Word-specific properties affect classification performance in Brain Computer Interfaces for decoding imagined speech...

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCI) show promise for decoding imagined speech.
• Current BCI applications face challenges with classification performance in real-world scenarios.
• Word-specific properties are known to modulate neural signals during speech processing.

Purpose of the Study:

• To investigate the effect of word-specific properties, specifically age of acquisition (AoA) and word frequency, on speech imagery (SI) classification performance.
• To determine if these properties influence the accuracy of distinguishing SI from an idle state.

Main Methods:

• Utilized 16 word prompts varying in AoA and word frequency.
• Employed a random forest classifier with 10-fold cross-validation.
• Compared classification performance between speech imagery (SI) trials and the idle state.

Main Results:

• Demonstrated highly significant effects of AoA and word frequency on classification performance.
• Found a significant interaction between AoA and word frequency impacting accuracy.
• Confirmed that word properties significantly influence classification accuracy in SI paradigms.

Conclusions:

• Word frequency and AoA are critical factors influencing classification accuracy in BCI applications for imagined speech.
• Selecting word prompts with optimized properties can substantially improve BCI performance.
• This finding has direct relevance for enhancing the efficacy of BCI systems in real-world applications.