This algorithm isolates and removes artifacts from electroencephalography (EEG) recordings in real-time, enabling the acquisition of clean data signals. EEG hardware is an essential tool for detecting brain waves in real-world mobile environments with high temporal resolution. Mobile brain imaging applications with EEG include neuro-rehabilitation, virtual reality entertainment, and brain-computer interfaces as assistive devices for everyday life. Unfortunately, the signals EEG measures are small, and movement-related artifacts and noise, such as eye blinking, muscle movement, and heartbeats, often corrupt the signals of interest. Recent improvements to the EEG hardware, such as using actively amplified electrodes, enable the recording of cleaner raw EEG signals, but motion-related artifacts remain a significant challenge for recording and analyzing brain activity during whole-body movement, limiting scientific research and potential commercial applications. The current gold standard for analyzing EEG data, independent component analysis, takes approximately 2 to 30 hours or requires a supercomputer, further limiting EEG applications.
Researchers at the University of Florida have developed a suite of cleaning algorithms for EEG data making identifying and removing latent and motion-related artifacts from EEG data in real time possible. These algorithms provide more accurate and reliable EEG data, including during movement, expanding the potential applications for EEG readings to previously inaccessible fields.
Real-time artifact-removing algorithms for obtaining clean electroencephalography (EEG) signals, even on moving subjects
These algorithms remove contaminating artifacts, including motion artifacts, muscle artifacts, and eye artifacts, from electroencephalography (EEG) data signals. The algorithms can run immediately after data collection, with data processing in seconds to minutes, or the algorithms running in real-time, with live cleaning of the EEG signals for real-time applications. The first algorithm leverages reference noise signals coming from concurrent noise recordings to identify and remove independent sources of noise activity corrupting the EEG signals. The second algorithm leverages pseudo-reference noise signals for applications where direct noise recordings are not feasible (only EEG signals are available). Both algorithms compare the corrupted EEG signals to reference noise signals in high-dimensional space and identify common noise components for subsequent removal.