Donghao Lv

Convolution kernel compensation (CKC) is often limited by poor noise robustness, weak tracking of time-varying features, and low stability in the decomposition of non-stationary surface electromyogram (sEMG) signals. To address these limitations, a dynamic manifold projection spatiotemporal-anchored weighted-gated CKC (DPSW-CKC) was proposed. A dynamic manifold gating mechanism was introduced. Informative components were adaptively distinguished from noise. A spatiotemporal anchoring strategy was incorporated. Stable tracking of motor unit action potential features in non-stationary signal was thus achieved. In addition, a peak-detection pre-screening scheme was adopted. Invalid warnings in conventional implementations were effectively suppressed. On this basis, a complete framework for non-stationary sEMG decomposition was established. Experiments were conducted on a simulated sEMG dataset. The mean peak-to-noise ratio reached 22.51 dB. The mean silhouette coefficient reached 0.99. The coefficient of variation was reduced to 16.80%. These results indicate that DPSW-CKC outperformed conventional CKC variants. Time-varying action potential features were reliably captured. Its engineering applicability was also validated.

Muscle fatigue has a major negative impact on human bodily health. Surface electromyography (sEMG) is widely used for fatigue assessment because of its noninvasive nature. The current methods of signal decomposition are limited in extracting fatigue signals. Methods for feature extraction often process leak information and lack sensitivity, making it difficult to evaluate fatigue. We propose a method to assess muscle fatigue using the combination of adaptive frequency evolution decomposition (AFED) and fractional fuzzy fluctuation-based dispersion entropy (F3DE). AFED dynamically adjusts bandwidth to track spectral changes, extracting fatigue signals from sEMG data. F3DE is a fluctuation-based dispersion entropy (FDE) that utilizes fuzzy functions and fractional order to improve sensitivity and robustness in the characterization of fatigue. Based on data collected from various subjects performing isometric contractions, we see that AFED can extract fatigue signals. In the feature fitting results, extracted features show improvements in both the slope and ${R} ^{2}$ compared to the original sEMG signals. F3DE demonstrates better accuracy and stability in capturing individual fatigue processes. It outperforms traditional features in fatigue recognition. In classification, the AFED and F3DE models achieved 97.5% accuracy and excellent performance in all metrics. In conclusion, the combination of AFED and F3DE improves the accuracy and robustness of muscle fatigue detection.

In the context of resource constrained embedded terminals, the current deep learning-based technology for detecting surface defects in strip steel faces challenges with low detection accuracy and efficiency. This study introduces a lightweight convolution neural network (CNN) model for detecting surface defects of strip steel, which is based on deep learning. Through the integration of non-parameter attention mechanism SimAM, inverse residual structure, and depth-wise convolution module, a lightweight attention mobile backbone network is developed to achieve optimal feature extraction. To address the challenge of detecting small surface defects the context augmentation module is introduced to provide more information for small defects detection by using multi-scale features. To improve the efficiency of feature fusion and reduce parameter redundancy, the GVFPN neck network is proposed. The network aims to represent and deal with multi-scale features objectively while minimizing costs. The experimental results indicate that on the NEU-DET dataset, the proposed network attains an mAP of 78.8% and FPS at 97 frame/s. Moreover, the model requires only 2.39 M parameters and 3.1 G FLOPs. Compared to YOLOv5s, the proposed network reduces parameters by 65.9% and FLOPs by 80.7%, while achieving a 1.3% higher mAP and a 30 frame/s increase in detection speed. These results effectively demonstrate the effectiveness and superiority of the proposed method.

Parameter selection in Variational Mode Decomposition (VMD) often faces challenges where existing empirical methods, trial-and-error approaches, or single-index optimization fail to achieve global optimal performance or are limited to specific signal domains. To tackle this issue, this paper proposes an approach based on an Improved Hippopotamus Optimization Algorithm (IHO) for optimizing VMD parameters. To comprehensively address the characteristics of various signals, a quality factor composed of relevance, permutation entropy, and distinctiveness is employed as the objective function, optimized using the Hippopotamus Optimization Algorithm (IHO) algorithm for VMD parameter tuning. Results demonstrate that IHO significantly enhances VMD parameter optimization compared to the original Hippopotamus Algorithm (HO). Across different signal types, IHO-VMD exhibits superior performance in gear fault detection compared to Ensemble Empirical Mode Decomposition (EEMD), and in surface electromyography (sEMG) signal denoising, it outperforms Second-Generation Wavelet Soft Thresholding methods, highlighting its universality and effectiveness.

This study introduces a novel data augmentation method aimed at addressing the limited availability of sample data for strip steel surface defects, utilizing an improved Auxiliary Classifier Generative Adversarial Network (ACGAN). Firstly, on the generator side, the improved ACGAN integrates an encoder that extracts essential features from images of strip steel surface defects and transforms them into latent vectors. These vectors serve as inputs for the generator, helping it learn the actual data distribution of strip steel defects more accurately. To enhance image resolution and reduce checkerboard artifacts common with transposed convolution, the generator uses a combination of transposed convolution and PixelShuffle for upsampling. Secondly, on the discriminator side, the convolutional network employs an alternating configuration of downsampling and standard convolutional layers, effectively reducing feature dimensions while preserving more local information. A Dropout regularization strategy is applied after each convolutional layer to boost the model's robustness. Finally, the stability of the learning process is reinforced by employing Kaiming initialization for both the generator and discriminator's convolutional and transposed convolution layers. Performance evaluations conducted using the Northeastern University surface defect database confirm the efficacy of the proposed method in generating accurate images of strip steel surface defects in small-sample scenarios. Notably, compared to the original ACGAN and other classical generative adversarial networks, the improved method demonstrated significant improvements, achieving reductions of 52.82% and 38.78% in FID and MMD metrics, respectively.