Due to the escalating network throughput and security risks, the exploration of intrusion detection systems (IDSs) has garnered significant attention within the computer science field. The majority of modern IDSs are constructed using deep learning techniques. Nevertheless, these IDSs still have shortcomings where most datasets used for IDS lies in their high imbalance, where the volume of samples representing normal traffic significantly outweighs those representing attack traffic. This imbalance issue restricts the performance of deep learning classifiers for minority classes, as it can bias the classifier in favor of the majority class. To address this challenge, many solutions are proposed in the literature. TDCGAN is an innovative Generative Adversarial Network (GAN) based on a model-driven approach used to address imbalanced data in the IDS dataset. This paper investigates the performance of TDCGAN by employing it to balance data across four benchmark IDS datasets which are CIC-IDS2017, CSE-CIC-IDS2018, KDD-cup 99, and BOT-IOT. Next, four machine learning methods are employed to classify the data, both on the imbalanced dataset and on the balanced dataset. A comparison is then conducted between the results obtained from each to identify the impact of having an imbalanced dataset on classification accuracy. The results demonstrated a notable enhancement in the classification accuracy for each classifier after the implementation of the TDCGAN model for data balancing.

Deep learning models, particularly transformer architectures, have revolutionized various computer vision tasks, including image classification. However, their performance under different types and levels of noise remains a crucial area of investigation. In this study, we explore the noise sensitivity of prominent transformer models trained on the ImageNet dataset. We systematically evaluate 22 transformer variants, ranging from state-of-the-art large-scale models to compact versions tailored for mobile applications, under five common types of image distortions. Our findings reveal diverse sensitivities across different transformer architectures, with notable variations in performance observed under additive Gaussian noise, multiplicative Gaussian noise, Gaussian blur, salt-and-pepper noise, and JPEG compression. Interestingly, we observe a consistent robustness of transformer models to JPEG compression, with top-5 accuracies exhibiting higher resilience to noise compared to top-1 accuracies. Furthermore, our analysis highlights the vulnerability of mobile-oriented transformer variants to various noise types, underscoring the importance of noise robustness considerations in model design and deployment for real-world applications. These insights contribute to a deeper understanding of transformer model behavior under noisy conditions and have implications for improving the robustness and reliability of deep learning systems in practical scenarios.

Swallowing is a complex neuromuscular activity regulated by the autonomic central nervous system, and impairment can lead to dysphagia, which is difficulty in swallowing. This research presents a novel approach that utilizes wireless, wearable technology for the continuous mechano-acoustic tracking of respiratory activities and swallowing. To address the challenge of accurately tracking swallowing amidst potential confounding activities or significant body movements, we employ two accelerometers. These accelerometers help distinguish between genuine swallowing events and other activities. By monitoring movements and vibrations through the skin surface, the developed device enables non-intrusive monitoring of swallowing dynamics and respiratory patterns. Our focus is on the development of both the wireless skin-interfaced device and an advanced algorithm capable of detecting swallowing dynamics in conjunction with respiratory phases. The device and algorithm demonstrate robustness in detecting respiratory patterns and swallowing instances, even in scenarios where users exhibit periodic movements due to disease or daily activities. Furthermore, peak detection using an adaptive threshold automatically adjusts to an individual’s signal strength, facilitating the detection of swallowing signals without the need for individual adjustments. This innovation has significant potential for enhancing patient training and rehabilitation programs aimed at addressing dysphagia and related respiratory issues.

In this project, the electroencephalogram (EEG) channel(s) is used to better characterize post-traumatic stress disorder (PTSD). For this aim, we applied boosting methods along with a combination of k-means and Support Vector Machine (SVM) models to find the diagnostic channels of PTSD cases and healthy subjects. We grouped 32 channels and 12 subjects (6 PTSD and 6 healthy controls) using k-means. Channels of the brain are grouped by the k-means clustering method to find the most similar part of the brain. This approach uses SVM by performing classification based on cluster classes are been mapped to EEG channels. This mapping uses information across all samples without the bias of using the outcome variable. The linear SVM found weights that distinguished channels within each subject for each cluster to compare the PTSD cases and healthy controls’ channel weights. It was found that the significant SVM weights of F4, F8, and Pz were smaller in subjects with PTSD than in healthy subjects. This new method can be used as a tool to better understand the relationship between EEG signals and diagnosis.

The development of robust circuit structures remains a pivotal milestone in electronic device research. This article proposes an integrated hardware–software system designed for the acquisition, processing, and analysis of surface electromyographic (sEMG) signals. The system analyzes sEMG signals to understand muscle function and neuromuscular control, employing convolutional neural networks (CNNs) for pattern recognition. The electrical signals analyzed on healthy and unhealthy subjects are acquired using a meticulously developed integrated circuit system featuring biopotential acquisition electrodes. The signals captured in the database are extracted, classified, and interpreted by the application of CNNs with the aim of identifying patterns indicative of neuromuscular problems. By leveraging advanced learning techniques, the proposed method addresses the non-stationary nature of sEMG recordings and mitigates cross-talk effects commonly observed in electrical interference patterns captured by surface sensors. The integration of an AI algorithm with the signal acquisition device enhances the qualitative outcomes by eliminating redundant information. CNNs reveals their effectiveness in accurately deciphering complex data patterns from sEMG signals, identifying subjects with neuromuscular problems with high precision. This paper contributes to the landscape of biomedical research, advocating for the integration of advanced computational techniques to unravel complex physiological phenomena and enhance the utility of sEMG signal analysis.

