Unsupervised Human Action Categorization Using Latent Dirichlet Markov Clustering

We present a novel unsupervised learning method for human action categories from video sequences using Latent Dirichlet Markov Clustering (LDMC). Video sequences are represented by a novel "bag-of-words" representation, where each frame corresponds to a "word". The algorithm automatically learns the probability distributions of the words and the intermediate topics corresponding to human action categories, and correlates actions over time. This is achieved by using Latent Dirichlet Markov Clustering (LDMC). Our approach builds on Hidden Markov Models (HMMs) and Latent Dirichlet Allocation (LDA), and overcomes their drawbacks on accuracy, robustness and computational efficiency. A collapsed Gibbs sampler is derived for offline learning with unlabeled training data, and significantly, a new approximation to online Bayesian inference is formulated to enable human action classification in new video data online in real-time. The strength of this mothod is demonstrated by unsupervised learning of human action categories and detecting irregular actions in different datasets.