Human Activity Recognition by Head Movement using Elman Network and Neuro-Markovian Hybrids

Traditionally, human activity recognition has been achieved mainly by the statistical pattern recognition techniques such as the Nearest Neighbor Rule (NNR), and the state-space methods, e.g. the Hidden Markov Model (HMM). This paper proposes three novel approaches – the use of the Elman Network (EN) and two hybrids of Neural Network (NN) and HMM, i.e. HMM-NN and NN-HMM, to recognize ten simple activities in an office environment. The sex, race and physique invariant feature vectors are extracted from tracking the subjects’ head movement over consecutive frames. Based on our database of 200 activity sequences, experimental results show that all the three proposed systems perform better than the two popular conventional methods. The HMM-NN system attained the best performance of 96.5%. The encouraging results not only reveal the performance improvement of combining NN and the traditional HMM, but also demonstrate our proposals’ greater potential in realizing recognition of continuous complex activities.