Comparing linear and quadratic models of the human auditory system using EEG.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• System identification using electroencephalography (EEG) is crucial for assessing human sensory processing.
• Linear methods like Visual/Auditory Evoked Spread Spectrum Analysis (VESPA/AESPA) are established for studying visual and auditory function.
• Previous research explored nonlinear VESPA but lacked a nonlinear AESPA counterpart.

Purpose of the Study:

• To investigate a nonlinear extension of the Auditory Evoked Spread Spectrum Analysis (AESPA) method.
• To quantify the contributions of linear and quadratic processes in EEG responses to auditory stimuli.
• To compare the predictive power of linear versus quadratic models for novel EEG data.

Main Methods:

• Developed and applied a nonlinear (quadratic) extension to the AESPA technique.
• Analyzed electroencephalography (EEG) data in response to novel auditory stimuli.
• Quantified the relative contributions of linear and quadratic signal components.

Main Results:

• The quadratic AESPA model demonstrated a statistically significant ability to predict novel EEG.
• The quadratic model (r=0.0418) showed a slight but significant improvement in prediction accuracy over the linear model (r=0.0361).
• While predictive accuracy was modest, the significance highlights the importance of nonlinear components.

Conclusions:

• A nonlinear (quadratic) extension of AESPA is feasible and provides valuable insights into auditory processing.
• Quadratic processes contribute significantly, albeit modestly, to EEG responses to auditory stimuli.
• This nonlinear approach advances the system identification of auditory sensory processing using EEG.