id: 05253001
dt: a
an: 05253001
au: Schneider, Daniel; Winkler, Thomas; Löffler, Jobst; Schon, Jochen
ti: Robust audio indexing and keyword retrieval optimized for the rescue
    operation domain.
so: Löffler, Jobst (ed.) et al., Mobile response. First international workshop
    on mobile information technology for emergency response, Mobile
    Response 2007, Sankt Augustin, Germany, February 22‒23, 2007. Revised
    selected papers. Berlin: Springer (ISBN 978-3-540-75667-5/pbk). Lecture
    Notes in Computer Science 4458, 135-142 (2008).
py: 2008
pu: Berlin: Springer
la: EN
cc: 
ut: 
ci: 
li: doi:10.1007/978-3-540-75668-2_15
ab: Summary: Due to the extreme environmental conditions, speech recognition in
    the rescue operation domain is a complex and difficult task. Various
    types of noise and speaker stress represent the main problems for the
    recognition engine. Within the SHARE project, a dedicated domain
    specific training corpus was recorded to improve the robustness of the
    audio indexing service. An experimental evaluation showed that a speech
    recognition model trained with a small amount of domain specific data
    outperforms models based on a large set of already available data from
    other domains. Using the domain specific models, the number of false
    alarms produced in a noisy testing environment could be reduced by
    80\%.
rv: