Multi-sensorial inputs for opto-acoustic identification with synergetic computers

With increasing computing power available, data stemming from different types of input sources can be used simultaneously to solve identification problems. In this paper we present an approach to combining optical lip reading and acoustic voice analysis for identifying either spoken numbers or speakers. Due to independence the different data sources, either lower error rates or a higher reliability of the results in comparison with simple optical lip reading can be expected. The classification of the preprocessed data is done by Synergetic Computers which have recently attracted increasing attention as robust algorithms for solving industrial classification tasks. Special potential of Synergetic Computers lies in their close mathematical similarity to self-organized phenomena in nature. Therefore they present a clear perspective for hardware realizations.