Speech Input Pre-Processing for Car Driver Robust Automatic Speech Recognition

This paper describes a pre-processing system that allows robust Automatic Speech Recognition (ASR) in car adverse environment. The pre-processing with only 3 microphones allows that only the driver provides commands to the car, the passenger voice is cancelled or reduced depending the situation and music does not need to be muted. An adaptive beamformer is combined with sources localization based on Direction-Of-Arrival (DOA) estimation. Additional specific adaptive filters are combined to remove remaining noises (engine, road, wind, music, etc.). The evaluation is done using real driving situations on a Toyota RAV4 car and with the Dragon Naturally Speaking (DNS) software in French to assess the recognition rate. The results show a recognition rate drastically improved.

  • Availability:
  • Supplemental Notes:
    • Abstract used with permission of ITS Japan. Paper No. 2139.
  • Corporate Authors:

    ITS Japan

    Tokyo,   Japan 
  • Authors:
    • Vrazic, Sacha
    • Sugae, Ippei
    • Inaba, Hisashi
    • Murakami, Yuichi
  • Conference:
  • Publication Date: 2013

Language

  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; References; Tables;
  • Pagination: 9p
  • Monograph Title: 20th ITS World Congress, Tokyo 2013. Proceedings

Subject/Index Terms

Filing Info

  • Accession Number: 01535029
  • Record Type: Publication
  • ISBN: 9784990493981
  • Files: TRIS
  • Created Date: Aug 20 2014 4:42PM