This project demonstrates a proof of concept for developing a means to remove the wires from Electroencephalograph (EEG) Brain to Computer Interface (BCI) systems while maintaining data integrity and increasing the speed of transmission. This paper uses Machine Learning techniques to develop an Encoder/Decoder pair. The Encoder pair learns the important information from the analog signal, reducing the amount of data encoded and transmitted. The Decoder ignores the noise and expands the transmitted data for further processing. This paper uses one channel from an EEG-BCI system and organizes the analog signal in 500 datapoint frames. The Encoder reduces the frames to 75 datapoints and after noise injection, the decoder successfully expands them back to virtually indistinguishable frames from the originals.
Fanfan, Ernst; Blankenship, Joe; Chakravarty, Summit; and Randolph, Adriane B., "Using Machine Learning Techniques to model Encoder/ Decoder Pair for Non-invasive Electroencephalographic Wireless Signal Transmission" (2022). SAIS 2022 Proceedings. 38.