The sound emission characteristics of unmanned aerial systems are of interest in many contexts. For a realistic representation of the directional characteristics and the radiated sound energy, it is useful if the aircraft is operated under realistic conditions. However, deriving emission-based acoustic quantities for a typical mode of operation such as cruise flight is difficult. In this paper, the directivity and sound power of a quadcopter drone are determined using a microphone array measurement of a single flight through a predefined corridor. The recorded data are processed to both reconstruct the drone’s flight path and characterize its acoustic emission. To verify the reliability of the presented method, the signal processing is tested with simulated data from moving sources with the directivity of a monopole and a dipole, respectively. Using different discretizations of the radiation direction space and with the evaluation of the frequency-dependent directivity factor, it is discussed how the sound radiation can be described as comprehensively as possible on the basis of as few quantities as possible.

Infant carriers play an important role in protecting child occupants from severe injuries caused by collisions, but the tension of harness webbing cannot be controlled properly most of the time. Infant carrier's user manual or instruction generally contains little information about the extent to which the adjusting belt should be pulled to cause the necessary webbing tension, and it is often neglected that the infants should be restrained securely. In order to improve public awareness, it is important to ascertain the effect of infant carrier's webbing tension on the occupant's chest accelerations. A testing scheme including 12 dynamic tests was devised and conducted, and test conditions were controlled strictly to ensure the accuracy and objectivity of results. P1.5 dummy's resultant and vertical chest accelerations were collected and analyzed. Both ISOFIX installation and seat belt installation methods were taken into consideration without lack of generality. Based on experimental validations, the relation between webbing's tensions and chest accelerations in frontal crash accidents was verified. Furthermore, suggestions were made about adjusting the webbing tension and the proper use of infant carriers.

This paper presents methods for detecting out-of-breath speech (OBS) under the shortness-of-breath condition due to physical load. It uses constant Q transform (CQT) for warping frequency spectrum non-linearly, which focuses on spectral saliencies of speech signal under the said condition. The existing works using deep neural networks (DNN) have spectrograms or prosodic features as input to detect the OBS speech. Two target labels, neutral and OBS, are commonly used for training DNN models. The labels, however, do not reflect the true exertion level in a speaker. In this work, a transfer-learning approach is proposed for estimating the physical exertion level. An open-source DNN model, which is trained on environmental sounds, is used for the same purpose. A multi-task learning (MTL) framework is applied for jointly learning the binary class labels and the exertion levels. The results show that the spectrogram with non-linear frequency warping performs better than its linear counterparts. Furthermore, with MTL, the classification performance F1-score improves by 3.54% and 12.67% over the single task learning (STL) using CQT-spectrogram and Mel-spectrogram, respectively.

In light of the NIMH’s Research Domain Criteria (RDoC), the advent of functional neuroimaging, novel technologies and methods provide new opportunities to develop precise and personalized prognosis and diagnosis of mental disorders. Machine learning (ML) and artificial intelligence (AI) technologies are playing an increasingly critical role in the new era of precision psychiatry. Combining ML/AI with neuromodulation technologies can potentially provide explainable solutions in clinical practice and effective therapeutic treatment. Advanced wearable and mobile technologies also call for the new role of ML/AI for digital phenotyping in mobile mental health. In this review, we provide a comprehensive review of the ML methodologies and applications by combining neuroimaging, neuromodulation and advanced mobile technologies in psychiatry practice. Additionally, we review the role of ML in molecular phenotyping, cross-species biomarker identification in precision psychiatry. We further discuss explainable AI (XAI) and causality testing in a closed-human-in-the-loop manner, and highlight the ML potential in multimedia information extraction and multimodal data fusion. Finally, we discuss conceptual and practical challenges in precision psychiatry and highlight ML opportunities in future research.

In a conventional power grid, energy theft is difficult to detect due to limited communication and data transition. The smart meter along with big data mining technology leads to significant technological innovation in the field of energy theft detection. This paper proposes a convolutional long and shortterm memory (ConvLSTM) based energy theft detection (ETD) model to identify electricity theft users. In this work, electricity consumption data is reshaped quarterly into a 2-dimensional matrix and used as the sequential input to the ConvLSTM. The convolutional neural network (CNN) embedded into the long short-term memory (LSTM) can better learn the features of the data on different quarters, months, weeks, and days. Besides, the proposed model incorporates batch normalization. This allows the proposed ETD model to support raw format electricity consumption data input, reducing training time and increasing the efficiency of model deployment. The final experimental results show that the proposed ConvLSTM model exhibits good robustness. It outperforms the multilayer perceptron (MLP) and CNN-LSTM in terms of performance metrics and model generalization capability.

