EEG signals (potential from brain) contain 'noise' due to EMG signals (potential from muscles). I was thinking of using the EMG as a common mode feedback prior to the amplification of the EEG signal. Thus, I would assume the EMG signal (EEG 'noise) can be removed when it enters the differential amplifier.
I am just wondering how possible this is. I am aware that EMG signals are much larger than EEG signals.
No -- sounds like a good idea, but separating EEG and EMG is much easier said than done. How do you intend to get "pure" EMG to use as your common Mode signal, especially if it shares a freq spectrum with EEG?
The real way to separate emg and eeg is filtering. EEG is lower frequency than EMG. What isn't low frequency, though, is movement artifact. If the EMG activity is caused by movement, and the movement also causes motion in the electrode-skin interface, this can cause trouble.
In state-of-the-art implementations, the EEG amplifiers run on relatively low gain, since the signals are then acquired by high-resolution ADCs. So EMG doesn't saturate them. In addition to EEG, the EMG signals are sensed from dedicated electrodes and acquired in addition to EEG. You'll also want EOG electrodes to similarly suppress the EOG signal.
EOG and EMG go into an adaptive linear (usually) model that estimates transfer function from each of those unwanted sources to each EEG electrode. Then these signals are passed through the estimated transfer functions and can be subtracted from EEG signals. There are various fairly robust schemes that can estimate those transfer functions in real time, and adapt as conditions change.
