Search
Search Funnelback University
- Refined by:
- Date: 2006
1 -
50 of
62
search results for TALK:PC53 20 |u:mi.eng.cam.ac.uk
where 0
match all words and 62
match some words.
Results that match 1 of 2 words
-
Complementary System Combination andGeneration for ASR Mark Gales 20…
mi.eng.cam.ac.uk/~mjfg/mjfg_tcstar.pdf5 Jul 2006: factorial HMMs [19], loosely coupled models [20], system combination [16]– synchronous: likelihoods combined at the state level. ... 0. 5. 10. 15. 20. 25. 30. 35. 40. 0 1000 2000 3000 4000 5000 6000 7000 8000 9000. % -
On Person Authentication by Fusing Visual and Thermal Face ...
mi.eng.cam.ac.uk/reports/svr-ftp/arandjelovic_AVSS06.pdf1 Sep 2006: a) Glasses ON. 0 20 40 60 80 100 1200.85. 0.9. ... Shashua. Learning over sets using kernel principalangles.JMLR, 4(10), 2003. [20] L. -
University of CambridgeEngineering Part IB Information Engineering…
mi.eng.cam.ac.uk/~cipolla/lectures/PartIB/old/2006-IB-search.pdf19 May 2006: Store additional branches ///. /// update bound and nearestNeighbour ///. }. return nearestNeighbour;. }. 20 Engineering Part IB: Paper 8 Image Search. -
Machine Learning for Speech & LanguageProcessing Mark Gales 28 ...
mi.eng.cam.ac.uk/~mjfg/FCSW_talk.pdf19 Jul 2006: 20 million words, 720 million frames of data);– rapid transcriptions/closed caption data. • ... 5. 10. 108. 64. 20. 24. 68. 10. 0. 0.05. 0.1. -
Face Set Classification using Maximally Probable Mutual Modes Ognjen…
mi.eng.cam.ac.uk/reports/svr-ftp/arandjelovic_ICPR06.pdf29 Apr 2006: λ(1)Dj = λ(2)Dj. (19). