Classification of Phonological Categories in Imagined Speech using Phase Synchronization Measure

Area of Science:

• Neuroscience
• Linguistics
• Biomedical Engineering

Background:

• Phonological categories are crucial for speech articulation.
• Brain-computer interfaces (BCIs) require distinct neural signals for effective use.
• Speech imagery-based BCIs face challenges in identifying suitable prompts.

Purpose of the Study:

• To investigate if phonological categories of imagined speech prompts elicit different cortical phase synchronization patterns.
• To determine if electroencephalography (EEG) data from imagined speech can be classified based on phonological categories.
• To assess the efficacy of neural network (NN) classifiers using Mean Phase Coherence (MPC) versus statistical parameters.

Main Methods:

• EEG data was collected during imagined speech of nasal and bilabial consonants.
• Mean Phase Coherence (MPC) was computed to measure phase synchronization across cortical regions.
• Shallow neural networks (NNs) were trained to classify EEG data based on MPC values and statistical parameters.

Main Results:

• The NN classifier trained on beta-band MPC values achieved superior classification accuracy compared to alpha-band, gamma-band MPC, or statistical parameters.
• Distinct phase synchronization patterns were observed between imagined nasal and bilabial consonants.
• The study demonstrates the discriminative potential of phonological features in EEG during speech imagery.

Conclusions:

• Phonological categories, specifically nasal versus bilabial consonants, induce discernible differences in cortical phase synchronization during speech imagery.
• Beta-band phase synchronization is a robust feature for classifying imagined speech prompts.
• This research provides a foundation for selecting more effective speech prompts in BCI applications, enhancing assistive technologies for individuals with disabilities.