Human is a marvelously adaptive creature, and many of our capabilities remain hidden and not well understood. Bio-feedback generally refers to the training of our body, mind, heart to learn, adapt and control specific stimuli from any or a combination of our senses. Bio-feedback may have been achieved or learned consciously, but also quite often achieved subconsciously or unconsciously. Scientists had tested and observed our fellow human beings' marvelous ability to learn, adapt various sensory stimuli and turned them into controlled responses, quite often without being able to decipher how. One professor said "we still don't know how, but it works."
Bio-feedback can usually be set up as follows: a specific bio-signal input is selected (whether EEG, ECG, EMG, or SCR), and such bio-signal may be filtered to extract or emphasize specific attributes (say, alpha-wave, beta-wave, super-beta-wave, theta wave, or a weighted combination of them) and displayed back to the user (most often through visual sense, but can also through aural, tact senses.) After some period of learning (consciously or unconsciously), the user generally learns to "control" the output toward the desired direction. In many other cases, humans figured out unanticipated "short-cuts" to "beat the system" unconsciously, actually defeating the intended bio-feedback training. One professor who had worked on EEG bio-feedback for decades, lamented that "all he had accomplished were actually EMG training." This can happen easily because EMG signal is typically 10 to 100 times stronger than actual EEG signals, and the user may unconsciously learn (without any intention to "cheat") how to manipulate minute facial muscles to mimic or "trick" the bio-feedback program's algorithms, if such algorithms do not distinguish between EMG and EEG signals. With the emergence of easily accessible EEG gears, and EEG-based games, it's quite easy for the users to be "mis-trained" to utilize minute facial muscle activations as valid game input signals (supposedly EEG signals.) A more robust brain-wave input device must be able to differentiate between EMG and EEG signals, as they do have very different spectral characteristics.