Context based error modeling for lossless compression of EEG signals using neural networks

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Lossless data compression is crucial for managing large datasets like electroencephalogram (EEG) signals.
• Traditional two-stage compression methods utilize predictors and encoders.
• Neural network predictors offer advanced capabilities for signal prediction.

Purpose of the Study:

• To investigate the application of context-based error modeling for neural network predictors in EEG signal compression.
• To evaluate the impact of error modeling on the efficiency of EEG compression algorithms.

Main Methods:

• Utilized context-based error modeling techniques applied to neural network predictors.
• Experimented with human EEG signals recorded under diverse physiological conditions.
• Analyzed the reduction in statistical redundancy within error signals post-prediction.

Main Results:

• Context-based error modeling significantly improved the compression efficiency of neural network-based predictive techniques for EEG signals.
• The bits per sample required for EEG compression with error modeling and entropy coding ranged from 2.92 to 6.62.
• Achieved a saving of 0.3 to 0.7 bits per sample compared to compression schemes without error modeling.

Conclusions:

• Context-based error modeling is an effective strategy for enhancing neural network predictors in EEG compression.
• The proposed method offers substantial improvements in compression efficiency, making it valuable for biomedical data management.