A machine learning approach to identify stochastic resonance in human perceptual thresholds

Area of Science:

• Cognitive Science
• Neuroscience
• Signal Processing

Background:

• Stochastic resonance (SR) enhances faint signals with optimal white noise, typically observed as a performance improvement curve in perceptual thresholds.
• Identifying SR is usually qualitative, relying on subjective visual examination of data by human judges, which is prone to error.
• Current methods for SR detection are subjective and lack quantitative rigor, hindering objective analysis.

Purpose of the Study:

• To develop and validate a quantitative, algorithmic method for detecting stochastic resonance (SR) in perceptual thresholds.
• To replace subjective, qualitative assessments with an objective, data-driven classification system.
• To improve the accuracy and reliability of SR identification in perceptual studies.

Main Methods:

• A logistic regression (LR) model was trained on engineered features derived from simulated SR performance data.
• The LR model was applied to 6 perceptual threshold test cases using parameters from experimental subject data.
• The algorithmic classification approach was compared against existing subjective methods for SR detection.

Main Results:

• The developed logistic regression algorithm quantitatively classified the exhibition of SR in perceptual thresholds.
• Algorithmic classifications demonstrated superior accuracy compared to subjective methods across 6 test cases (p < 0.05).
• The study successfully implemented and validated the algorithmic classification process.

Conclusions:

• Algorithmic classification provides an effective means to identify SR in perceptual thresholds.
• This quantitative approach offers a rigorous and objective alternative to subjective visual inspection.
• The findings establish a more reliable method for detecting stochastic resonance in scientific research.