Activity recognition using incremental learning

This paper presents an unsupervised incremental learning approach for activity recognition. Activity recognition is important because ambient intelligent spaces need to recognize the activity of the inhabitant before it can provide the appropriate support or assistance. However, building a knowledgebase of appropriate support is difficult, tedious and expensive. It is not guaranteed to be complete, therefore, it is unable to handle novel situations. In this paper an unsupervised incremental algorithm was used on an 82-hour activity corpus of daily living was gathered by having a male inhabitant occupy the living space for three to four hours at a time. Accuracy is 93.04%