Efficient model training for HMM-based person identification by gait

In gait-based person identification, statistical methods such as hidden Markov models (HMMs) have been proved to be effective. Their performance often degrades, however, when the amount of training data for each walker is insufficient. In this paper, we propose walker adaptation and walker adaptive training, where the data from the other walkers are effectively utilized in the model training. In walker adaptation, maximum likelihood linear regression (MLLR) is used to transform the parameters of the walker-independent model to those of the target walker model. In walker adaptive training, we effectively exclude the inter-walker variability from the walker-independent model. In our evaluation, our methods improved the identification performance even when the amount of data was extremely small.