I have a number of questions regarding EEG signal analysis. Even after looking in a number of websites I am confused regarding it's usage. Please suggest with clear explanation.
1) Suppose, I have an eeg data with 100 Hz sampling frequency, recorded for 30 seconds. This means that we have 30 * 100 = 3000 samples of data. For these 3000 samples of data, in order to plot it's graph against time, how do I calculate the time data from this ? Do I simply divide the sample points by 3000 which makes (time_at_1_sec = 1/3000, time_at_2_sec = 2/3000 etc) ? Or Do i divide the sample points by 100 (time_at_1_sec = 1/100, time_at_2_sec = 2/100 etc) And continue with (1+1/100) after it reaches 100 sample points ?
2) While calculating alpha, beta, delta bands power, do we need to average the final value of power spectrum from all the channels of eeg data ?
3) After cleaning the noise using FFT and obtaining the final spectra, how to separate the individual band power from the final data ? Is there any function ?
I am not into this field but at least I could answer the first question for the time being :)
the time axis:
t = (sample index/total number of samples) * total sampling time
or, t = sample index / sampling rate
so for example, the sample #300 occurs at t = 300/100 = 3 s
I hope you get a complete answer to your questions as soon as possible.
1. t_sec = 1/Fs * (sample index);
2. averaging is one of the methods and you can also use other statistical methods such as maximum likelihood
After cleaning the noise using FFT and obtaining the final spectra, how to separate the individual band power from the final data ? Is there any function ?
Ans:
After pre processing the EEG data , use wavelet transform with different levels , so that our required low frequency information can be extracted. further used for analysis
You can see video: http://links.exponenta.ru/c/cPo/bPg/DsZ61-DN7hu7Di4DxgY2A3/b/X3Y/F/5892613a
1) Dividing the sample points by the sampling rate gives you the exact time in seconds.
2)Alpha,beta, delta and theta bands are calculated based on moving average bins with an overlap. If your sampling rate is 100 then you may average it as 1 to 100, 50 to 150 and take the power of all the bands in this span. The overlap percentage may be chosen as per the requirements.
3) There are methods like filter bank construction using simple band pass filters and also using wavelet decomposition. This gives you the filter banks based on your sampling rate. Better have a sampling rate of 256Hz to implement that.
- Badges
- Science topic