Over the last decade, video-enabled mobile devices have become almost ubiquitous, while advances in markerless pose estimation allow an individual’s body position to be tracked across the frames of a video. Previous work by this and other groups has shown that pose-extracted kinematic features can be used to reliably measure motor impairment in Parkinson’s disease (PD). This presents the prospect of developing an asynchronous and scalable, video-based assessment of motor dysfunction. Crucial to this endeavour is the ability to automatically recognise the class of an action being performed, without which manual labelling is required. Representing the evolution of body joint locations as a spatio-temporal graph, we implement a deep-learning model for frame-level classification of activities performed according to part 3 of the Movement Disorder Society Unified PD Rating Scale (MDS-UPDRS). We train and validate this system using a dataset of n=7220 video clips, recorded at 5 independent sites. This approach achieves human-level performance in classifying and labelling periods of activity within monocular video clips. Our framework could support clinical workflows and patient care at scale through applications such as quality monitoring of clinical data collection, automated labelling of video streams, or a module within a remote self-assessment system.

This paper is written to compare different classification models for sentiment analysis

Abstract— Although mobile robots have on-board sensors to  perform navigation, path planning can benefit from fixed, or  infrastructure, cameras that capture activity analytics  continuously and from non-robot perspectives. We describe an approach that collects statistics of motion that repeat at some  period (usually a day) and perform path planning to avoid  expected activity in time and space The same statistics are used to learn preferred human paths and plan robot paths on these at  times of low human activity. Temporal filtering is performed in  cascade to efficiently extract long- and short-term activity in two time dimensions (isochronal and chronological) for use in global  and local path planning. We compare our lightweight activity  detection approach to neural network object detection methods  and propose an activity-gated approach that combines activity and  object detection efficiently. We deployed our approach in the ROS  robot software development framework by augmenting the cost  map of static objects with dynamic regions determined from  activity. We describe benefits and constraints of this combination.

A document by Edern Ollivier . Click on the document to view its contents.

The Contrastive Learning for blind Hyperspectral Unmixing (CLHU) is a self-supervised deep learning approach for blind hyperspectral unmixing. Unlike existing deep learning methods that rely on reconstruction capabilities, CLHU leverages the input-endmembers relationship for abundance estimation. The endmembers are dynamically updated during training, controlled by two regularization factors. Extensive experiments demonstrate CLHU’s effectiveness, achieving state-of-the-art performance in hyperspectral unmixing. This novel approach offers a promising perspective for the field and holds potential for further enhancements in hyperspectral unmixing tasks.

With the rapid penetration of electric vehicles (EVs), the security of EV charging systems is becoming an important issue. This paper addresses cyber-physical protection of EV charging systems from the perspective of control theory. By using a residual generation technique, an attack-resilient EV charging system is designed, which detects false data injection under the denial of service (DOS) attack. The effectiveness of the proposed attack-resilient EV charging system is verified experimentally by using a hardware-in-the-loop testbed, developed by the IEEE 802.15.4 wireless sensors.

Arguments about the appropriate valuation of life in public policy have generally moved from accounting resource costs to revealed preference valuations. In medical and public health, the valuation of the ‘quality” life experience over time (a very different concept) has focused on equivalents to an unconstrained life experience. The user perspective on these valuations is largely missing and has substantial implications for both individuals and organizations. As these summary measures are fundamental to current triage and assessment processes, they have considerable importance for individuals. This highlights a number of different channels that can bias the assessment of individuals in specific circumstances when AI support is used, as a wide range of inferences from diverse data is then brought to bear and amplifies health data security scope and importance. These intermediate aggregate channel security vulnerabilities have not previously been highlighted, in spite of the importance attached to these summary measures in public health. The collision of totally individualized assumed impacts and end user consequences are amplified rapidly by Ai methods in support of decisions, and the security and privacy-if any left-of individual medical AND life data are thus jointly at cyber risk.

A document by Michael Hasselbeck . Click on the document to view its contents.

The development of sustainable and resilient networking and communication technologies, artificial intelligence, dispersed computing, and beyond 5G are expected to play a crucial role in the realization of Industry 5.0. More specifically, antennas are recognized as a vital technology to support the continuously growing demand for 5G/6G connectivity and data rates. In the development of wireless communication systems, an effective antenna design is critical. In this work, we present the design and modeling of a hexagonal microstrip antenna for using in ultra-wide band applications such as 5G mobile networks. The suggested antenna has a frequency range of 3.3 GHz to 4.5 GHz, dimensions of 20 mm × 14 mm, and is built on a FR-4 substrate, with a relative permittivity of 4.3, and of 1.6 mm in height. It is expected that this antenna will operate in two notched bands with resonance frequencies of 3.3 GHz and 4.5 GHz and a voltage standing wave ratio (VSWR) of less than 2. The proposed antenna, which has been expanded to be a multiple-input multiple-output (MIMO) antenna, was created and modeled using CST Microwave Studio. The comparison against the recent works is given to show the efficiency of the proposed design.

