Search
Search Funnelback University
- Refined by:
- Date: 2007
1 -
35 of
35
search results for TALK:PC53 20 |u:mi.eng.cam.ac.uk
where 0
match all words and 35
match some words.
Results that match 1 of 2 words
-
Discriminative Models for Speech Recognition Mark Gales 1 February ...
mi.eng.cam.ac.uk/~mjfg/talk_ita.pdf22 Feb 2007: LanguageTraining. ML MPE. English (WER%) 16.0 13.1Arabic (WER%) 22.9 20.0. Mandarin (CER%) 14.4 12.7. ... 3-emitting state HMMs. ClassifierTraining Componentsλ α 10 20. HMM ML – 29.4 27.3C-Aug ML CML 24.2 –. HMM MMI – 25.3 24.8C-Aug MMI CML 23.4 -
paper.dvi
mi.eng.cam.ac.uk/~mjfg/liao_ICASSP07.pdf15 Aug 2007: Compensation Noise Est. Type %WER. None — 20.8. JointML VTS 19.1ML Joint 18.8. ... The best training scheme was. the JAT system, which exceeded matched and multistyle with VTSperformance at both 20 and 14 dB SNR. -
UNSUPERVISED TRAINING FOR MANDARIN BROADCAST NEWS AND…
mi.eng.cam.ac.uk/~mjfg/wang_ICASSP07.pdf22 Jun 2007: Two thresholds were used in the show level selection. The firstwas a show-level confidence score of 55%, the second a thresholdof 20% for the percentage of English. ... The gains from using discriminative training weremuch reduced. MPE unsupervised -
SPEECH RECOGNITION SYSTEM COMBINATION FOR MACHINE TRANSLATION M.J.F.…
mi.eng.cam.ac.uk/~mjfg/gales_ICASSP07.pdf22 Jun 2007: In addition a maximum segmentlength of 20 seconds was imposed by splitting longer segments atthe longest inter-word gap. ... System Comb. CER%level bnmd06 bcmdr06CU — 8.0 21.8BBN 8.6 23.9BBNCU — 7.5 20.7CUBBN char 7.1 21.3. -
PREDICTIVE LINEAR TRANSFORMS FOR NOISE ROBUST SPEECH RECOGNITION…
mi.eng.cam.ac.uk/~mjfg/gales_ASRU07.pdf15 Aug 2007: Here. Σ(m)-1 = A(r)TΣ̃(m)-1diag A. (r) (20). and Σ̃(m)diag is a diagonal covariance matrix. ... System # iter SNR(dB)20 15 10 50 31.2 50.4 79.2 89.7. -
Discriminative Models for Speech RecognitionM.J.F. Gales Cambridge…
mi.eng.cam.ac.uk/~mjfg/mjfg_ita.pdf22 Feb 2007: m=1. γjm(t) (20). For the second derivative,. ckn Tcjm log(p(O1:T ; λ)) = (21)1. ... model. No data or feature whitening was performed. Classifier Criterion Componentsλ α 10 20. -
UNSUPERVISED TRAINING FOR MANDARIN BROADCAST NEWS AND…
mi.eng.cam.ac.uk/research/projects/AGILE/publications/wang_ICASSP07.pdf10 Oct 2007: Two thresholds were used in the show level selection. The firstwas a show-level confidence score of 55%, the second a thresholdof 20% for the percentage of English. ... The gains from using discriminative training weremuch reduced. MPE unsupervised -
Unsupervised Language Model Adaptation for Mandarin…
mi.eng.cam.ac.uk/reports/svr-ftp/mrva_icslp06.pdf20 Jan 2007: BN data and 20 hours of BC data. The language model is describedin section 3.2. -
SPEECH RECOGNITION SYSTEM COMBINATION FOR MACHINE TRANSLATION M.J.F.…
mi.eng.cam.ac.uk/research/projects/AGILE/publications/gales_ICASSP07.pdf10 Oct 2007: In addition a maximum segmentlength of 20 seconds was imposed by splitting longer segments atthe longest inter-word gap. ... System Comb. CER%level bnmd06 bcmdr06CU — 8.0 21.8BBN 8.6 23.9BBNCU — 7.5 20.7CUBBN char 7.1 21.3. -
Discriminative Models for Speech RecognitionM.J.F. Gales Cambridge…
mi.eng.cam.ac.uk/~mjfg/mjfg_ita_margin.pdf30 May 2007: Tl(w) = log (P (w)) (20). and constrain all the α’s associated with this for all models tobe the same, αl. ... model. No data or feature whitening was performed. Classifier Criterion Componentsλ α 10 20. -
IEEE Proo f IEEE TRANSACTIONS ON AUDIO, SPEECH, AND ...
