Zero UI Part 1: The Science Behind Nissan’s Mind-controlled Car
At the last CES, Nissan unveiled its “B2V” (Brain-to-Vehicle) device, a headset designed to anticipate driver actions.
Nissan claim that they will be able to detect driver actions 300ms before they occur, allowing the vehicle to anticipate braking or turning actions.
After our resident Brain-computer Interface (BCI) expert analysed the footage, they concluded that this device functions by painlessly measuring the driver’s surface-EEG over their motor cortex and detecting an Event-related Desynchronization (ERD) – a disruption in brain activity caused by the motor cortex “booting up” out of sleep mode in preparation for physical activity.
Contrary to what you might think, the most important part of a BCI is its software. The hardware that is used to measure brain activity has plateaued in the last few decades, and though they are becoming cheaper and there are some variations in design, fundamentally everyone makes do with the same silver chloride electrodes used to obtain brain activity.
What makes the difference between a BCI working and not working is the software used to clean up and interpret the incredibly weak signals from the electrodes. Nissan will be using a proprietary combination of digital signal processing and machine classification techniques to decode and identify whether the ERDs it’s detecting come from the driver’s feet, or left or right arms, though the best-performing techniques are usually Wavelet Transform or Common Spatial Patterns in conjunction with Support Vector Machines or Neural Nets.
Wavelet Transform and Common Spatial Patterns are both Blind Source Separation (BSS) techniques. BSS is when you take a composite signal, and without knowing what its components look like, automatically split it into its original constituent parts – in this case brain activity and a whole heap of noise. Common Spatial Patterns and Neural Nets on the other hand are machine learning techniques where you train the computer to identify patterns until it constructs its own classifier.
Nissan aren’t the first company to use BCI to control a car, our own BCI expert having done it for two separate companies, but they may be the first to bring it to market. It's also part of a growing trend of Zero UI product design, something we'll be touching on in later blog posts.