EMOTIV EPOC has 14 EEG sensors of which 8 are positioned around the frontal and prefrontal lobes, which by virtue of their location pick up signals from facial muscles and the eyes. Most EEG systems treat these signals as noise and they are filtered or ignored when interpreting the signals. The EMOTIV detection system also filters these signals out before interpreting the brain signals, however, we also use these signals to classify which muscle groups are causing them, we call this Smart Artifacts.
We have developed efficient classifiers to detect many facial expressions, including blink, left wink, right wink, raised eyebrows (surprise), furrowed brows (frown), smile and clenched teeth.
Our Facial Expressions are detected from muscle noise, except for eye rotations which arise from the fact that the eye is electrically polarised and the moving dipole also creates a detectable elecrical signal. It is a challenge with conventional EEG systems to eliminate the effects of muscle signal from the brain patterns and most medical EEGs require the patient to sit very still so they can see the brain signals with high enough integrity to diagnose functional problems (and even then much of the data is discarded due to blinks and other involuntary motions such as swallowing.
We took a different approach, which is that there is valuable information about the user's facial expression which can be derived from the pattern of muscle signals, and we developed specific classifier systems to allocate different muscle group activation patterns to specific expressions. We are then able to apply some filtering to our brain signals so we have a better chance of seeing the real brain signal through the muscle noise because we can tell what kinds of muscle signal are occurring.
We use a combination of filtering and specific brain pattern features which are less affected by muscle movements to derive the underlying brain behavior.