mi.eng.cam.ac.uk/~mjfg/kai_ASP06.pdf4 Apr 2007: If eigenvoices are updated using the maximumlikelihood eigenspace (MLES) approach [20], it is equivalentto updating CAT cluster mean matrices and leaving covariancematrices unchanged [5]. ... Int. Conf. Speech Language Processing, 1998, pp.1771–1774. -
JOURNAL OF IEEE TRANS. ACOUST., SPEECH, SIGNAL PROCESSING, JULY ...
mi.eng.cam.ac.uk/research/projects/AGILE/publications/sim_SAP06.pdf10 Oct 2007: µm =xnm xdm Dµ̂mβnm βdm Dm. (20). Wmpem =Ynm Ydm Dm(Σ̂m µ̂mµ̂′m). ... Combining equations (20)and (21) gives the full covariance statistics in terms ofDm,. -
HYBRID SYSTEMS FORRECONSTRUCTION OF FREEHAND 3D ULTRASOUND DATA R. ...
mi.eng.cam.ac.uk/reports/svr-ftp/housden_tr574.pdf30 Mar 2007: 0 20 40 60 80 100 120 1400.5. 0. 0.5. 1. ... es). 0 20 40 60 80 100 120 1400.5. 0. 0.5. -
ROTATIONAL MOTION INSENSORLESS FREEHAND 3D ULTRASOUND R. J. Housden,…
mi.eng.cam.ac.uk/reports/svr-ftp/housden_tr587.pdf5 Oct 2007: 0.4. peak. offs. et (. mm. ). depth (mm)0 10 20 30 40. ... IEEETransactions on Ultrasonics, Ferroelectrics, and Frequency Control, 33(3):257–264, May 1986. [20] G. -
Non-rigid Photometric Stereo with Colored Lights Carlos Herńandez1…
mi.eng.cam.ac.uk/research/projects/VideoNormals/NonRigidPhotometricStereoWithColoredLights_iccv2007.pdf16 Aug 2007: Wecanadvectdt(u, v) to estimate(ut, vt) using the followingequation from [20]. (uj , vj ) = (uj1, vj1)dj1(. ... Thirdrow: texture-mapping is obtained by advecting frame-by-frame 2D optical flow [20]. -
Recent Progress in Large Vocabulary ContinuousSpeech Recognition: An…
mi.eng.cam.ac.uk/~mjfg/icassp06_tutorial.pdf22 Feb 2007: Typically for HTK (natural logs)– α in range 10 to 16– β in range 0 to 20. ... Cambridge UniversityEngineering Department. ICASSP 2006 Tutorial 20. Recent Progress in LVCSR: An HTK Perspective cMark Gales & Phil Woodland, 2006. -
Freehand ultrasound elastography with a 3D probe G.M. Treece, ...
mi.eng.cam.ac.uk/reports/svr-ftp/treece_tr577.pdf19 Jun 2007: We achieve this in a system which matches the fastestreported processing time of about 20 seconds (Anand et al., 2007), even though this technique onlyemployed 1D windows. ... 3D strain images were calculated in a postprocessing step, lasting typically -
journal.dvi
mi.eng.cam.ac.uk/~mjfg/kcs_csl.pdf4 Apr 2007: This model will be referenced to aspMPE [20]. Similar to fMPE, the time dependent scale factor is a weightedcontribution from the centroid specific scales where the weights are given bythe posteriors ... 20. The fMPEMPE system improved the baseline by -
IEEE TRANS. ON SAP, VOL. ?, NO. ??, ????? ...
mi.eng.cam.ac.uk/~mjfg/kai_ASP07.pdf23 Apr 2007: Thus. q(θ, T ) = q(θ|O, H)q(T |O, H) (20). This assumption is necessary to obtain a tractable mathemati-cal form. ... The optimisation of the boundis then only based on possible hypotheses for the uth block.The variational distributions in equation (20) -
CAMBRIDGE UNIVERSITY ENGINEERING DEPARTMENT TRANSFORMATION STREAMS…
mi.eng.cam.ac.uk/~mjfg/mjfg_csl.pdf3 May 2007: Bias transforms have been used for adaptation in speechrecognition, but have been shown to be less effective than more complex transforms [20]. ... PhD thesis, ICSI,UC Berkely, 1999. [20] L R Neumeyer, A Sankar, and V V Digalakis. -
K. Yu, M.J.F. Gales and P.C. Woodland Cambridge University ...
mi.eng.cam.ac.uk/research/projects/AGILE/publications/yu-interspeech07.pdf10 Oct 2007: data. 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10. 10. 20. 30. -
tech.dvi
mi.eng.cam.ac.uk/~mjfg/liao_tr552.pdf21 Sep 2007: a) Noise mean n=0. 10 0 10 20 300. 0.05. 0.1. ... 10 0 10 20 300. 0.05. 0.1. 0.15. p(y). Corrupted Speech, y. -
IEEE TRANS. ON SAP, VOL. ?, NO. ??, ????? ...
