Measuring Performance of Virtual Keyboards Based on Cyclic Scanning

This paper presents an exhaustive study into the different topologies of virtual ambiguous keyboards that operate by scanning techniques, analyzing the text entry average time (t_c) and the average number of user inputs (UI_c) per character. An mathematical model shows that in comparison with unambiguous one, text entry, in multi-tap mode, doesn´t offers better performance, because both t_c and UI_c are greater in them. Another method of text entry, called Tnk (Text in n keys), offers improvement with respect to unambiguous keyboards. But solely highly ambiguous key-board (4-keys keyboards) shows a jointly reduction in t_c and UI_c. Results obtained with the model do to focus on highly ambiguous keyboard. This paper demonstrate, using simulation with extensive text, that character prediction with TnK mode only have better performance than unambiguous keyboard with character prediction in UI_c parameter. Another techniques of text entry are also studied.