Data augmentation plays a crucial role in deep learning-based medical imaging analysis, but manually designing tailored data augmentation strategies for each dataset is impractical. Although automatic data augmentation (ADA) techniques have been explored, they often focus solely on data augmentation without considering the importance of neural architecture. Similarly, neural architecture search (NAS) methods mainly concentrate on optimizing the neural architecture while overlooking the impact of data augmentation. However, both data augmentation and neural architecture are interrelated and should be considered together. The joint optimization of data augmentation and neural architecture can lead to improved model performance by harnessing the complementary effects of customized data augmentation strategies and compatible neural architectures. Despite this, the seamless integration of data augmentation and neural architecture search remains under-explored. To address this research gap, we propose \medpipe, an approach that enables end-to-end joint search of data augmentation and neural architecture. We introduce a compact data augmentation search space and unify data augmentation and neural architecture into a cohesive network. This allows simultaneous exploration, optimizing their synergy for enhanced performance. Experimental evaluation on nine medical datasets highlights the necessity of the joint search for data augmentation and neural architecture, demonstrating the superior performance of our approach. Our work opens up possibilities for future applications in diverse medical domains.

—Cross-project defect prediction (CPDP) makes use of cross-project (CP) data to overcome the lack of data necessary to train well-performing software defect prediction (SDP) classifiers in the early stage of new software projects. Since the CP data (known as the source) may be different from the new project’s data (known as the target), this makes it difficult for CPDP classifiers to perform well. In particular, it is a mismatch of data distributions between source and target that creates this difficulty. Transfer learning-based CPDP classifiers are designed to minimize these distribution differences. The first Transfer learning-based CPDP classifiers treated these differences equally, thereby degrading prediction performance. To this end, recent research has proposed the Weighted Balanced Distribution Adaptation (W-BDA) method to leverage the importance of both distribution differences to improve classification performance. Although W-BDA has been shown to improve model performance in CPDP, research to date has failed to consider model performance in light of increasing target data or variances in data sampling. We provide the first investigation of when and to what extent the effect of increasing the target data and using various sampling techniques have when leveraging the importance of both distribution differences. We extend the initial W-BDA method and call this extension the W-BDA+‘ method. To evaluate the effectiveness of W-BDA+‘ for improving CPDP performance, we conduct eight experiments on 18 projects from four datasets where data sampling was performed with different sampling methods. We evaluate our method using four complementary indicators (i.e., Balanced Accuracy, AUC, F-measure and G-Measure). Our findings reveal an average improvement of 6%, 7.5%, 10% and 12% for these four indicators when W-BDA+‘ is compared to five other baseline methods (including W-BDA), for all four of the sampling methods used. Also, as the target to source ratio is increased with different sampling methods, we observe a decrease in performance for the original W-BDA, with our W-BDA+ approach outperforming the original W-BDA in most cases. Our results highlight the importance of adjusting for data imbalance and having an awareness of the effect of the increasing availability of target data in CPDP scenarios.

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.

An FDTD scheme for simulating wave propagation inside Cole-Cole, Havriliak-Negami, or Raicu dispersive media is proposed. The main difficulty in FDTD implementations for such media is the fact that the time-domain polarization relations are fractional integrodifferential equations. To circumvent this difficulty, the dispersive susceptibility of the medium is modeled by Pade approximants. It is proved that under certain conditions the Pade approximant is equal to a weighted sum of Debye terms, while the physical restrictions for positive weights and relaxation times are fulfilled with certainty. Hence, a set of first-order auxiliary differential equations, which approximate the original fractional polarization relation, is derived and directly embedded in the FDTD scheme. In contrast to previous works, the derivation of the Debye expansion of the original dispersive permittivity is carried out in a straightforward manner without involving nonlinear optimization techniques. Comparisons between analytical and FDTD results of the reflection and transmission coefficients as well as of the transfer functions inside the dispersive media illustrate the efficiency of the proposed scheme, over wideband frequency domains

As data-driven AI systems become increasingly integrated into industry, concerns have recently arisen regarding potential privacy breaches and the inadvertent leakage of sensitive user data through the exploitation of these systems. In this paper, we explore the intersection of data privacy and AI-powered document analysis systems, presenting a comprehensive benchmark of well-known privacy-preserving methods for the task of document image classification. In particular, we investigate four different privacy methods—Differential Privacy (DP), Federated Learning (FL), Differentially Private Federated Learning (DP-FL), and Secure Multi-Party Computation (SMPC)—on two well-known document benchmark datasets, namely RVL-CDIP and Tobacco3482. Furthermore, we investigate the performance of each method under a variety of configurations for thorough benchmarking. Finally, the privacy strength of each approach is assessed by subjecting the private models to well-known membership inference attacks. Our results demonstrate that, with sufficient tuning of hyperparameters, Differential Privacy (DP) can achieve reasonable performance on the task of document image classification while also ensuring rigorous privacy constraints, both in standalone and federated learning setups. On the other hand, while FL-based approaches present less implementation complexity and incur little to no loss in performance on the task, they do not offer sufficient protection against privacy attacks. By rigorously benchmarking various privacy approaches, our study paves the way for integrating deep document classification models into industrial pipelines while meeting regulatory and ethical standards, including GDPR and the AI Act 2022.

The datas in this survey were collected from reference papers.

