Abnormal Activity Recognition Based on HDP-HMM Models

Detecting abnormal activities from sensor readings is an important research problem in activity recog-nition. A number of different algorithms have been proposed in the past to tackle this problem. Many of the previous state-based approaches suffer from the problem of failing to decide the appropriate number of states, which are difficult to find through a trial-and-error approach, in real-world applica-tions. In this paper, we propose an accurate and flexible framework for abnormal activity recogni-tion from sensor readings that involves less hu-man tuning of model parameters. Our approach first applies a Hierarchical Dirichlet Process Hid-den Markov Model (HDP-HMM), which supports an infinite number of states, to automatically find an appropriate number of states. We incorporate a Fisher Kernel into the One-Class Support Vector Machine (OCSVM) model to filter out the activi-ties that are likely to be normal. Finally, we derive an abnormal activity model from the normal activ-ity models to reduce false positive rate in an unsu-pervised manner. Our main contribution is that our proposed HDP-HMM models can decide the ap-propriate number of states automatically, and that by incorporating a Fisher Kernel into the OCSVM model, we can combine the advantages from gener-ative model and discriminative model. We demon-strate the effectiveness of our approach by using several real-world datasets to test our algorithm’s performance.