This paper presents a new method for the design of separable-denominator 2-D IIR filters with nearly linear phase in the passband. The design method is based on a balanced realization model reduction technique. The nearly linear-phase 2-D IIR filter is designed using 2-D model reduction from a linear-phase 2-D FIR filter, which serves as the initial filter. The structured controllability and observability Gramians Ps and Qs serve as the foundation for this technique. onal positive-definite matrices that satisfy 2-D Lyapunov equations. An efficient method is used to compute these Gramians by minimizing the traces of Ps and Qs under linear matrix inequality (LMI) constraints. The use of these Gramians ensures that the resulting 2-D IIR filter preserves stability and can be implemented using a separable-denominator 2-D filter with fewer coefficients than the original 2-D FIR filter. Numerical examples show that the proposed method compares favorably with existing techniques.

Within the broad problem known as automatic music transcription, we considered the specific task of automatic drum transcription (ADT). This is a complex task that has recently shown significant advances thanks to deep learning (DL) techniques. Most notably, massive amounts of labeled data obtained from crowds of annotators have made it possible to implement large-scale supervised learning architectures for ADT. In this study, we explored the untapped potential of these new datasets by addressing three key points: First, we reviewed recent trends in DL architectures and focused on two techniques, self-attention mechanisms and tatum-synchronous convolutions. Then, to mitigate the noise and bias that are inherent in crowdsourced data, we extended the training data with additional annotations. Finally, to quantify the potential of the data, we compared many training scenarios by combining up to six different datasets, including zero-shot evaluations. Our findings revealed that crowdsourced datasets outperform previously utilized datasets, and regardless of the DL architecture employed, they are sufficient in size and quality to train accurate models. By fully exploiting this data source, our models produced high-quality drum transcriptions, achieving state-of-the-art results. Thanks to this accuracy, our work can be more successfully used by musicians (e.g., to learn new musical pieces by reading, or to convert their performances to MIDI) and researchers in music information retrieval (e.g., to retrieve information from the notes instead of audio, such as the rhythm or structure of a piece).

In this paper, a new optimization method for the design of nearly linear-phase two-dimensional infinite impulse (2D IIR) digital filters with a separable denominator is proposed. A design framework for 2D IIR digital filters is formulated as a nonlinear constrained optimization problem where the group delay deviation in the passband is minimized under prescribed soft magnitude constraints and hard stability requirements. To achieve this goal, sub-level sets of the group delay deviations are utilized to generate a sequence of filters, from which the one with the best performance is selected. The quality of the obtained filter is evaluated using three quality factors, namely, the passband magnitude quality factor Qh and the group delay deviation quality factor Qτ, while the third one is a new quality factor Qs that assesses the performance in the stopband relative to the minimum filter gain in the passband. The proposed framework is implemented using the interior-point (IP) method in a MATLAB environment, and the experimental results show that filters designed using the proposed method have good magnitude response and low group delay deviation. The performance of the resulting filters is compared with the results of other methods.

Experimental data from the SACLANTCEN 1993 Mediterranean Experiment are reviewed to assess the reduction of the search space for the localization and tracking of an acoustic source in a three-dimensional environment. Key to this goal is the availability of an initial estimate of source range and depth (called the 2D initial guess); an ambiguous estimate of source bearing can be obtained from the 2D initial guess through Environmental Signal Processing, and the ambiguity can be removed by searching for the source only in the range/bearing regions where bearing estimates are higher. This search provides a new estimate of source range and a single bearing, which together with the estimate for source depth constitute the center of the reduced search space for source localization and tracking. The suggested approach is tested on experimental data from the SACLANTCEN experiment considering different frequencies, as well as a stationary and a moving source.

Currently, Internet of Things (IoT) has become common in various applications, including smart factories, smart cities, and smart homes. In them, wireless local-area networks (WLANs) are widely used due to their high-speed data transfer, flexible coverage ranges, and low costs. To enhance the performance, the WLAN configuration should be optimized in dense WLAN environments where multiple access points (APs) and hosts exist. Previously, we have studied the active AP configuration algorithm for dual interfaces using IEEE802.11n and 11ac protocols at each AP under non-channel bonding (non-CB). In this paper, we study the algorithm considering the channel bonding (CB) to enhance its capacity by bonding two channels together. To improve the throughput estimation accuracy of the algorithm, the reduction factor is introduced at contending hosts for the same AP. For evaluations, we conducted extensive experiments using the WIMENT simulator and the testbed system using Raspberry Pi 4B APs. The results show that the estimated throughput is well matched with the measured one, and the proposal achieves the higher throughput with a smaller number of active APs than the previous configurations.