Recognizing Human Behaviour from Temporal Sequential Data with Activity Assignment

A probabilistic system in an ambient assisted living environment is automatically built to detect human behaviour. The focus lies on the early prediction of human activities based on domotic sensor data and on general activity assignment. First recurrent patterns are detected using the Temporal-Pattern (T-Pattern) algorithm and further a probabilistic finite-state automaton is generated out of the patterns. Afterwards the patterns are assigned to specific defined human activities with the help of Fuzzy Logic. The needed rules are learned automatic from an annotated dataset.