Frequency domain forecasting approach for latency reduction in direct human-computer interaction

The problem of latency reduction in direct human-computer interaction is considered and formulated as a trajectory prediction problem. To solve the problem, the predictor is constructed as a frequency-domain approximation of the non-casual ideal predictor. This approximation can be computed analytically, or obtained as an optimization task. An adaptive modification of the forecasting algorithm is proposed taking into account possible variations in user behavior. Experimental results illustrate the applicability of the proposed solution.