Signal classification algorithm in motor imagery based on asynchronous brain-computer interface

• Published in: IEEE International Instrumentation and Measurement Technology Conference (Online) pp. 1 - 5
• Main Authors: Jiang, Yu, He, Jingyan, Li, Dandan, Jin, Jing, Shen, Yi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2019
• Subjects: Associated common spatial patterns; Asynchronous brain-computer interface; Brain-computer interfaces; Electroencephalography; Feature extraction; Motor imagery; ROC curve; Spatial filters; Synchronous motors; Task analysis; Training
• ISSN: 2642-2077
• DOI: 10.1109/I2MTC.2019.8826883

In asynchronous Brain-Computer Interface (BCI), the subjects imagine autonomously to control devices via the system analyzing the electroencephalogram (EEG) recorded from the scalp. The continuous EEG of motor imagery (MI) based asynchronous BCI is processed, associated spatial filter is constructed to extract the EEG features. The classification result at each time is combined with the results of neighboring period, preventing the instabilities of the final classification resulting from short durations of the imagery tasks. When extended to the condition of two kinds of events for purpose of practical use, it reaches that in asynchronous BCI, detection sensitivity of left-right hand MI is above 90%, classification accuracy is above 83%, and false positive rate is below 32%; detection sensitivity of hand-foot MI is above 87%, classification accuracy is above 91%, and false positive rate is below 44%, which can guide the practical use in asynchronous BCI classification.