EEG-based Classification of Imaginary Mandarin Tones

Speech imagery based brain-computer interface (BCI) has the potential to assist patients with communication disorders to recover their speech communication abilities. Mandarin is a tonal language, and its tones play an important role in language perception and semantic understanding. This work studied the electroencephalogram (EEG) based classification of Mandarin tones based on speech imagery, and also compared the classification performance of speech imagery based BCIs at two test conditions with visual-only and combined audio-visual stimuli, respectively. Participants imagined 4 Mandarin tones at each condition. Common spatial patterns were applied to extract feature vectors, and support vector machine was used to classify different Mandarin tones from EEG data. Experimental results showed that the tonal articulation imagination task achieved a higher classification accuracy at the combined audio-visual condition (i.e., 80.1%) than at the visual-only condition (i.e., 67.7%). The results in this work supported that Mandarin tone information could be decoded from EEG data recorded in a speech imagery task, particularly under the combined audio-visual condition.