Hanyang University announced on the May 18th that a team of Professor Im Chang-hwan of the Department of Biomedical Engineering developed a virtual electroencephalogram with artificial intelligence (AI) technology that is used for image transfer or voice modulation. 

When using the technology, it is expected to be utilized in many different brain engineering fields as it can significantly improve the Brain-Computer Interface (BCI) performance that makes external devices controlled only by thoughts.

The neural style transfer technology is once spotlighted through the application of deep learning and art in 2015, and it means a technology that learns artistic styles from pieces by van Gogh and Rembrandt and modifies random images to paintings as if the aforementioned authors drew them. 

When the technology is applied to voice signals, modifying your own voice to a certain celebrity's voice is possible. The neural style transfer technology was conventionally used for image and voice transfer and has never been applied to an electroencephalogram, an electric signal developed in the brain.

Professor Im's team suggested a new neural network 'S2S-StarGAN' that enables electroencephalogram transfer based on the StarGAN model, used for image transfer fields. Professor Im's team succeeded in transferring a short 16-second electroencephalogram signal that is measured in a resting state to 'steady-state visually evoked potentials' (SSVEP), a special electroencephalogram. SSVEP is an electroencephalogram generated at the occipital lobe when one stares at a visual stimulus that blinks regularly, and it is most widely used in the BCI field.

Professor Im's team recognized that when using SSVEP-based BCI that was made by the S2S-StarGAN model, one can improve the accuracy by 3.4% on average and a maximum of 10% higher than the previous methods. Also, comparing that it required an individual data collection process that takes more than 5 minutes before using the BCI system, when using the new technique, the BCI using efficiency will be largely improved as one needs to collect electroencephalograms for around 15 seconds. 

Professor Im said, "Korea's BCI technology is already world-class. As the world is interested in the BCI field due to Neuralink of Elon Musk, when enough support is guaranteed, Korea's BCI research can lead the world technologies."

Professor Im is Korea's first BCI researcher and published in around 200 international journals including succeeding in communicating with completely locked-in syndrome (CLIS) patients, whose communication with the external world is completely disconnected, with electroencephalogram in 2019.

This research result, supported by the SW Computing Industry Source Technology Development Project and the Artificial Intelligence Graduate School Supporting Project by the Institute for Information & Communication Technology Planning & Evaluation (IITP), was posted online to an international journal in the AI application field, 「Expert Systems With Applications」on the 13th.