Transfer Learning for Brain-Computer Interfaces in Cyber-Physical Systems.

Brain-Computer Interfaces (BCI) have emerged as a promising technology for enabling seamless human-machine commnication. They bypass the need for explicit effort from a user and thus are apt as interfacing technology for wearables, IoT systems, body-area networks, etc. On account of this, BCIs have seen a steady rise in entertainment, health and wellness, and security applications. However, despite their growing popularity, they suffer from a key problem that prevents them from being a ubiquitous communication modality. BCIs often generalize poorly to new users owing to brain signals exhibiting a lot of variability among individuals. This makes creating standardized BCI models a challenge as they need to be re-trained/calibrated for each new end-user. In this paper, we present a method that not only shrinks this re-calibration overhead to a tiny fraction of the original but also provides a generalization accuracy that approaches that of intra-user models. We use transfer learning to transfer relevant parts of our detection model using divergence calculations in the signal probability space and use few-shot learning to adapt the model to a new user. We demonstrate our method on a dataset collected in our lab and obtain state-of-the-art results.