I try use SSVEP paradigm to simple experiments. Unfortunately, results are unsatisfactory. I recorded data from O1 and O2 electrodes.
1. without LED blinks (4 seconds)
2. 7Hz LED blinks (4 seconds)
3. 10Hz LED blinks (4 seconds)
4. 16 Hz LED blinks (4 seconds)
5. 23 Hz LED blinks (4 seconds)
Next, I use FFT to get spectrums of this signals, but there are no any peaks (on the selected frequencies) as in many articles about SSVEP. What am I doing wrong? Any good references?
Well there are several things that can be improved. I suggest you to use Oz and POz instead. Also using only the FFT is too simple, I suggest you to use the signal to noise ratio between the baseline (no led blinks) and the stimulation (led blink). Similar to what you are doing, at leas for the basic: http://journal.frontiersin.org/article/10.3389/fncir.2013.00027/full
There are several possibilities you need to consider in that case.
First, you need to check whether visual stimuli are correctly presented with high precision. How many trials did you record? If it was just a single trial, try to record more. Also, some subjects do not show strong peaks. So, testing more subjects would be better. Recording EEG signals with more channels (e.g., Oz and POz) are also recommended. SSVEP responses are somewhat dependent on electrode locations among different subjects.
I think that O1 and O2 are good choices, even Oz channel could be also considered. Note that the stimuli should be generated 50/50 on / off. Furthermore, i consider that some people presents some problems to recognize evoked potentials. Also, you can find some alternatives to classify SSVEPs as in this paper:
Conference Paper A Comparison of Techniques and Technologies for SSVEP Classification
- Badges
- Science topic