BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Robust speech recognition / Towards better probabilistic models of
  speech - a discussion - Speaker to be confirmed
DTSTART:20130405T110000Z
DTEND:20130405T123000Z
UID:TALK43599@talks.cam.ac.uk
CONTACT:Catherine Breslin
DESCRIPTION:BRIDGING THE GAP BETWEEN SPEECH ENHANCEMENT AND RECOGNITION\, 
 Takuya Yoshioka\n\nOne major goal of ASR is to accurately transcribe targe
 t utterances even in acoustically adverse environments and this has been t
 ackled with a range of approaches. The approaches to noise robust ASR syst
 ems can be broadly classified into two categories. One is based on speech 
 enhancement\, which attempts to clean speech signals that are corrupted by
  noise. The speech enhancement approaches are usually optimized using SNR 
 or related critetia. The approaches in the other category\, including feat
 ure/model-space VTS\, JUD\, and SPLICE\, are more tailored to ASR in the s
 ense that they exploit statistical models of clean feature vectors for eff
 ective compensation for noise. In this short talk\, I will argue that thes
 e two types of approaches have\ndifferent advantages and present a novel g
 eneral scheme for integrating them. I will also present applications of th
 is scheme to meeting transcription and reverberant speech recognition. I t
 ry to explain the basic concept without using too many equations so that t
 he talk would be enjoyable for everyone in this group.\n\nTOWARDS BETTER P
 ROBABILISTIC MODELS OF SPEECH: WHY SAMPLED TRAJECTORIES SOUND BAD AND HOW 
 TO FIX THEM - A DISCUSSION\,\nMatt Shannon\n\nMatt will present some preli
 minary experimental results and some thought provoking speech samples for 
 discussion.
LOCATION:Department of Engineering - LR6
END:VEVENT
END:VCALENDAR