Feature selection in classification of eye movements using electrooculography for activity recognition.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Activity recognition is crucial for applications like patient monitoring and human-computer interfaces.
• Effective feature selection is vital for accurate data mining and machine learning models.
• Electrooculography (EOG) signals offer a rich source for analyzing eye movements in human-computer interactions.

Purpose of the Study:

• To investigate the efficacy of Differential Evolution (DE) for selecting informative features from EOG signals.
• To improve the accuracy of activity recognition by optimizing feature subsets.
• To explore various feature extraction techniques including clearness-based and minimum redundancy maximum relevance (mRMR) features.

Main Methods:

• Utilized Differential Evolution (DE), an efficient evolutionary algorithm, for feature selection.
• Analyzed electrooculography (EOG) signals to extract eye movement-related features.
• Compared DE-based features with clearness-based and mRMR features for classification.

Main Results:

• Differential Evolution (DE) demonstrated effectiveness in identifying informative features from EOG data.
• The proposed DE-based feature selection approach led to improved classification performance for activity recognition.
• Enhanced accuracy in recognizing faultless activities was achieved through optimized feature subsets.

Conclusions:

• Differential Evolution (DE) is a powerful tool for feature selection in EOG-based activity recognition.
• Optimized feature selection using DE significantly enhances classification accuracy.
• The study highlights the potential of DE for developing robust and reliable activity recognition systems.