian mesh smoothing by introducing a bilateral weighting scheme for each vertex. This enhancement keeps the simplicity of the Laplacian smoothing structure while mitigating the limitations of traditional weighting functions, offering improved performance in terms of accuracy and efficiency. The proposed method demonstrates potential for better estimating reference forms for freeform surfaces, which is the first step for surface characterization in advanced manufacturing.
This paper presents an innovative approach for fall detection, a significant concern in elder care, using vision-based techniques and video analysis. By employing and comparing supervised machine learning algorithms for recognising falls, The paper examines the impact of different environmental conditions on the fall detection system, focusing on illumination, an aspect previously overlooked in the field. The study introduces a vision-based fall detection method using Human Pose Estimation (HPE) models, specifically MoveNet, for feature extraction from human gestures and temporal moving features. Selected machine learning algorithms and neural network models are then trained and compared using these features to recognise video events such as falls and non-falls. The presented results show promising 70.6% accuracy and real-time model efficiency. This study’s findings hold significant potential for enhancing timely fall detection in real-world scenarios.
Aiming at dynamic scheduling in multi-train timetable, a optimization model of multi-train timetable and speed profile optimization for energy conservation is proposed. The proposed approach integrates train operation trajectory optimization with the scheduling of arrival and departure times at stations. At each instance, the highest-priority train operation process is selected, and optimization is performed for the operation of only one train between stations. Various constraints are considered to ensure safe train operation and prevent conflicts with preceding trains. By collaboratively regulating and optimizing train arrival/ departure time, speed profile, traction/braking force, minimizing train traction energy consumption and minimizing power peak can be achieved. The objectives are combined into a single-objective nonlinear optimization problem through weighted summation, which is then solved using the differential evolution algorithm. The simulation verification conducted on the Yizhuang Line demonstrates that this method enables more efficient utilization of energy while also recovery from the delay. The train traction energy consumption was reduced by 11.35%. This work provides an important technical support and guidance for realizing the green, low-carbon and sustainable development of urban rail transit in the future.
In alignment with the UK's ambitious target to achieve zero-carbon power generation by 2050, this study explores the integration of Hybrid Renewable Energy Systems (HRES) within grid-connected microgrids. This research addresses the techno-economic viability and power quality impacts of HRES to promote sustainable energy solutions. Employing a robust simulation framework using HOMERPro and MATLAB, the study meticulously evaluates HRES configurations to determine the most cost-effective and efficient system design. The findings in this research underscore their potential in enhancing grid stability, reducing carbon emissions, and contributing economically via lower energy costs. Results indicate significant improvements in financial metrics, achieving a Levelised Cost of Energy (LCOE) of 0.157 pound/kWh, showcasing its economic advantage against conventional energy sources. Additionally, the optimized system demonstrated a Total Harmonic Distortion (THD) level below 5%, ensuring compliance with stringent power quality standards. This research not only supports the UK's energy transition goals but also provides a replicable model for similar energy systems worldwide, highlighting the critical role of integrated renewable solutions in achieving global decarbonization objectives.