A computationally efficient framework to fingerprint real-world Bluetooth devices is presented in this work. Despite the active research in this topic, a generalizable framework suitable for real-world deployment in terms of performance in a new and evolving environment as well as hardware efficiency of the architecture is lacking. We propose an embedding-assisted attentional framework (Mbed-ATN) suitable for fingerprinting actual Bluetooth devices and analyze its generalization capability in different settings and demonstrate the effect of sample length and anti-aliasing decimation on its performance. The embedding module serves as a dimensionality reduction unit that maps the high dimensional 3D input tensor to a 1D feature vector for further processing by the ATN module. Furthermore, unlike the prior research in this field, we closely evaluate the complexity of the model and test its fingerprinting capability with real-world Bluetooth dataset collected under a different time frame and experimental setting while being trained on another. Our study reveals a 7.3x and 65.2x lesser memory usage with the proposed Mbed-ATN architecture in contrast to Oracle at input sample lengths of M=10 kS and M=100 kS respectively. Further, the proposed Mbed-ATN showcases a 16.9x fewer FLOPs and 7.5x lesser trainable parameters when compared to Oracle. Finally, we show that when subject to anti-aliasing decimation and at greater input sample lengths of 1 MS, the proposed Mbed-ATN framework results in a 5.32x higher TPR, 37.9 % fewer false alarms, and 6.74x higher accuracy under the challenging real-world setting.

Ultra-Wide Band (UWB) radios provide high-precision distance measurements, however are still prone to ranging errors due to the environment. For localization system design and testing, this leads to cost-intense field trials in which localization methods are validated. As an alternative, publicly available UWB datasets exist, which however are also limited to only certain environmental conditions. In order to provide a more general evaluation basis, we present and discuss a procedure which allows the synthetic generation of UWB distance measurements with respect to theoretical error sources and parameterization based on a conducted measurement campaign. This contribution builds an understanding of commonly present errors for wireless distance measurements and their effects on ranging residual distributions in the distance domain. In addition, we discuss the influences of non-line-of-sight and multipath conditions. Based on this work, distance measurements with respect to the multipath richness of the propagation environment can be generated, helping to flexibly evaluate positioning methods for various emerging applications for telematics systems in the consumer, industrial, or transportation sector. These application fields are ranging from unmanned aerial vehicles, in-house parking systems, and intelligent freight wagons to connected aircraft cabins.

Neuromorphic computing is going to be the new standard in low power AI applications. The integration between new neuromorphic hardware and traditional microcontroller is an open challenge. Here we present an interface board and a communication protocol that allows the communication between different devices using a microcontroller unit (Arduino Due) in the middle. Our compact printed circuit board (PCB) board links different devices as a whole system and provides power supply for the entire system using batteries as power source. Specifically, we have connected a Dynamic Vision Sensor (DVS128), SpiNNaker board and a servo motor, creating a platform for intercepting incoming objects. The data rate of the implemented interface board is 24.64k symbols/s and the latency for generating commands is about 11ms. The complete system is run only by batteries making it very suitable for robotic applications.

This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible. The main disadvantage of calculations on quantum computers is their non-determinism. For optimization problems, of course, it is possible to get surprisingly good results but without a guarantee of the actual optimality of the result, i.e, an assurance that there exists no better solution which could potentially be found by the quantum machine. In this paper, we propose an novel approach that provides such a guarantee of optimality. We generate a solution that is optimal in the strict mathematical sense, without probabilistic considerations. To accomplish this, we use the D-Wave quantum machine working as a sampler implementing quantum annealing – an approach considered a hardware metaheuristic – to obtain upper and lower bounds on the value of the objective function of the problem under consideration. Then, we use the Branch and Bound scheme controlled by quantum annealing. This allows us to obtain very quickly – in constant time – the boundaries of the considered subproblems. Our approach is a combination of calculations realized on a quantum machine controlled by procedures implemented on a classical computer, allowing us to generate optimal solutions to the NP-hard problems of task scheduling on a single machine with a total weighted tardiness criterion.  The proposed quantum algorithm shows a significant advantage over the classical CPU approach both in terms of calculation time and the number of generated solution tree nodes.

