Wireless World Research and Trends Magazine

Advancements in Deep-Learning-Based Object Detection in Challenging Environments

Abhishek Mukhopadhyay — 2024-03-21

This article focuses on object detection in challenging environments, where objects of interest need to be detected in images captured under unconstrained conditions. These environments can include outdoor scenes with varying lighting, weather conditions, and background clutter. Object detection in such scenarios is crucial for applications like autonomous driving, surveillance, and industrial production inspection. The paper explores existing techniques for object detection in challenging environments and proposes novel solutions to enhance its performance. One key aspect emphasized in the paper is the need for relevant datasets to validate models under unconstrained scenarios. While several datasets already exist, the research identifies gaps in specific situations and addresses them by proposing new datasets, such as the Indian Lane Dataset for autonomous vehicles. The proposed object detection framework leverages spatial information to detect objects in challenging environments, and its performance is evaluated against benchmark datasets and the newly proposed datasets. A real-world application demonstrates significant improvements in object detection performance in natural environments.

Emotion Detection from Speech: A Comprehensive Approach Using Speech-to-Text Transcription and Ensemble Learning

Fatima M Inamdar — 2024-03-21

The field of text-to-emotion analysis is investigated in this study, which uses an interactive methodology to reveal subtle emotional insights in textual data. The research explores the complex relationship between language and emotion using sophisticated methods without focusing on any particular frontend or backend technology. The research attempts to improve our understanding of how literary information transmits emotional subtleties by emphasizing a broad but methodical examination. The lack of mentions of particular libraries and backends indicates an emphasis on the general ideas and techniques used in text-to-emotion analysis.

The findings demonstrate the possibility of deriving significant emotional context from text, opening doors for applications in a variety of fields where user sentiment analysis is essential. This study adds to the body of knowledge on emotional intelligence in computational linguistics and lays the groundwork for future developments in text analysis techniques.

Beam-Tracking Challenges in THz Communications

Giorgos Stratidakis — 2024-03-21

In recent years, the demand for high data rates has increased drastically. In response to this increasing demand, higher frequencies, such as millimeter wave (mmWave) and terahertz (THz) have been considered that offer much larger bandwidth and potentially much higher data rates. Nevertheless, with the increase in frequency, the pathloss increases significantly, with the non-line-of-sight (nLoS) incurring attenuation levels that can dramatically reduce the quality of a wireless link. The use of directional antennas to counteract the increased pathloss is widely accepted. However, highly directional beams are prone to misalignment and a method to track the tracking object (TO) of the link with consistency, accuracy and low overhead is needed. To this end, the design of beam-tracking algorithms has been proposed. In this work, the main parameters that affect the reliability of beam-tracking are presented along with the challenges that beam-tracking algorithms need to address. Furthermore, the performance merits with respect to three reliability parameters are presented in the case of a simple beam-tracking algorithm.

Coherence Bandwidth of Rice Channels for Millimeter Wave and Sub-Terahertz Applications

Werner Mohr — 2024-03-21

6G research in the millimeter wave and sub-Terahertz domain is targeting very wideband systems with significantly higher data throughput than for 5G systems. Multipath propagation under shadowing conditions is affecting radio propagation, where multipath propagation results in frequency-selective fading, which is characterized by the coherence bandwidth and the time variation by the coherence time. In these frequency ranges shadowing can be overcome by additional means in the network deployment such as reflectors, RIS arrays or repeaters, which provide at the receiver a channel impulse response with a strong component (Rice type channel). Coherence bandwidth and coherence time are well-known for Rayleigh channels. However, both parameters for Rice channels versus the Rice factor K′ are not available. This paper is investigating the coherence bandwidth and time for Rice channels based on an approximative approach for the fading statistics. With the proposed correlation criterion, the coherence bandwidth and time tend to infinity from a Rice factor around K′≥4=ˆ6 dB. These relations are provided by approximative functions for K′≥0 starting at the Rayleigh channel.

Adversarial Machine-Learning-Enabled Anonymization of OpenWiFi Data

Samhita Kuili — 2024-03-21

Data privacy and protection through anonymization is a critical issue for network operators or data owners before it is forwarded for other possible use of data. With the adoption of Artificial Intelligence (AI), data anonymization augments the likelihood of covering up necessary sensitive information; preventing data leakage and information loss. OpenWiFi networks are vulnerable to any adversary who is trying to gain access or knowledge on traffic regardless of the knowledge possessed by data owners. The odds for discovery of actual traffic information is addressed by applied conditional tabular generative adversarial network (CTGAN). CTGAN yields synthetic data; which disguises as actual data but fostering hidden acute information of actual data. In this paper, the similarity assessment of synthetic with actual data is showcased in terms of clustering algorithms followed by a comparison of performance for unsupervised cluster validation metrics. A well-known algorithm, K-means outperforms other algorithms in terms of similarity assessment of synthetic data over real data while achieving nearest scores 0.634, 23714.57, and 0.598 as Silhouette, Calinski and Harabasz and Davies Bouldin metric respectively. On exploiting a comparative analysis in validation scores among several algorithms, K-means forms the epitome of unsupervised clustering algorithms ensuring explicit usage of synthetic data at the same time a replacement for real data. Hence, the experimental results aim to show the viability of using CTGAN-generated synthetic data in lieu of publishing anonymized data to be utilized in various applications.

On the Impact of CDL and TDL Augmentation for RF Fingerprinting Under Impaired Channels

Omer Melih Gul — 2024-03-21

Cyber-physical systems have recently been used in several areas (such as connected and autonomous vehicles) due to their high maneuverability. On the other hand, they are susceptible to cyber-attacks. Radio frequency (RF) fingerprinting emerges as a promising approach. This work aims to analyze the impact of decoupling tapped delay line and clustered delay line (TDL+CDL) augmentation-driven deep learning (DL) on transmitter-specific fingerprints to discriminate malicious users from legitimate ones. This work also considers 5G-only-CDL, WiFi-only-TDL augmentation approaches. RF fingerprinting models are sensitive to changing channels and environmental conditions. For this reason, they should be considered during the deployment of a DL model. Data acquisition can be another option. Nonetheless, gathering samples under various conditions for a train set formation may be quite hard. Consequently, data acquisition may not be feasible. This work uses a dataset that includes 5G, 4G, and WiFi samples, and it empowers a CDL+TDL based augmentation technique in order to boost the learning performance of the DL model. Numerical results show that CDL+TDL, 5G-only-CDL, and WiFi-only-TDL augmentation approaches achieve 87.59%, 81.63%, 79.21% accuracy on unobserved data while TDL/CDL augmentation technique and no augmentation approach result in 77.81% and 74.84% accuracy on unobserved data, respectively.