Abstract

The information transfer rate (ITR) is widely accepted as a performance metric for generic brain-computer interface (BCI) spellers, while it is noticeable that the communication speed given by ITR is actually an upper bound which however can never be reached in real systems. A new performance metric is therefore needed. In this paper, a new metric named average time consumption per character (ATCPC) is proposed. It quantifies how long it takes on average to type one character using a typical synchronous BCI speller. To analytically derive ATCPC, the real typing process is modelled with a random walk on a graph. Misclassification and backspace are carefully characterized. A close-form formula of ATCPC is obtained through computing the hitting time of the random walk. The new metric is validated through simulated typing experiments and compared with ITR. Firstly, the formula and simulation show a good consistency. Secondly, ITR always tends to overestimate the communication speed, while ATCPC is more realistic. The proposed ATCPC metric is valid. ATCPC is a qualified substitute for ITR. ATCPC also reveals the great potential of keyboard optimization to further enhance the performance of BCI spellers, which was hardly investigated before.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call