Our detections vary in accuracy. Each detection has its own 'accuracy' figure and indeed the accuracy expressed is an average across users, therefore for some people the detection is more accurate than the quoted figure and for others the accuracy is less.
Cognitiv can achieve very high accuracy for a single command and the accuracy becomes less when more actions are added. But in this case the accuracy is also strongly dependent on the user's skill level and ability to remember the mental state used in training.
Affectiv detections (i.e. Performance Metrics) vary significantly. Excitement is very accurate, Frustration and Engagement a little less and Meditation covers some types of meditation quite well and others less so (for example, meditation styles may have high or low cognitive involvement).
Overall our accuracy is determined by an index calculated within specified time segments. A True Positive (TP) is a correct activation of an event, True Negative (TN) is a period of time where no actions are attempted and none are detected, and False Negatives (FN) and False Positives (FP) are failures to trigger in the required period, or a trigger occurring incorrectly. Accuracy is determined by (TP+TN)/(TP+FP+TN+FN)
Measurement accuracy of the device is technically: minimum voltage resolution 0.51uV, 14 bit accuracy, 0.2-43Hz bandwidth.