Temporal pattern recognition using fuzzy clustering
Todd K. Moon · 1994
In many problems it is necessary to recognize patterns of time sequences of feature vectors where the training vectors and the test vectors are not temporally aligned. In this paper the author presents a fuzzy clustering approach to this temporal pattern recognition. Observation classification vectors are embedded into a larger vector that explicitly shows the time dependence. Clustering is done both in this larger space and in the observation space to give "state" and "output" spaces similar to those used in HMM modeling. Recognition is accomplished by finding the best match in state order to the clustering.>