Cross-and-Diagonal Networks: An Indirect Self-Attention Mechanism for Image Classification.

In recent years, computer vision has witnessed remarkable advancements in image classification, specifically in the domains of fully convolutional neural networks (FCNs) and self-attention mechanisms. Nevertheless, both approaches exhibit certain limitations. FCNs tend to prioritize local information, potentially overlooking crucial global contexts, whereas self-attention mechanisms are computationally intensive despite their adaptability. In order to surmount these challenges, this paper proposes cross-and-diagonal networks (CDNet), innovative network architecture that adeptly captures global information in images while preserving local details in a more computationally efficient manner. CDNet achieves this by establishing long-range relationships between pixels within an image, enabling the indirect acquisition of contextual information. This inventive indirect self-attention mechanism significantly enhances the network's capacity. In CDNet, a new attention mechanism named "cross and diagonal attention" is proposed. This mechanism adopts an indirect approach by integrating two distinct components, cross attention and diagonal attention. By computing attention in different directions, specifically vertical and diagonal, CDNet effectively establishes remote dependencies among pixels, resulting in improved performance in image classification tasks. Experimental results highlight several advantages of CDNet. Firstly, it introduces an indirect self-attention mechanism that can be effortlessly integrated as a module into any convolutional neural network (CNN). Additionally, the computational cost of the self-attention mechanism has been effectively reduced, resulting in improved overall computational efficiency. Lastly, CDNet attains state-of-the-art performance on three benchmark datasets for similar types of image classification networks. In essence, CDNet addresses the constraints of conventional approaches and provides an efficient and effective solution for capturing global context in image classification tasks.

Multi-Functional Reconfigurable Intelligent Surfaces for Enhanced Sensing and Communication.

In this paper, we propose a reconfigurable intelligent surface (RIS) that can dynamically switch the transmission and reflection phase of incident electromagnetic waves in real time to realize the dual-beam or quad-beam and convert the polarization of the transmitted beam. Such surfaces can redirect a wireless signal at will to establish robust connectivity when the designated line-of-sight channel is disturbed, thereby enhancing the performance of wireless communication systems by creating an intelligent radio environment. When integrated with a sensing element, they are integral to performing joint detection and communication functions in future wireless sensor networks. In this work, we first analyze the scattering performance of a reconfigurable unit element and then design a RIS. The dynamic field scattering manipulation capability of the RIS is validated by full-wave electromagnetic simulations to realize six different functions. The scattering characteristics of the proposed unit element, which incorporates two p-i-n diodes have been substantiated through practical implementation. This involved the construction of a simple prototype and the subsequent examination of its scattering properties via the free-space measurement method. The obtained transmission and reflection coefficients from the measurements are in agreement with the anticipated outcomes from simulations.

Robust sensing suite for measuring temporal dynamics of surface temperature in sewers.

Sewerage systems are paramount underground infrastructure assets for any nation. In most cities, they are old and have been exposed to significant microbial induced corrosion. It is a serious global problem as they pose threats to public health and economic repercussions to water utilities. For managing sewer assets efficaciously, it is vital to predict the rate of corrosion. Predictive models of sewer corrosion incorporate concrete surface temperature measurements as an observation. However, currently, it has not been fully utilized due to unavailability of a proven sensor. This study reports the feasibility of infrared radiometer for measuring the surface temperature dynamics in the aggressive sewer conditions. The infrared sensor was comprehensively evaluated in the laboratory at different environmental conditions. Then, the sensor suite was deployed in a Sydney based sewer for three months to perform continuous measurements of surface temperature variations. The field study revealed the suitability of the developed sensor suite for non-contact surface temperature measurements in hostile sewer conditions. Further, the accuracy of the sensor measurements was improved by calibrating the sensor with emissivity coefficient of the sewer concrete. Overall, this study will ameliorate the present sewer corrosion monitoring capabilities by providing new data to models predicting sewer corrosion.

Influence of multiple dynamic factors on the performance of myoelectric pattern recognition.

