Lip reading using fuzzy logic network with memory

This paper proposes a new approach for lip reading. Most existing systems for lip reading utilize a kind of neural network or hidden Markov models as classifiers. We propose a new approach where fuzzy combinational networks with fuzzy flip-flop memories are combined into one network. Our model introduces a hierarchical structure of this network, where single layers are contextually dependent. Experiments with fuzzy flip-flop network propose a new approach in the process of automatic lip-reading system where time dependence in inputs series is modeled with memories. Such approach provides possibilities for continuous speech recognition.