Are these steps correct for analysis of eeg signal ?
1) Import the data in matlab.
2) Filter out the signal using bandpass filter (1-4 Hz for delta, 8 - 13 Hz for alpha etc.)
3) Create Welch PSD or use FFT to create the spectra and get the peak power in the desired alpha, delta etc. ranges.
Please help me to clear out and confirm these steps. Thank you
These are more like random signals so suggest:
1 Import data say into Scilab or Matlab
2. Then use Psd eg Burg method ( ref
chapter 14 of Proakis DSP)
Good luck
Cheers
Hi Rasa, what you describe should work.
You could also consider wavelet transforms.
You may find attached pdf on a similar topic on psd analysis
useful.
Cheers
There are 3 main sources of information that can be used to extract features from EEG signals: 1. Spatial information, 2. Spectral (frequency) information, 3. Temporal information.
It also depends on your approach of analysis whether you want a multi channel approach or a single channel approach.
One of the most widely used method to analyze EEG data is to decompose the signal into functionally distinct frequency bands, such as delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–100 Hz) using a filter bank. You can use multichannel methods like CCA as well.
Hello Rasa, Swati Banerjee is right in proposing a structured bottom-up approach, which is proven to work.
1. Using signal processing which works on EEG
In each of the bands described in her contribution, you can perform a mathematically easy to understand signal processing, which takes away "permanent" components and focuses better on significant events.
Such body of knowledge has been established since the 1970s in the most studied signal, which happens to be biological, and driven by the human mind/brain by the way: speech. Speech has been studied in all directions because of the Telephony business. I have been involved in the later stage of this (see my publications on speech coding, detecting events in speech, like glottal closure, which is a biologically significant event).
When I was doing my speech focused PhD research, I also experimented my algorithms and methods on other biological signals beyond speech: ECG (electrocardiogrammes), EMG (electromyogrammes) and EEG. Surprise, suprise, my algorithms worked even better on ECG, EMG, EEG than on speech. Maybe because speech is a very elaborate human construction, language, and it has the depth of human complexity at its core, whereas a heart, well is "just for beating" (beating the right beat, and don't forget to breathe...), well that would be naive of course...
The EEG is very tricky.
To give you a flavour of models likely to work: autoregressive models , transforming s(n) into s'(n)= s(n)- sum( a(i)*s(n-i) for i=1 to N). You project the signal on its past values... You tune the a(i) i=1...N to minimise the residual energy (sum of squares of s'(n)).
2. Specific signal processing methods developed for EEG
Now there is a specific body of knowledge, of use in all areas of signal processing, but which emerged first with brain research, EEG related.
The brilliant concept can be found in details in the seminal article by Bandt-Pompe on permutation entropy ("Permutation entropy — a natural complexity measure for time series"). The idea of permutation entropy is simple: for a signal v(n), how many steps do you need to re-order the values of v(0), v(1),..v(n) as w(0), w(1)... w(n) such that w(i)
Using adaptive sensors we may extract any bio signal,extraction is a feature of measure of different energy level within short interval
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts