Temporal Independent Component Analysis for Automatic Artefact Removal from EEG
|Course Name||Temporal Independent Component Analysis for Automatic Artefact Removal from EEG|
Upon successful completion of this module, you will be able to:
- Define the word “artifact”
- Identify what is meant by “Independent ComponentAnalysis (ICA)”
- Identify the function of a “Support Vector Machine (SVM)”
Course InformationOne of the most significant issues associated with EEG analysis is the high contamination of the recorded signals with various artifacts, both from the subject and from equipment interference. These artifacts must be removed prior to subsequent processing of the EEG. One of the more common methods recently employed for artifact removal is Independent Component Analysis (ICA).
Since many applications, such as Brain- Computer Interfaces (BCIs), require online and realtime processing of EEG signals, it is ideal if the removal of artifacts is performed in an automatic fashion. This paper discusses the application of a specific extension of ICA, Temporal Decorrelation Source Separation (TDSEP), for the problem of automatic artifact removal from EEG signals. The method makes explicit use of the temporal structure of the estimated sources in automating the process. A Support Vector Machine (SVM) is trained to classify the estimated sources into EEG and artifacts based on their auto-correlation structure. The performance of the method has been assessed on artificial mixtures of EEG and artifact signals (heart-beat and eye movements). High classification rates, in the order of >90%, were achieved for both the EEG and artifact signals. The method also appears to remain robust with inter-subject variability, with classification rates remaining at similar high levels.