(Springer-Verlag, 2003-03) Pulasinghe, K; Watanabe, K; Kiguchi, K; Izumi, K
In this article, a fuzzy neural network (FNN)-
based approach is presented to interpret imprecise natural
language (NL) commands for controlling a machine. This
system, (1) interprets fuzzy linguistic information in NL
commands for machines, (2) introduces a methodology to
implement the contextual meaning of NL commands, and
(3) recognizes machine-sensitive words from the running
utterances which consist of both in-vocabulary and out-ofvocabulary words. The system achieves these capabilities
through a FNN, which is used to interpret fuzzy linguistic
information, a hidden Markov model-based key-word spotting system, which is used to identify machine-sensitive
words among unrestricted user utterances, and a possible
framework to insert the contextual meaning of words into
the knowledge base employed in the fuzzy reasoning process. The system is a complete system integration which
converts imprecise NL command inputs into their corresponding output signals in order to control a machine. The
performance of the system specifications is examined by
navigating a mobile robot in real time by unconditional
speech utterances.
