Interactive Korean speech recognition for a mobile robot

Abstract

This paper is a study on the implementation of a speech recognition mobile robot to which the interactive Korean speech recognition techniques are applied. The speech command uttered the sentential connected word and is asserted through the wireless mic system. This speech signal transferred LPC-Cepstrum and short-time energy which are computed from the received signal on the DSP board to a notebook PC. In the notebook PC, the DP matching technique is used for the recognizer and the recognition results are transferred to the motor control unit which outputs pulse signals corresponding to the recognized command and drives the stepping motor. It has been implicitly assumed that, for connected word recognition, each word in the string can be followed by any other word in the string. However, for some connected word recognition tasks, there is an explicit set of rules governing which words can logically follow other words to form valid sentences in the language. In this paper, a grammar network was applied to reduce the recognition speed of the recognizer, so that real time recognition is realized. The misrecognized command is revised by interface revision through conversation with the mobile robot. The recognition system asks the user for help whenever there are two or more recognition candidates whose likelihood scores are all high and it is difficult to resolve small differences in the string. The recognizer asks the user to verify the first choice decision; if it is not verified, the recognizer asks the user to verify the second choice, etc. Alternatively, the recognition system can list the confusing candidates and ask the user to choose the correct string based on the string index number in the list.