A thin ScAlN layer was deposited on high resistivity (111) oriented silicon and two port surface acoustic wave (SAW) devices were manufactured, using advanced nanolithographic techniques, on this material. The Surface Acoustic Wave and Spin Wave (SAW/SW) coupling was performed via a thin ferromagnetic layer (Ni) placed between the interdigitated transducers (IDTs) of the SAW device. Since the phase velocity in this material is lower than in the Si substrate, both Rayleigh (4.67 GHz) and Sezawa (8.05 GHz) propagation modes could be observed. The amplitude of the S21 parameter around the two resonances was measured for values of the magnetic field µ0H from -280 to +280 mT, at different angles (θ) between the SAW propagation direction and the magnetic field direction. A maximum decrease of 2.54 dB occurred in S21 for the Rayleigh mode at µ0H = -90 mT, and of 7.24 dB for the Sezawa mode at µ0H =-203 mT, both at θ = 45°. These values were extracted from time gated processing of the frequency domain raw data.

Deep Learning (DL) models can be trained to automatically learn the underlying features of the traffic patterns in IoT networks to detect complex botnet attacks. However, the performance of a neural network model largely depends on the set of hyperparameters that is used for the model development. In this paper, an algorithm is proposed to determine the optimal set of hyperparameters (the numbers of hidden layers and hidden units, the learning rate, the optimiser, the activation function, the batch size, and the number of epochs) for efficient DL-based botnet detection in IoT networks. The DL models employ a Deep Neural Network (DNN) architecture for binary and multi-class classification. DNN-based botnet detection models are developed and experiments are performed with the Bot-IoT and N-BaIoT datasets to validate the effectiveness of the hyperparameter optimisation method. Experiment results showed that the proposed method produced DNN models that achieved high botnet attack detection rates, low false alarm rates, and near real-time computation speed.

Deep learning-based object detection technology is actively studied, and non-maximum suppression (NMS) is an algorithm used to remove redundant object detection. NMS creates a bounding box for objects detected using a fixed ratio to determine the probability of an object being present. However, this constraint does not solve the difficulty in detecting objects that significantly overlap or are too small. Soft-NMS is an improvement. Nevertheless, free parameter values are manually chosen to provide non-optimal results. Our proposed Adaptive-NMS algorithm (1) calculates the optimal free parameters in real-time and (2) provides for multi-class object detection in highly dense images, improving results over Soft-NMS up to 12.2%.

Fully differential topologies are well-suited for biopotential amplifiers, mainly for single-supply battery-powered circuits such as portable wearable devices where a reduced number of parts is desired. A novel fully differential biopotential amplifier is proposed with the goal of providing electrode offset rejection, bandwidth limitation, and a temporal response compliant with biomedical standards with only a single commercial quad operational amplifier (OA) integrated circuit.

The power quality of distribution network side directly affects the charging efficiency and safety of electric vehicles, so excellent power quality detection approach will contribute to improve the charging quality of electric vehicles. However, present power quality decomposition technologies are mainly suitable for single disturbance signal detection, but not for composite power quality signals under different contextual conditions. To resolve these problems and improve the accuracy and reliability of waveform decomposition, a context-based two-phase decomposition method is proposed. At first, we develop a hybrid-model decomposition framework for analyzing composite disturbance signals to improve the decomposition performance of voltage sag waveform. Then, we design a context-awere mechanism to adapt to multiple types of composite disturbance signals in different regions, thereby make the model has better adaptability. Furthermore, We are the first time to propose a voltage sag waveform decomposition method for studying the impact of power quality in low-voltage distribution networks on electric vehicle charging. The experimental results demonstrate that compared with other methods, the proposed two-phase model is more applicable in decomposing composite power quality disturbance signal including voltage sag events, as well as attain significant decomposition results.

Benefiting from the development of image sensors and the popularity of light-emitting diode (LED) lighting technology, visible light positioning (VLP) technology based on image sensors has ushered in vigorous development and broad prospects, which can provide low-cost and high-accuracy position service. However, the existing approaches require dense LEDs or sensors such as gyroscopes to assist positioning, which limit the area and lower the accuracy of positioning because of the errors from imperfect sensors. In this paper, we propose a simultaneous localization and calibration VLP method based on double coplanar circular LED lights aiming to get rid of the dependence on additional sensors and dense LED transmitters. By the pinhole camera model and the perspective projection of circle, our proposed method extends the available position area and relaxes the required quantity of LED to two. The experiment result shows that our system has a mean 3D positioning accuracy of 7.91cm, a mean angle error of less than 1.6°, and an average latency of 182ms on mobile devices.

This paper summarizes the technical activities of the IEEE Task Force on Solving Large Scale Optimization Problems in Electricity Market and Power System Applications. This Task Force was established by the IEEE Technology and Innovation Subcommittee to first review the state-of-the-art of the security-constrained unit commitment (SCUC) business model, its mathematical formulation, and solution techniques in solving electricity market clearing problems. The Task Force then investigated the emerging challenges of future market clearing problems and presented efforts in building benchmark mathematical and business models.