An integrated approach for a robust command and control application on the motorcycle

Automatic speech recognition for command and control in human-computer interaction (HCI) is used in many applications today. For noisy environments, like the car or motorcycle environment, word error rates of the speech recognition increase and, hence, the reliability of such systems decreases. To achieve an overall robust and reliable system, the implementation of a noise reduction algorithm or the adaptation of the acoustic models alone is usually not sufficient. In this paper the robust speech recognition system for command and control interaction for the European projectMoveOn is presented in detail. Several central aspects for achieving a reliable system are discussed. These comprise robust speech capturing, layout and adaptation of the system to the domain and purpose, training of appropriate acoustic models, a fast decoder concept, and additional preprocessing algorithms. We present and evaluate a robust base system and appropriate signal preprocessing for a command and control application on police motorcycles. The results show that a reliable command and control system for HCI on a motorcycle is generally feasible.