Over-the-air computing (AirComp) has recently attracted considerable attention as an efficient method of data fusion by integrating uncoded communication transmissions with computation thanks to the signal superposition offered by the multiple access channels. However, appropriate processing is required to neutralize the wireless channel effect. As, internet-of-things (IoT) applications through low-cost devices is the main target of AirComp, perfect availability of channel state information (CSI) is not always practical, there is the need to investigate the effect of imperfect CSI on AirComp. Specifically, we present novel closed-form expressions for tight approximations that can be used to design and evaluate AirComp systems. Furthermore, we design a general optimization framework that takes into account both magnitude and phase errors in the CSI. Finally, a pilot retransmission policy is designed, that offers trade-off between resources cost and the gain in the accuracy of the computations. In order to validate its application, a utility function of the cost of retransmission is introduced, namely, Retransmission Policy Cost (RPC), which can incorporate the power or throughput cost opposing to the expected gain of the selected policy. Simulations show the deterioration caused by the imperfect CSI and highlight the added value of the proposed policy under various system conditions.Â

A metasurface reflector unit cell is proposed for achieving near-complete-range of complex reflectance (i.e., magnitude and phase) for versatile beamforming capabilities. The unit cell is based on coplanar coupled resonators: a split ring resonator (SRR) with a lumped capacitor and a lumped resistor, and a dipole ring resonator (DRR) with another lumped capacitor. The DRR inserted inside the SRR creates a coupled resonance configuration which results in an enhanced complex reflectance range at the desired frequency. To provide a physical insight and explain the operation principle of the structure, the response of the unit cell is modeled as a coupled Lorentz oscillator via the effective surface susceptibilities, where a unique plasma, damping constant, and resonant frequency can be attributed to each resonator. The proposed unit cell is demonstrated in an array configuration for linear-polarized beamforming, where full-wave simulations are used to demonstrate beam-steering, gain control, side-lobe level control and dual- and triple-beam generation, as illustrative examples. Finally experimental demonstration is performed to validate the full-wave results and obtain in-depth electrical characterization of the reflectors.

We present a new machine learning algorithm, Latent Similarity, and use it to predict subject (endo)phenotypes from fMRI data. fMRI can be used to predict dysfunctional mental states. In addition, endophenotypes are known to be predictive of disease status, and are of interest in developmental studies. However, fMRI studies often suffer from small cohort size and high feature dimensionality, making reproducible prediction challenging. The innovation of our algorithm is to combine a kernel similarity function with metric learning to increase the robustness of prediction. Our algorithm becomes robust by utilizing the N squared connections between the N subjects in the cohort, rather than the features of the N subjects themselves. We identify important functional connections in the default mode and uncategorized functional networks for predicting age, sex, and intelligence. We also find that only a few connections contain most of the information required for any predictive task. We believe that our algorithm can be beneficial in small sample size, high noise, high dimensionality settings.

As an essential tool for describing dynamic systems, the concept of derivatives has been deeply rooted in all aspects of modern science. In engineering practice, due to the complexity and unknownness of the process of interest and the discreteness nature of the collected data, people need a method to obtain the derivative information of the dynamic process from these discrete data, that is, numerical differentiation method. Unfortunately, the disturbance introduced from the dynamic process itself and the data acquisition process make the collected data noisy. The existing numerical differentiation methods are susceptible to such noise in the data, making them perform poorly when faced with data from the real world. In this paper, a new numerical differentiation method is proposed. It works like a 1-D FIR filter whose tap coefficients are generated based on signal and can estimate (higher-order) derivatives from noisy sampled data of given functions. The proposed method shows much better robustness against noise and consistency against signal frequency varying than existing methods.

The massive device’s connectivity and the great data traffic demand are growing explosively in our daily lives, especially in the Internet of Things environment. For this purpose, the main objective of operators is to find the most innovative and powerful solution that can support all these needs. The sliced fifth-generation mobile network (sliced 5G) is considered a promising architecture using a new radio access technology called New Radio (NR), which can offer significantly higher data rate experiences. In addition, using Network Slicing (NS) architecture will allow more flexibility and isolation to the predefined infrastructure. The NS isolation effect’s benefit is eliminating inter-slices interference since they will be independent of each other. This paper aims to resolve the user-association problem in a sliced 5G (NR)-IoT network. Hence, the problem is formulated as an Unconstrained-Markov Decision Process (U-MDP) model. Then, the U-MDP association algorithm is proposed to find the optimal association among users and the different slices. Numerical simulation results validate the theoretical proposed model and prove her efficiency in improving the global system performance while respecting all devices’ Quality of Service.