.png){kind=icon}
.png){kind=icon}
Return
Group Sparse and Super Resolution Time-Frequency-Based Method for Emotion Recognition
.png){kind=icon}
PRE
AI
DOI：10.1016/j.dsp.2025.105761.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A hybrid EEG denoising method is proposed by combining group sparse mode decomposition (GSMD) with the Bhattacharyya Distance. This method improves signal quality by effectively distinguishing noise from relevant brain activity. This method combines the strengths of statistical distance measures and morphological techniques to achieve more effective noise reduction in EEG signals. • The performance of emotion recognition is significantly improved by time-frequency images generated using superlet (SLT) and adaptive superlet transform (ASLT). The SLT and ASLT provide superior time-frequency localization compared to traditional techniques such as Short-Time Fourier Transform (STFT) or wavelet transforms, which may have fixed resolution or limited adaptability. • The study proposes a state-of-the-art Super Resolution Neural Network (SRNET). SRNET is trained on high-resolution time-frequency images. The SRNET is lightweight and end-to-end trainable on high-resolution Superlet-derived inputs. SRNET is efficient in extracting pinpoint features. • The SRNET achieves a superior classification accuracy of 99.63% while requiring less computational time, making it highly suitable for real-time brain-computer interface (BCI) applications.
Journal
D
IF:
3
Papers:
653
Citations:
0