Detecting event-related recurrences by symbolic analysis: applications to human language processing

Area of Science:

• Neuroscience
• Complex Systems Analysis
• Dynamical Systems Theory

Background:

• Quasi-stationarity is a common feature in complex dynamical systems, including brain activity.
• Event-related potentials (ERPs) are known to exhibit quasi-stationary states.
• Detecting these states from time series data is crucial for understanding brain dynamics.

Purpose of the Study:

• To elaborate a recent approach for detecting quasi-stationary states in time series data.
• To address challenges in recurrence analysis, specifically optimizing neighborhood size and symbol identification for sequence alignment.
• To identify quasi-stationary brain states within single-subject ERPs.

Main Methods:

• Utilizing recurrence analysis and symbolization methods to detect quasi-stationary states as recurrence domains.
• Applying a maximum entropy criterion to optimize the size of recurrence neighborhoods.
• Employing a Hausdorff clustering algorithm for identifying symbols across different realizations for sequence alignment.

Main Results:

• Developed a robust method for detecting quasi-stationary states in time series data.
• Successfully optimized key parameters in recurrence analysis for improved accuracy.
• Generated recurrence domains as partition cells that effectively represent quasi-stationary brain states in ERPs.

Conclusions:

• The proposed method provides an effective way to identify and characterize quasi-stationary brain states from ERP data.
• This approach enhances the analysis of complex dynamical systems, particularly in neuroscience.
• The findings contribute to a deeper understanding of brain dynamics and quasi-stationary phenomena.