Jiayan Huang

Objective. Miniature electrocardiogram (ECG) devices can rapidly and accurately acquire real-time cardiac signals, enabling timely warnings for patients with heart disease. To achieve accurate arrhythmia classification on resource-constrained ECG edge devices, we propose EdgeECG, an ultra-lightweight neural network designed for deployment on low-power microcontrollers. Approach. EdgeECG is first designed with a compact convolutional architecture to ensure compatibility with resource-limited embedded platforms. A dual criterion pruning (DCP) strategy is then introduced to evaluate weight importance using both magnitude and median deviation for more precise model compression. In addition, quantization is applied to reduce storage cost and improve deployment efficiency on an STM32F103 microcontroller. Main results. Experimental results on the benchmark MIT-BIH arrhythmia dataset show that EdgeECG achieves an overall classification accuracy of 98.07%, outperforming several representative methods. In patient-independent record-level evaluation, the accuracy reaches 80.73%. The model contains only 2680 parameters and is successfully deployed on an STM32F103 microcontroller with 64 KB SRAM, achieving an inference latency of 0.025 s and an energy consumption of 0.156 mJ per inference. In addition, DCP reduces the number of non-zero parameters by nearly 50% while maintaining promising classification performance. Significance. EdgeECG provides an effective solution for five-class ECG arrhythmia classification on severely resource-constrained edge devices. Its compact architecture, effective pruning strategy, and successful deployment on an STM32F103 demonstrate its potential for low-power edge-based ECG analysis and wearable cardiac monitoring applications. The source code of EdgeECG is publicly available at: https://github.com/jyanhuang/EdgeECG-code .

The rapid and accurate classification of lychee varieties is crucial for improving production efficiency and optimizing market supply. Especially for the main production areas of lychee, efficient lychee classification is more urgent. However, there is currently no publicly available comprehensive and diverse lychee benchmark dataset for precise training of classification models. To fill this gap, this work constructs a comprehensive lychee image dataset (Lychee13-3634), which covers 13 varieties and 3634 images. Different from the general fruit datasets, which show significant differences in features between their fruit images, Lychee13-3634 highlights minor inter-class differences among various lychee varieties. Based on this dataset, we applied 20 advanced deep learning-based classification models to validate its availability and effectiveness. Meanwhile, we comprehensively evaluated and provided meaningful insights about all models. Experimental results show that EfficientNetv2 has the best classification performance with an accuracy of up to 99.90%. Besides, we further comprehensively analyzed the balance of Lychee13-3634, and the corresponding experiments demonstrate that a more balanced dataset usually leads to better classification performance of the model. In summary, Lychee13-3634 provides benchmark training data for the lychee image classification task and demonstrates the effective application of existing deep learning classification models, providing reference and inspiration for other agricultural product image recognition research. Our Lychee13-3634 and all evaluation models are available at https://github.com/jyanhuang/Lychee13-3634 .

Driver drowsiness electroencephalogram (EEG) signal monitoring can promptly alert drivers of their drowsiness in edge-end Consumer Electronics (CE), thereby reducing the probability of traffic accidents. Graph convolutional networks (GCNs) have shown significant advancements in processing the non-stationary, time-varying, and non-Euclidean nature of EEG signals. However, the existing single-channel EEG adjacency graph construction process lacks interpretability, which hinders the ability of GCNs to effectively extract adjacency graph features, thus affecting the performance of drowsiness monitoring. To address this issue, we propose an edge-end Lightweight Dual Graph Convolutional Network (LDGCN). Specifically, we are the first to incorporate neurophysiological knowledge to design a Baseline Drowsiness Status Adjacency Graph (BDSAG), which characterizes driver drowsiness status. Additionally, to express more features within limited EEG data, we introduce the Augmented Graph-level Module (AGM). This module captures global and local information at the graph level, ensuring that BDSAG features remain intact while enhancing effective feature expression capability. To further enhance the application of CE, we apply Adaptive Pruning Optimization (APO) to the channels and neurons of LDGCN, which nearly halves the inference latency. Experiments on benchmark datasets demonstrate that LDGCN offers the best trade-off between monitoring performance and hardware resource utilization, outperforming existing state-of-the-art algorithms.

