Human ECoG analysis during speech perception using matching pursuit: a comparison between stochastic and dyadic...

Area of Science:

• Neuroscience
• Signal Processing
• Computational Auditory Neuroscience

Background:

• Human electroencephalography (EEG) is crucial for studying neural correlates of cognitive processes.
• Gamma activity (> 70 Hz) plays a role in speech perception but is challenging to detect.
• The matching pursuit (MP) algorithm is a signal processing technique for analyzing non-stationary signals.

Purpose of the Study:

• To apply the matching pursuit (MP) algorithm for detecting induced gamma activity in human EEG during speech perception.
• To evaluate the MP algorithm's efficacy in identifying subtle power changes at high gamma frequencies.
• To compare the performance and computational efficiency of stochastic versus dyadic dictionaries within the MP algorithm.

Main Methods:

• Utilized the matching pursuit (MP) algorithm on human EEG data recorded during speech perception tasks.
• Employed both stochastic and dyadic dictionaries for signal decomposition.
• Analyzed time-frequency power plots averaged over 100 trials to assess dictionary performance.

Main Results:

• The MP algorithm successfully detected induced gamma activity, particularly small power changes at high gamma frequencies (> 70 Hz).
• Time-frequency power plots generated by dyadic and stochastic MP dictionaries showed high similarity (> 98.5%) despite inherent frequency biases.
• The dyadic MP approach demonstrated significantly faster computational performance compared to the stochastic MP.

Conclusions:

• The matching pursuit (MP) algorithm is a valuable tool for analyzing high-frequency gamma activity in EEG during speech perception.
• Dyadic dictionaries provide a computationally efficient and accurate alternative to stochastic dictionaries for MP analysis in this context.
• These findings enhance the utility of EEG and MP for investigating neural dynamics of auditory processing.