運用觸覺刺激的腦機介面，干擾認知功能運作的程度較運用視覺或聽覺刺激的腦機介面為低，而在現實世界中有很高的應用價值。先前文獻對於觸覺型腦機介面能否達到與視覺型相當的動作心像類型辨識率尚無結論；此外，運用極少量腦波頻道數是否能達到高動作心像類型辨識率亦屬未知。本研究試圖建立一個腦機介面架構，僅運用四個腦波頻道資訊來進行動作心像類型辨識，然後比較其辨識觸覺和視覺誘發動作心像時之正確率。我們發現應用共同空間型態(common spatial pattern)作為特徵提取器，以及線性區辨分析(linear discriminant analysis)為分類器(classifier)，對觸覺與視覺誘發的左、右手動作心像區辨正確率是相當的；而僅運用C3、C4、Fp1、和Fp2四個頻道的情況下，觸覺與視覺型腦機介面的辨識正確率都可達到85%。本研究之發現可作為未來發展高效能動作心像辨識腦機介面系統之基礎。;Haptic-based Brain Computer Interface (BCI) has great values in real-world applications as it is less intrusive than visual or auditory based BCI. It is not clear from previous literature whether haptic-based BCI can achieve equivalent or even better accuracies when applied to the classification of motor imagery. In addition, it was also not clear whether high classification accuracy can be achieved in haptic cue based motor imagery BCI with few channels. The current study sets out to establish a BCI framework using only four channels for motor imagery classification, and to compare the accuracies between the haptic versus the visual cue based BCI. We demonstrated that using common spatial pattern (CSP) as feature extractor and linear discriminant analysis (LDA) algorithm as the classifier, the classification accuracy of the haptic cue based motor imaginary BCI can reach comparable level as the visual-based one. We also demonstrated that with only four EEG channels (C3, C4, Fp1, and Fp2), the mean accuracy of both haptic and visual cue based motor imaginary BCI can reach a high level (~ 85%). The current findings can serve as the foundation for efficient BCI implementation in motor imagery classification for future research and real-world applications.
