Multi-Resolution Wavelet Fractal Analysis and Subtask Training for Enhancing Few-Shot Noisy Brainwave Recognition

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Internet of Things (IoT) integration revolutionizes healthcare monitoring.
• Portable electroencephalography (EEG) offers convenience but faces challenges with noisy data and inter-subject variability.
• Existing methods struggle with generalizability due to signal variability and insufficient training data.

Purpose of the Study:

• To develop a novel framework for enhancing EEG-based recognition in IoT healthcare monitoring.
• To address challenges of noisy samples, high inter-subject variability, and poor generalizability in EEG data.
• To improve the accuracy and robustness of brainwave detection from real-time, variable EEG signals.

Main Methods:

• Implemented a multi-resolution data analysis framework using wavelet fractals for feature extraction.
• Utilized continuous wavelet transform (CWT) and recombination to expand original data and augment training samples.
• Employed a deep learning (DL) subtask learning approach, incorporating wavelets at various scales for efficient model generalization.

Main Results:

• The proposed DL-based method effectively extracts features from small-scale and noisy EEG data.
• Demonstrated significant improvements in healthcare monitoring performance.
• Successfully mitigated the impact of external noise on EEG signal analysis.

Conclusions:

• The novel framework enhances EEG recognition accuracy in IoT healthcare settings.
• The approach effectively handles noisy and variable EEG data, improving generalizability.
• This method offers a robust solution for advanced brainwave detection in real-world healthcare applications.