3 Different ways of system control
3.4 System control via brain-computer interface (BCI)

Systems control using BCI is one of the one challenges for future systems control options.

BCI has more advantages as privacy (no loud sounds, no visible gestures) and low computation (only very few data are captured and processed when comparing e.g. to video).

But BCI has at the moment also major drawbacks as comfort issue (the headband, headset or headcap is needed) and mental effort (certain mental effort in most cases necessary to “generate” the control signals).

image
A conceptual scheme of a BCI

General scheme of the simple BCI system is depicted on Fig. 1. The first stage is a signal acquisition. EEG measures electric brain activity during synaptic excitations in the neurons. EEG signals are captured non-invasively using electrodes on the scalp. After the signal acquisition, signals are to be pre-processed. In general, the acquired brain signals are contaminated by noise and other artefacts caused by bio signals or external signals like power line, etc. After obtaining the “noise-free” signals in the signal enhancement phase, essential features from the brain signals are to be extracted. The most common feature extraction methods used with EEG signals include the discrete orthogonal transforms. Once the proper features are extracted, there must be a method that classifies the signal into desired classes. There are many categories of classification techniques as: generative (Gaussian Mixture Model -GMM), discriminative (Neural networks- NN, Support Vector machines- SVM), non-parametric i.e. sample based (K nearest neighbour - KNN), etc. Each method has its pros and cons and thus must be selected based on the application requirements. The Control interface stage of BCI uses the classification output as a control signal. The approaches can be divided into: endogenous (based on self-regulation of brain rhythms and potentials without external stimuli) and Exogenous (uses the neuron activity elicited in the brain by an external stimulus). The most frequently used methods include slow cortical potentials (SCP), sensorimotor rhythms, visual evoked potentials (VEP) including steady-state VEPs (SSVEPs), and P300. Particular devices or processes can be operated using the control interface. BCI using sensorimotor rhythms use Mu brain waves present imagination of motoric actions (e.g. hand movement). This is endogenous BCI. On the other hand, the SSVEP is exogenous BCI. SSVEP is based on the property that the visual cortext “resonated” according to the frequency of the visual stimulus that the subject observes. Thus, using this method there are needed e.g. on screen movement control buttons, each blinking with different frequency. As the user looks at the particular button, the signal captured on visual cortex contains this frequency, so the estimation at which button the user is looking can be done. Another frequently used method is the P300. It is based on the fact, that positive peak that appears in the EEG approximately 300 ms after the presentation of a rare stimulus. Though we have an external stimulus here, P300 is considered to be an endogenous BCI, as the occurrence peak links not to the physical attributes of a stimulus, but reflect processes involved in stimulus evaluation or categorization.