Specific events, e.g. finger flexions, generate synchronized neuronal responses in the brain that can be measured on the scalp with EEG. However, a voltage response to a single event typically generates a low-amplitude response relative to the background EEG. One way to enhance the signal-to-noise ratio is to repeatedly record the EEG in response to the same event. By time-locking the event with the EEG and averaging the individual event-related EEGs, the random background noise will be reduced (stochastic noise) and the event-related potential (or ERP) can be extracted from the EEG (Woodman, 2010). An example of an ERP to repeated index finger flexions is shown in Fig. 1. Such a wave doesn't need 10,000 repeats. 25 may be sufficient.
If you characterize this ERP and determine specific features of it (duration, amplitude, latency after event etc.) you can generate algorithms that detect these features in the raw EEG, without the need to do repeated measures. In fact, thinking of a motor movement (e.g. grasp that coffee mug) generates very similar EEG activity as the factual act of grasping that mug. By recording the ERPs using events like think of grasping that coffee mug from a person missing a hand, and subsequently designing the necessary algorithms to detect that waveform near-realtime in the person's EEG, one can design brain-computer interfaces to control prosthetic limbs (Guger & Pfurtscheller, 2016).
Fig. 1. ERP to voluntary index finger flexions. Motor activity was recorded from the motor area. Source: Ball et al. (1999)
- Ball et al. NeuroImage (1999); 10: 682–94
- Guger & Pfurtscheller, AAATE 5th European Conference Advancement of Assistive Technology (2016)
- Woodman, Atten Percept Psychophys (2010); 72(8): 10.3758