BCI-driven spelling devices
In 1999, the Tübingen BCI group developed the Thought-Translation-Device (TTD; Birbaumer et al., 1999), a system that could be operated by patients suffering from ALS through the modulation of brain rhythms. Binary decisions made by the BCI were used to select letters in a procedure where the alphabet was iteratively split into halves. The achieved spelling rate was about 0.5 char/min. Since then, other groups have developed BCI-driven spelling devices based on the detection of voluntarily patterns of activity in the spontaneous EEG. These systems can operate synchronously (Birbaumer et al., 1999; Obermaier et al., 2003) or asynchronously (Millán, 2003; Millán et al., 2004a; Scherer et al., 2004; Müller and Blankertz, 2006; Williamson et al., 2009). Interestingly, one patient suffering from severe cerebral palsy could operate the Graz system at about 1 char/min. In the case of Millán's approach, trained subjects have taken 22.0 s on average to select a letter, including recovery from errors, with peak performances of 7.0 s per letter. Particularly relevant is the spelling system developed by the Berlin group in cooperation with the University of Glasgow, called Hex-o-Spell (Williamson et al., 2009), which illustrates how a normal BCI can be significantly improved by state-of-the-art HCI principles. The idea for Hex-o-Spell was taken from the Hex system which was designed for use on mobile devices augmented with accelerometers, where tilt control was used to maneuver through a hexagonal tessellation. The text entry system is controlled by the two mental states imagined right hand movement and imagined right foot movement. Expert subjects achieved typing speed of up to 7.5 char/min. A recent development in the field of HCI, inspired by a similar approach to Hex, is the Nomon selection system, based on the use of phase angle in clock-like displays (Broderick and MacKay, 2009). Still another speller designed upon efficient HCI principles is DASHER (Wills and MacKay, 2006).
Most BCI spelling devices, especially those actually used by disabled people, are based on the detection of potentials that are evoked by external stimuli rather than spontaneous mental states. The most prominent is the approach that elicits a P300 component (Donchin et al., 2000). In this approach, all characters are presented in a matrix, and the symbol which the user focuses her/his attention on can be predicted from the brain potentials that are evoked by random flashing of rows and columns. Similar P300-based spelling devices have since been extensively investigated and developed (e.g., Sellers et al., 2006; Nijboer et al., 2008; Silvoni et al., 2009).