Search

19 Apr 2001: However, themain drawbacks are (1) new views cannot be added easilyat a later time, and (2) part of the structure will alwaysremain invisible under circular motion. ... The 3D models thus obtained canbe refined incrementally by adding new arbitrary views

24 Jul 2003: After eval03 found SPron lattices from scratch (new word lattices/modelmarked lattice MPE regen) helps by only another 0.1% absolute. ... variable number of Gaussians per state: 0.3% abs. • new MPE lattice generation/regeneration procedure: 1.2% abs.

27 Sep 2000: Subsequently we present thespeaker adaptation tests using the new front-end, followed by de-tails about language models used. ... Young S.J. (1998) The 1997HTK Broadcast News Transcription System.Proc. DARPABroadcast News Transcription and Understanding

25 Oct 2000: A new consistencyfunction is proposedthat usesa statisticalcomparisoninsteadof the voxelcentroidsamplingthat wasinitially proposed.This is importantwhenthereis moredetailin theimagesthancanbestoredin a voxel representation.Thenew functionis

12 May 2003: Sequential Monte CarloMethods in Practice. Statistics for Engineering and Information Science Series.Springer-Verlag, New York, 2001. ... J. Julier, J. K. Uhlmann, and H. F. Durrant-Whyte. A new approach for filteringnonlinear systems.

21 Sep 2006: 3. Feature extraction is done within a. t1. on the new weights (i.e. ... 5 Conclusion. A new method is proposed to recognise continuouselementary gestures in this paper and it can be applied tosign language recognition.

21 Sep 2006: This meanswe can start from an initially small model and test the ‘rele-vance’ of each new input vectori sequentially. ... 5 Conclusion. In this paper, a new appearance-based method is pro-posed to track an object which may undergo appearancechange.

Main/Speech/SeminarVideos > WebLeftBarExample. Toolbox. No permission to viewSpeech Group. Members. Research. Publications. Seminars. Joining the Speech Group. Software. Machine Intelligence Homepage. 2009 University of Cambridge Department of

9 Aug 2005: e-mail:th223,pcw,ge204,dp10006 @eng.cam.ac.uk. ABSTRACT. This paper discusses new features integrated into the CambridgeUniversity HTK (CU-HTK) system for the transcription of conver-sational telephone ... This is followed by adescription of the data

8 Mar 2000: Unfortunately thePorter algorithm suffers from some problems for example mapping“news” to “new” but not equating “government” and “governmen-tal”. ... We therefore used a new set of HTKtriphone HMMs in an HTK-2 recognition setup.

21 Apr 2006: To demonstrate this formally, we introduce in Figure 1(b) a new setof axes labelled xp and yp which are simply the x and y axes rotated anticlockwise through an3.

2 Nov 2000: 2 Finding the TopicsA mixture of broadcast news and newswire text was usedas training data for the topic model, with 144 million wordsof Broadcast News text and 25 million words of ... Whittaker and S.J. Young, “The 1997 HTKBroadcast News Transcription

19 May 2005: Experimental results are pre-sented on both broadcast news (BN) and conversational telephonespeech (CTS) transcription tasks. ... Development ofthe CU-HTK 2004 Broadcast News Transcription System,Summited to ICASSP’05. [10] K.

10 Jan 2005: This baseline was trained onCTS and broadcast news transcripts (500M words al-together). ... 3.2. RESULTS ON FISHER CORPUS. Fisher is a new database of CTS data released by LDCin 2003/04.

27 May 2004: INTRODUCTION. The majority of state-of-the-art speech recognition systems aretrained on found data, for example broadcast news and telephoneconversations. ... This is consistent with the gains that were obtainedon the Broadcast News task [11].

10 Apr 2000: These methods were shown tobe an important part of our overall recognition systemfor broadcast news. ... The 1996 Hub-4 Broadcast News Transcription develop-ment data was used for all the experiments reported inthis paper.

5 May 2000: It is also interesting tonote that the improvement is consistent over the varioustypes of data found in broadcast news. ... Johnson, T.R. Niesler,A. Tuerk, E.W.D. Whittaker, and S.J. Young. The1997 HTK Broadcast News Transcription System.

27 Jul 2020: Contact:Title LEARNING NEW ARTICULATOR TRAJECTORIES FOR A SPEECH PRODUCTION MODEL USING ARTIFICIAL NEURAL NETWORKS.

27 Jul 2020: Contact:Title A NEW FREQUENCY SHIFT FUNCTION FOR REDUCING INTER-SPEAKER VARIANCE. Abstract: tuerk_eurospeech93b.

27 May 2004: The large rescoring lattices produced in the first stage are thenrescored with the adapted triphones and new lattices are gener-ated on output. ... Niesler, A.Tuerk, E.W.D. Whittaker and S.J. Young, “The 1997 HTKBroadcast News Transcription System”