ABSTRACT
In Chapter 8, we saw how simple activity recognition is being handled. In this chapter, we discuss complex activity recognition. Our approach will focus on how to represent complex activities in terms of a sequences of simpler activities. We present a general framework for constructing models for complex activity recognition. The models are
hierarchically structured and represent events at different granularities. We also show how to partition the observation space so that low-level, simple events are defi ned automatically according to the statistics of the training data.
