Proposal and implementation of a spoken word recognizer using utterance normalization and multiple templates on a single VLSI chip