mi.eng.cam.ac.uk/research/projects/AGILE/publications/liu_ASL07.pdf10 Oct 2007: development data, dev01sub, was used. The test set contains 20 Switchboard I and 20 Switchboard. ... 20 52 - - 33.6†. BIC (ρ = 0.5) 36.6 34.9 33.4. -
Real-time Freehand 3D Ultrasound Calibration P-W. Hsu, R. W. ...
mi.eng.cam.ac.uk/reports/svr-ftp/hsu_tr565.pdf19 Feb 2007: Ultrasound in Medicine & Biology, 20(9):923–936, 1994. A. Fenster, D. B. -
AN INTELLIGENT INTERFACEFOR FREEHAND STRAIN IMAGING J. E. Lindop, ...
mi.eng.cam.ac.uk/reports/svr-ftp/lindop_tr578.pdf25 May 2007: This issue also affects exhaustive searching basedon correlation coefficient or SAD, where in both cases the imposition of continuity constraints –whether explicitly [20] or implicitly [14] — substantially reduces the rate of -
PhD Thesis
mi.eng.cam.ac.uk/~mjfg/thesis_ky219.pdf16 Nov 2007: 2.2 A composite HMM constructed from two individual HMMs 13. 2.3 State clustering for single Gaussian tri-phones 20. -
IEEE TRANS. ON SAP, VOL. ?, NO. ??, ????? ...
mi.eng.cam.ac.uk/research/projects/AGILE/publications/kai_ASP07.pdf10 Oct 2007: Thus. q(θ, T ) = q(θ|O, H)q(T |O, H) (20). This assumption is necessary to obtain a tractable mathemati-cal form. ... The optimisation of the boundis then only based on possible hypotheses for the uth block.The variational distributions in equation (20) -
��������� �� � �� �������������� ��������������� �������� ���� �����…
mi.eng.cam.ac.uk/~mjfg/pmc_SAP.pdf15 Aug 2007: Diagonal. Extended. Standard Diagonal. 5 10 15 20 25 30 3510. -
Discriminative Complexity Control and Linear Projections for Large…
mi.eng.cam.ac.uk/~mjfg/thesis_xl207.pdf16 Nov 2007: 2.3.5 Decoding Algorithms 20. 2.4 Linear Projection Schemes 22. 2.4.1 Linear Discriminant Analysis 22. ... oτ µ(jm))>. τ γjm(τ )(2.20). -
A Gaussian Mixture Model Spectral Representation for Speech…
mi.eng.cam.ac.uk/~mjfg/thesis_mns25.pdf16 Nov 2007: 2.2.6 Scoring and confidence 20. 2.3 Noise robustness 20. 2.3.1 Noise robust features 21. ... the model, i.e. maximise the function I :. I 3 / (2.20)Unfortunately, there exists no closed form solution for the optimisation of the function above. -
� ������� � ��� � ������� ����������������������� !�"� ��� ...
mi.eng.cam.ac.uk/~mjfg/thesis_nds1002.pdf16 Nov 2007: ". #$! % &! '$#%. ( )& #-,.-. /1032546087:9<;>=@?AB08CEDGF@HJIK2>4. LNMPORQTSVUWXSLNY[ZTZT][. _PUabcdOfehgig[j. kmlonpnKqsrptpuvtplowyxznp{-|}ltKtKqwVrtp-q. kdqsVrpqsq@wVikdwtKwVrwV[_lowynKwVa. "!#$% '&)( ,-. /. -
����������� ��� ������������� ������� ����…
mi.eng.cam.ac.uk/~mjfg/liao_SPEECOM.pdf8 Oct 2007: 180 190 200 210 220 230. -10. 0. 10. 20 Corrupted SpeechClean SpeechFE-Joint EstimateFE-Joint /– Std. ... 10. 0. 10. 20 Corrupted SpeechClean SpeechFE-Joint EstimateFE-Joint /– Std. Dev. -
PhD Thesis
mi.eng.cam.ac.uk/~mjfg/thesis_kcs23.pdf16 Nov 2007: 2.2.3 Speaker Adaptive Training and Adaptation 20. 2.2.4 Decoding Strategies 22. ... 2000. 3000. 4000. 5000. 0. 20. 40. 60. 80. 100. Figure 1.3 Short Time Fourier Transform on a speech waveform to obtain a spectogram. -
ATK An Application Toolkit for HTK Version 1.6 Steve ...
mi.eng.cam.ac.uk/research/dialogue/ATK_Manual.pdf20 May 2007: 20. process will have to perform. For example, a component process may have several input buffers and it. ... AMONITOR DISPYORIGIN 20 top-left y origin of display. AMONITOR DISPWIDTH 460 width of a display panel. -
Linear Gaussian Models for Speech Recognition Antti-Veikko Ilmari…
mi.eng.cam.ac.uk/~mjfg/thesis_avir2.pdf16 Nov 2007: 2.5.2 Precision Matrix Modelling 20. 2.6 Segment Models 21. 2.6.1 Stochastic Segment Model 21. ... to reduce boundary effects [20] in the subsequent processing. The fast Fourier transform is.
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.