Development of SSVEP-BMI for controlling LEGO MINDSTORMS car

• ¹ Meiji University, Department of Electronics and Bioinformatics, Japan
• ² Meiji University, Electrical Engineering Program, United States

Introduction: LEGO Mindstorms provides variety of sensors and motors to be embedded and thus is promising for prototype modelling of brain-controlled robots. However, there is a lack of information on how to interface Mindstorms with brain-derived signals obtained from an EEG system. We developed Brain Machine Interface (BMI) based on Steady-state visual evoked potential (SSVEP) that controls a remote control car built with LEGO Mindstorms NXT (LEGO car) by use of g.tec g.USBamp bioamplifier. Methods: To send a command to the LEGO car, a user looked at one of the three LED visual stimuli blinking at different frequency of 6, 7, or 8Hz. EEG was continuously measured from 3 electrodes attached to the occipital area. Our in-house developed unsupervised SSVEP classifier determined the frequency of attended stimulus based on the frequency spectrum and the averaged waveform of EEG signals. Each of the 3 LED stimuli was assigned to one of the commands of moving forward, turning right, or turning left so that the user can control the LEGO car by looking at specific LED stimulus. Instrument Control Toolbox of MATLAB was employed to directly send commands to the LEGO car from the g.tec Simulink user interface via Bluetooth connection. The LEGO NXT has a wireless protocol called Direct Command to control the NXT from external Bluetooth devices. Using the Direct Command, we can activate/deactivate the motors and change their speed. To prevent unwanted operation during muscle activity or state without attention to any of the LED stimulus, we provided a mechanical button to the user and he/she could send a command to the LEGO car only when the button was pressed. Results and Discussion: The LEGO car system was tested by more than 20 novice BMI users who visited our open lab in the university festival held in November 2014. Approximately 75% of them could complete a run of U-shaped circuit (approximately requires 2 straight run, 1 turning right, 1 straight run, 1 turning right, and another 2 straight run to finish) after a short period of practice about 3 minutes. Incorporating LEGO Mindstorms in the g.tec EEG system has benefit for BMI researchers because they can concentrate on the BMI algorithms without being bothered by hardware configuration and can further integrate BMI and other sensor-based technologies such as computer vision with minimum effort. Introduction: LEGO Mindstorms provides variety of sensors and motors to be embedded and thus is promising for prototype modelling of brain-controlled robots. However, there is a lack of information on how to interface Mindstorms with brain-derived signals obtained from an EEG system. We developed Brain Machine Interface (BMI) based on Steady-state visual evoked potential (SSVEP) that controls a remote control car built with LEGO Mindstorms NXT (LEGO car) by use of g.tec g.USBamp bioamplifier. Methods: To send a command to the LEGO car, a user looked at one of the three LED visual stimuli blinking at different frequency of 6, 7, or 8Hz. EEG was continuously measured from 3 electrodes attached to the occipital area. Our in-house developed unsupervised SSVEP classifier determined the frequency of attended stimulus based on the frequency spectrum and the averaged waveform of EEG signals. Each of the 3 LED stimuli was assigned to one of the commands of moving forward, turning right, or turning left so that the user can control the LEGO car by looking at specific LED stimulus. Instrument Control Toolbox of MATLAB was employed to directly send commands to the LEGO car from the g.tec Simulink user interface via Bluetooth connection. The LEGO NXT has a wireless protocol called Direct Command to control the NXT from external Bluetooth devices. Using the Direct Command, we can activate/deactivate the motors and change their speed. To prevent unwanted operation during muscle activity or state without attention to any of the LED stimulus, we provided a mechanical button to the user and he/she could send a command to the LEGO car only when the button was pressed. Results and Discussion: The LEGO car system was tested by more than 20 novice BMI users who visited our open lab in the university festival held in November 2014. Approximately 75% of them could complete a run of U-shaped circuit (approximately requires 2 straight run, 1 turning right, 1 straight run, 1 turning right, and another 2 straight run to finish) after a short period of practice about 3 minutes. Incorporating LEGO Mindstorms in the g.tec EEG system has benefit for BMI researchers because they can concentrate on the BMI algorithms without being bothered by hardware configuration and can further integrate BMI and other sensor-based technologies such as computer vision with minimum effort.