In recent years, the development of natural language process (NLP) technologies and deep learning hardware has led to significant improvement in large models (LLMs). The ChatGPT, state-of-the-art LLM built on GPT-3.5 GPT-4, shows excellent capabilities general understanding reasoning. Researchers also tested GPTs a variety NLP-related tasks benchmarks got results. With exciting performance daily chat, researchers began explore capacity ChatGPT expertise that requires professional education...

CNN has demonstrated remarkable performance in EEG signal detection, yet it still faces limitations terms of global perception. Additionally, due to individual differences signals, the generalization ability epilepsy detection models is week. To address this issue, paper presents a cross-patient method utilizing multi-head self-attention mechanism. This first utilizes Short-Time Fourier Transform (STFT) transform original signals into time-frequency features, then local information using...

With the development of 5G new radio (NR) applications in high-speed scenarios, such as non-terrestrial networks (NTN), Doppler shift systems is significant. In this paper, an estimation scheme for frequency offset with large range orthogonal division multiplexing (OFDM) proposed. The proposed firstly takes advantage 2π-periodicity phase between two pilot OFDM symbols to estimate a set candidate offsets. It then uses autocorrelation sequence determine final estimated offset. This method...

Introduction In recent years, the use of EEG signals for seizure detection has gained widespread academic attention. Aiming at problem overfitting deep learning models due to small number signal data during epilepsy detection, this paper proposes an method that combines augmentation and learning. Methods First, Adversarial Mixup Data Augmentation (AMDA) is used realize augmentation, which effectively enriches training samples. To further improve classification accuracy robustness a...

Remote sensing image dehazing (RSID) aims to remove nonuniform and physically irregular haze factors for high-quality restoration. The emergence of CNNs Transformers has taken extraordinary strides in the RSID arena. However, these methods often struggle demonstrate balance adequate long-range dependency modeling maintaining computational efficiency. To this end, we propose first lightweight network on mamba-based model called RSDhamba field RSID. Greatly inspired by recent rise Selective...

Background: Corn is the main grain crop grown in China, and ear shape index of corn an important parameter for breeding new varieties, including length, diameter, row number ears, grains per ear, so on. Objective: In order to solve problem limited field view associated with computer detection against a complex background, this paper proposes panoramic splicing method ears which can splice 10 images at same time, improve information collection efficiency, display comprehensive information,...

In this article, we consider incentive mechanism designs in asynchronous federated learning (FL) systems. With the consideration of unique characteristics inherent FL, such as dynamic participating and multiminded IoT nodes mobile users (MUs), requirements model training (i.e., accuracy convergence time), limited uplink bandwidth, formulate considered system an online design problem, where each MU is not only a buyer for communication resource but also seller computation service. To address...

With the rapid growth of number Internet Things (IoT) devices, support for massive access is becoming vital future low power wide area network (LPWAN). Up to now, LoRa™ and NB-IoT are two leading LPWAN technologies that use very distinct random (RA) schemes, is, grant-free RA scheme grant-based scheme, respectively. The low-overhead LoRa makes it suitable light traffic load, while more moderate load as could provide high reliability in data-transmission stage active terminals without...

Cardiac arrhythmia (CA) is a severe cardiac disorder that results in significant number of fatalities worldwide each year. Conventional electrocardiography (ECG) devices are often unable to detect symptoms during patients’ hospital visits due their intermittent nature. This paper presents wearable ECG processor for detection. The utilizes Hilbert transform-based R-peak detection engine detection, Haar discrete wavelet transform (HDWT) unit feature extraction, and Hybrid classifier combines...

To address the shortcomings of wireless data transmission in aquaculture water quality monitoring, we have developed an intelligent detection device based on STM32 microcontroller system as processing unit and TurMass technology. This can detect real-time transmit to user end for monitoring so that users monitor real-time. Through rigorous systematic experimental measurements, accuracy detected by was found be extremely high, indicating it used practical applications.

Pansharpening refers to the fusion of a panchromatic (PAN) image with high spatial resolution and multispectral (LRMS) low obtain (HRMS) image, which is beneficial visual display geographic research. Recently, many deep learning (DL) methods have been proposed address pansharpening problem, but still few examples DL-based techniques are designed from perspective better receptive field while scale features greatly varies among different ground objects. In this paper, we mainly focus on...

Pansharpening, which attempts to obtain a high-resolution multispectral (HR-MS) image by fusing panchromatic (PAN) with low-resolution (LR-MS) image, is critical yet difficult remote sensing processing task. In this study, we present novel convolution operation, self-guided spatial-channel adaptive (SSCAConv), for pansharpening. Unlike the reported convolutions that only focus on spatial details, our SSCAConv also considers channel specificity generating an individual kernel each patch...

Despite significant progress has been made in image deraining, existing approaches are mostly carried out on low-resolution images. The effectiveness of these methods high-resolution images is still unknown, especially for ultra-high-definition (UHD) images, given the continuous advancement imaging devices. In this paper, we focus task UHD and contribute first large-scale deraining dataset, 4K-Rain13k, that contains 13,000 pairs at 4K resolution. Based conduct a benchmark study processing...

Massive multiple input output (MIMO) enabled grant-free random access (mGFRA) stands out as a promising (RA) solution, thus effectively addressing the need for massive in machine-type communications (mMTCs) while ensuring high spectral efficiency and minimizing signaling overhead. However, bottleneck of mGFRA is mainly dominated by orthogonal preamble collisions, since pool small fixed-sized. In this paper, we explore potential non-orthogonal preambles to overcome limitations enhance success...

<div class="section abstract"><div class="htmlview paragraph">The advent of the low-altitude economy represents a novel economic paradigm that has emerged in recent years response to technological advancement and an expanding social demand. The is currently undergoing period rapid development, which underscores importance ensuring safety airfield operations. To enhance operational efficiency, unmanned aerial vehicles (UAVs) can be utilized for inspection surrounding area, runway...

The purpose of this study was to design an unmanned patrol service in combination with artificial intelligence technology solve the problem underground vehicle patrol. This used Raspberry Pi development board, L298N driver chip, camera, and other major hardware equipment transform remote control car. Python as programming language. By writing code, car could be driven under computer keyboard camera turned on for data collection. Keras neural network library quickly build a model, collected...