Abstract: In order to improve the recognition correct rate for audiovisual bimodal speech recognition in noise environment, this paper designs and implements a speech recognition system. This system simulates car environment and puts the speaking video information into speech recognition system. This system consists of three parts, model training, offline recognition and online recognition. Online recognition using voice as humancomputer interaction, has the function of adaptive user. Offline recognition has an analysis of data at different levels for system to produced, it is very suitable for bimodal speech recognition.

Key words: bimodal speech recognition, speech recognition, vehicular control