SASG-DA: Sparse-Aware Semantic-Guided Diffusion Augmentation For Myoelectric Gesture Recognition

Surface electromyography (sEMG)-based gesture recognition plays a critical role in human-machine interaction (HMI), particularly for rehabilitation and prosthetic control. However, sEMG-based systems often suffer from the scarcity of informative training data, leading to overfitting and poor generalization in deep learning models. Data augmentation offers a promising approach to increasing the size and diversity of training data, where faithfulness and diversity are two critical factors to effectiveness. However, promoting untargeted diversity can result in redundant samples with limited utility. To address these challenges, we propose a novel diffusion-based data augmentation approach, Sparse-Aware Semantic-Guided Diffusion Augmentation (SASG-DA). To enhance generation faithfulness, we introduce the Semantic Representation Guidance (SRG) mechanism by leveraging fine-grained, task-aware semantic representations as generation conditions. To enable flexible and diverse sample generation, we propose a Gaussian Modeling Semantic Sampling (GMSS) strategy, which models the semantic representation distribution and allows stochastic sampling to produce both faithful and diverse samples. To enhance targeted diversity, we further introduce a Sparse-Aware Semantic Sampling (SASS) strategy to explicitly explore underrepresented regions, improving distribution coverage and sample utility. Extensive experiments on benchmark sEMG datasets, Ninapro DB2, DB4, and DB7, demonstrate that SASG-DA significantly outperforms existing augmentation methods. Overall, our proposed data augmentation approach effectively mitigates overfitting and improves recognition performance and generalization by offering both faithful and diverse samples.

Enlightenment Period Improving DNN Performance

In the early stage of deep neural network training, the loss decreases rapidly before gradually leveling off. Extensive research has shown that during this stage, the model parameters undergo significant changes and their distribution is largely established. Existing studies suggest that the introduction of noise during early training can degrade model performance. We identify a critical "enlightenment period" encompassing up to the first 4% of the training cycle (1--20 epochs for 500-epoch training schedules), a phase characterized by intense parameter fluctuations and heightened noise sensitivity. Our findings reveal that strategically reducing noise during this brief phase--by disabling data augmentation techniques such as Mixup or removing high-loss samples--leads to statistically significant improvements in model performance. This work opens new avenues for exploring the relationship between the enlightenment period and network training dynamics across diverse model architectures and tasks.