Adaptive Touch Panel User Interface by Type-Based Approach Using Particle Filters

A new method for adaptation of touch panel user interface is proposed. The method employs a type-based approach effectively using prior information on user population which is represented by mixture model. An advantage of the approach is that we only need to estimate weights of types which is numerically far less expensive than estimating values of all attributes of the user. State space model is formalized to estimate the weights of types by supposing smoothness prior to time evolution of the weights and mixture model with the time-varying weights. Particle filter is used to estimate the weights based on observation series of user operations up to current time. An experiment on touch panel user interface demonstrates the efficiency of the proposed method.