RESUMEN
Electroencephalogram (EEG) signals are pivotal in clinical medicine, brain research, and neurological disorder studies. However, their susceptibility to contamination from physiological and environmental noise challenges the precision of brain activity analysis. Advances in deep learning have yielded superior EEG signal denoising techniques that eclipse traditional approaches. In this research, we deploy the Retentive Network architecture - initially crafted for large language models (LLMs) - for EEG denoising, exploiting its robust feature extraction and comprehensive modeling prowess. Furthermore, its inherent temporal structure alignment makes the Retentive Network particularly well-suited for the time-series nature of EEG signals, offering an additional rationale for its adoption. To conform the Retentive Network to the unidimensional characteristic of EEG signals, we introduce a signal embedding tactic that reshapes these signals into a two-dimensional embedding space conducive to network processing. This avant-garde method not only carves a novel trajectory in EEG denoising but also enhances our comprehension of brain functionality and the accuracy in diagnosing neurological ailments. Moreover, in response to the labor-intensive creation of deep learning datasets, we furnish a standardized, preprocessed dataset poised to streamline deep learning advancements in this domain.
RESUMEN
In speaker recognition tasks, convolutional neural network (CNN)-based approaches have shown significant success. Modeling the long-term contexts and efficiently aggregating the information are two challenges in speaker recognition, and they have a critical impact on system performance. Previous research has addressed these issues by introducing deeper, wider, and more complex network architectures and aggregation methods. However, it is difficult to significantly improve the performance with these approaches because they also have trouble fully utilizing global information, channel information, and time-frequency information. To address the above issues, we propose a lighter and more efficient CNN-based end-to-end speaker recognition architecture, ResSKNet-SSDP. ResSKNet-SSDP consists of a residual selective kernel network (ResSKNet) and self-attentive standard deviation pooling (SSDP). ResSKNet can capture long-term contexts, neighboring information, and global information, thus extracting a more informative frame-level. SSDP can capture short- and long-term changes in frame-level features, aggregating the variable-length frame-level features into fixed-length, more distinctive utterance-level features. Extensive comparison experiments were performed on two popular public speaker recognition datasets, Voxceleb and CN-Celeb, with current state-of-the-art speaker recognition systems and achieved the lowest EER/DCF of 2.33%/0.2298, 2.44%/0.2559, 4.10%/0.3502, and 12.28%/0.5051. Compared with the lightest x-vector, our designed ResSKNet-SSDP has 3.1 M fewer parameters and 31.6 ms less inference time, but 35.1% better performance. The results show that ResSKNet-SSDP significantly outperforms the current state-of-the-art speaker recognition architectures on all test sets and is an end-to-end architecture with fewer parameters and higher efficiency for applications in realistic situations. The ablation experiments further show that our proposed approaches also provide significant improvements over previous methods.
