Recognition of the idle state using individualized attention features for asynchronous brain-computer interfaces

Brain–computer interface (BCI), sometimes called a neural control interface or direct neural interface, is a direct communication pathway between an enhanced or wired brain and an external device. BCI provides an innovative pathway to help people suffering from motor disabilities, such as amyotrophic lateral sclerosis, spinal cord injury, and brainstem stroke. Technically, in a BCI system, electroencephalogram signals are usually used due to its non-invasive and inexpensive assay, ease of use, and acceptable temporal resolution. A review of published literature reveals that three types of brainwave-based models: i.e. event-related potentials, motor-imagery potentials, and steady-state visual evoked potentials (SSVEPs), are most frequently explored. Of these, the SSVEP-based BCI systems have been illustrated to having relatively high information transfer rate (ITR) and less training demand. As such, they are widely used for control of different robotic devices. In real-world applications, an asynchronous SSVEP system, which allows the subject to start and stop performing the mental task freely, is more flexible and natural to operate. However, the major difficulty of implementing the asynchronous BCI lies in recognizing the control state and idle state with high accuracy. Recent studies have devoted various efforts to model their electroencephalogram patterns.

Recent publications have demonstrated successful asynchronous control of the SSVEP systems; unfortunately, the number of false triggered commands ought to be reduced to avoid damages to the robot and the surroundings. In another study, the amplitude of the SSVEP signal has been shown to correlate with attention paid to the flickering stimuli. Moreover, it has been acknowledged that difference in attention level may be helpful to recognize the control and idle states. Nonetheless, the validity of recognizing the idle state using attention features has not been thoroughly investigated. In fact, in a previous study aimed at addressing this shortfall, an optimized complex network (OCN) method that yielded higher accuracy rate than the existing algorithms was developed. However, the performance of the OCN method was significantly influenced by the frequency band for filtering the EEG data. To address this, Yanshan University researchers: Mr. Wei Zhang, Dr. Tianyi Zhou, Assistant Professor Jing Zhao and Mr. Bolun Ji, in collaboration with Associate Professor Zhengping Wu at the Sanjiang University, China, proposed a novel Individualized Frequency Band based Optimized Complex Network (IFB-OCN) method to enhance the performance of discriminating the control and idle states. Their work is currently published in the research journal, Journal of Neuroscience Methods.

The main goal of the study was to individualize the filter band of the OCN method to enhance its classification performance. To realize these, the researchers adopted an experimental procedure consisting of two sessions: i.e. the control session and the idle session. The subject conducted an SSVEP task by focusing on the target stimulus in the control session, and watched the real-world robot feedback in the idle session. The performance of the proposed IFB-OCN method was estimated using classification accuracy and ITR with different data length.

Remarkably, the experimental results obtained showed that the IFB-OCN method achieved the highest average accuracy of 93.5 % with the data length of 4s, and achieved the highest ITR of 47.3 bits/min with the data length of 0.5 s. Compared with the standard OCN method utilizing no filter, the proposed IFB-OCN method delivered higher accuracies and information transfer rate regardless of data length.

In summary, the study demonstrated a novel IFB-OCN method designed to distinguish the idle and the control states in the asynchronous BCI system. In the proposed IFB-OCN method, the individualized frequency bands with the first three highest accuracies were selected for each subject, and their OCN features were extracted and integrated for classification. When compared with existing methods, the presented IFB-OCN method recognized the control and idle states using a single FPz channel rather than the occipital channels, and outperformed the existing algorithms in the accuracy of detecting the attention level. In a statement to Medicine Innovates, Dr. Jing Zhao said their proposed IFB-OCN method is efficient in recognizing the idle state and has a great potential for enhancing the asynchronous BCIs.