The project should be reported in a concise but high standard research-like article (around 8 pages).
The report should be returned before the End of the Year.
I - Characterisation of the frequency content
- Specify the frequency content of the data using Fourier analysis. Keep in mind that the sampling frequency is 100 Hz.
- Filter out the 60 Hz power line and the base-line wander.
- Extend the analysis into time-frequency representation both in Fourier and Wavelet bases. Justify your choice of Wavelet basis. Explain the reasons for the differences between the two approaches.
II - Segmentation
Segment the data in portions having:
You may choose any segmentation method from the book. Justify your choice.
III - Artifact detection and removal
The MEG data (fs = 148.5 Hz) in MEG.mat has a number of artifacts in it. These include ocular (both blinking and horizontal saccades), cardiac and myographic contaminations. You still have a weak (but real) digital watch. Underneath it all, there is some beta activity (artificially added :-) ). Your task is to:
- Identify all the artifacts and alpha activity named above;
- Make use of any method studied in the book (in case of despair you can get one or two artifacts from your preferred bag of tricks);
- Use ICA for the detection of, at most, 3 artifacts, including the digital watch. (you will need to do some dimension reduction, explain why it is important)
- Bonus question: there may be some additional artifact that hasn't been mentioned... finding any will be acknowledged.