A multi-target brain-computer interface based on code modulated visual evoked potentials

• Published in: PloS one Vol. 13; no. 8; p. e0202478
• Main Authors: Liu, Yonghui, Wei, Qingguo, Lu, Zongwu
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 17.08.2018; Public Library of Science (PLoS)
• Subjects: Accuracy; Adult; Algorithms; Binary codes; Biology and Life Sciences; Brain; Brain research; Brain-Computer Interfaces; Computer and Information Sciences; Computer applications; Correlation; Correlation analysis; Correlation coefficient; Correlation coefficients; Electroencephalography; Engineering; Engineering and Technology; Evoked Potentials, Visual - physiology; Feasibility studies; Female; Frequency spectrum; Human-computer interface; Humans; Implants; Information transfer; Medicine and Health Sciences; Methods; Models, Neurological; NMR; Nuclear magnetic resonance; Pseudorandom binary sequences; Research and Analysis Methods; Signal to noise ratio; Target groups; Target recognition; User training; Visual evoked potentials; China
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0202478

The number of selectable targets is one of the main factors that affect the performance of a brain-computer interface (BCI). Most existing code modulated visual evoked potential (c-VEP) based BCIs use a single pseudorandom binary sequence and its circularly shifting sequences to modulate different stimulus targets, making the number of selectable targets limited by the length of modulation codes. This paper proposes a novel paradigm for c-VEP BCIs, which divides the stimulus targets into four target groups and each group of targets are modulated by a unique pseudorandom binary code and its circularly shifting codes. Based on the paradigm, a four-group c-VEP BCI with a total of 64 stimulus targets was developed and eight subjects were recruited to participate in the BCI experiment. Based on the experimental data, the characteristics of the c-VEP BCI were explored by the analyses of auto- and cross-correlation, frequency spectrum, signal to noise ratio and correlation coefficient. On the basis, single-trial data with the length of one stimulus cycle were classified and the attended target was recognized. The averaged classification accuracy across subjects was 88.36% and the corresponding information transfer rate was as high as 184.6 bit/min. These results suggested that the c-VEP BCI paradigm is both feasible and effective, and provides a new solution for BCI study to substantially increase the number of available targets.