Hand motion classification using surface Electromyogram (EMG) signals has been widely studied for the control of powered prosthetics in laboratory conditions. However, clinical applicability has been limited, as imposed by factors like electrodes shift, variations in the contraction force levels, forearm rotation angles, change of limb position and many other factors that all affect the EMG pattern recognition performance. While the impact of several of these factors on EMG parameter estimation and pattern recognition has been considered individually in previous studies, a minimum number of experiments were reported to study the influence of multiple dynamic factors. In this paper, we investigate the combined effect of varying forearm rotation angles and contraction force levels on the robustness of EMG pattern recognition, while utilizing different time-and-frequency based feature extraction methods. The EMG pattern recognition system has been validated on a set of 11 subjects (ten intact-limbed and one bilateral transradial amputee) performing six classes of hand motions, each with three different force levels, each at three different forearm rotation angles, with six EMG electrodes plus an accelerometer on the subjects' forearm. Our results suggest that the performance of the learning algorithms can be improved with the Time-Dependent Power Spectrum Descriptors (TD-PSD) utilized in our experiments, with average classification accuracies of up to 90% across all subjects, force levels, and forearm rotation angles.

Towards limb position invariant myoelectric pattern recognition using time-dependent spectral features.

Recent studies in Electromyogram (EMG) pattern recognition reveal a gap between research findings and a viable clinical implementation of myoelectric control strategies. One of the important factors contributing to the limited performance of such controllers in practice is the variation in the limb position associated with normal use as it results in different EMG patterns for the same movements when carried out at different positions. However, the end goal of the myoelectric control scheme is to allow amputees to control their prosthetics in an intuitive and accurate manner regardless of the limb position at which the movement is initiated. In an attempt to reduce the impact of limb position on EMG pattern recognition, this paper proposes a new feature extraction method that extracts a set of power spectrum characteristics directly from the time-domain. The end goal is to form a set of features invariant to limb position. Specifically, the proposed method estimates the spectral moments, spectral sparsity, spectral flux, irregularity factor, and signals power spectrum correlation. This is achieved through using Fourier transform properties to form invariants to amplification, translation and signal scaling, providing an efficient and accurate representation of the underlying EMG activity. Additionally, due to the inherent temporal structure of the EMG signal, the proposed method is applied on the global segments of EMG data as well as the sliced segments using multiple overlapped windows. The performance of the proposed features is tested on EMG data collected from eleven subjects, while implementing eight classes of movements, each at five different limb positions. Practical results indicate that the proposed feature set can achieve significant reduction in classification error rates, in comparison to other methods, with ≈8% error on average across all subjects and limb positions. A real-time implementation and demonstration is also provided and made available as a video supplement (see Appendix A).

Muscle computer interfaces for driver distraction reduction.

Driver distraction is regarded as a significant contributor to motor-vehicle crashes. One of the important factors contributing to driver distraction was reported to be the handling and reaching of in-car electronic equipment and controls that usually requires taking the drivers' hands off the wheel and eyes off the road. To minimize the amount of such distraction, we present a new control scheme that senses and decodes the human muscles signals, denoted as Electromyogram (EMG), associated with different fingers postures/pressures, and map that to different commands to control external equipment, without taking hands off the wheel. To facilitate such a scheme, the most significant step is the extraction of a set of highly discriminative feature set that can well separate between the different EMG-based actions and to do so in a computationally efficient manner. In this paper, an accurate and efficient method based on Fuzzy Neighborhood Discriminant Analysis (FNDA), is proposed for discriminant feature extraction and then extended to the channel selection problem. Unlike existing methods, the objective of the proposed FNDA is to preserve the local geometrical and discriminant structures, while taking into account the contribution of the samples to the different classes. The method also aims to efficiently overcome the singularity problems of classical LDA by employing the QR-decomposition. Practical real-time experiments with eight EMG sensors attached on the human forearm of eight subjects indicated that up to fourteen classes of fingers postures/pressures can be classified with <7% error on average, proving the significance of the proposed method.

Driver drowsiness classification using fuzzy wavelet-packet-based feature-extraction algorithm.

Driver drowsiness and loss of vigilance are a major cause of road accidents. Monitoring physiological signals while driving provides the possibility of detecting and warning of drowsiness and fatigue. The aim of this paper is to maximize the amount of drowsiness-related information extracted from a set of electroencephalogram (EEG), electrooculogram (EOG), and electrocardiogram (ECG) signals during a simulation driving test. Specifically, we develop an efficient fuzzy mutual-information (MI)- based wavelet packet transform (FMIWPT) feature-extraction method for classifying the driver drowsiness state into one of predefined drowsiness levels. The proposed method estimates the required MI using a novel approach based on fuzzy memberships providing an accurate-information content-estimation measure. The quality of the extracted features was assessed on datasets collected from 31 drivers on a simulation test. The experimental results proved the significance of FMIWPT in extracting features that highly correlate with the different drowsiness levels achieving a classification accuracy of 95%-- 97% on an average across all subjects.