The detection of esophageal motility disorders serves as a crucial diagnostic indicator for preventing esophageal diseases. Currently, the swallowing cloud image obtained by high-resolution esophageal manometry is mainly used for esophageal motility analysis; however, manual analysis of swallowing cloud images presents numerous drawbacks, including time-intensive processes, labor-intensive procedures, and subjectivity in diagnosis. In response to these challenges, we propose an esophageal motility analysis model based on deep learning techniques to facilitate preliminary diagnosis of swallowing cloud images, thereby augmenting diagnostic efficiency for medical practitioners. Initially, we construct an esophageal motility dataset. Subsequently, we introduce a light weight model, termed PortableNet, derived from MobileNet architecture. The research findings demonstrate that the PortableNet model achieves a recognition accuracy of 93.99% with only 4.3 million parameters. In terms of classification performance, the model also achieves a precision of 94.73%, a recall of 93.39%, and an F1-score of 94.06%. Compared to the MobileNet architecture model, there is a 5.05% increase in recognition accuracy and a 58.65% reduction in size. These results collectively indicate that the PortableNet model delivers superior overall performance, offering both high accuracy and efficiency in recognizing esophageal motility swallowing cloud images. Such advancements can significantly enhance diagnostic efficiency among medical professionals, thereby offering valuable insights for future research endeavors.

Objective. Signal denoising methods based on deep learning have been extensively adopted on electroencephalogram devices. However, they are unable to deploy on edge-based portable or wearable (P/W) electronics due to the high computational complexity of the existed models. To overcome such issue, we propose an edge-based lightweight Kalman filter network (EKFNet) that does not require manual prior knowledge estimation. Approach. Specifically, we construct a multi-scale feature fusion module to capture multi-scale feature information and implicitly compute the prior knowledge. Meanwhile, we design an adaptive gain estimation module that incorporates long short-term memory and sequential channel attention module to dynamically predict the Kalman gain. Furthermore, we present an optimization strategy utilizing operator fusion and constant folding to reduce the model’s computational overhead and memory footprint. Main results. Experimental results show that the EKFNet reduces the sum of the square of the distances by at least 12% and improves the cosine similarity by at least 2.2% over the state-of-the-art methods. Besides, the model optimization shortens the inference time by approximately 3.3×. The code of our EKFNet is available at https://github.com/cathnat/EKFNet. Significance. By integrating Kalman filter with deep learning, the approach addresses the parameter-setting challenges in traditional algorithms while reducing computational overhead and memory consumption, which exhibits a good tradeoff between algorithm performance and computing power.

Underwater images often suffer from degradation due to scattering and absorption. With the development of artificial intelligence, fully supervised learning-based models have been widely adopted to solve this problem. However, the enhancement performance is susceptible to the quality of the reference images, which is more pronounced in underwater image enhancement tasks because the ground truths are not available. In this paper, we propose a light-enhanced pixel fusion network (LEPF-Net) to solve this problem. Specifically, we first introduce a novel light enhancement block (LEB) based on the residual block (RB) and the light enhancement curve (LE-Curve) to restore the cast color of the images. The RB is adopted to learn and obtain the feature maps from an original input image, and the LE-Curve is used to renovate the color cast of the learned images. To realize the superb detail of the repaired images, which is superior to the reference images, we develop a pixel fusion subnetwork (PF-SubNet) that adopts a pixel attention mechanism (PAM) to eliminate noise from the underwater image. The PAM adapts weight allocation to different levels of a feature map, which leads to an enhancement in the visibility of severely degraded areas. The experimental results show that the proposed LEPF-Net outperforms most of the existing underwater image enhancement methods. Furthermore, among the five classic no-reference image quality assessment (NRIQA) indicators, the enhanced images obtained by LEPF-Net are of higher quality than the ground truths from the UIEB dataset.