Intention estimation using time-varying fuzzy Markov models

We propose intention estimation using time-varying fuzzy Markov models. Based on human non-verbal information, such as gestures or posture change, we vary the probability between states of the model to improve the accuracy of estimation. The time-varying fuzzy Markov model therefore composes of two part. First, we define the initial probability of the fuzzy Markov model according to human experience. We then adjust the probability according to the actual time-varying life environment estimate the human intention. The advantages of the approach are: non-verbal information is core of human intention; time-varying probability improves estimation accuracy; and fuzzy inference consider practical human experience. The comparison of simulations for both fixed fuzzy Markov model and time-varying fuzzy Markov model reveals the latter is more accurate in estimating human intention.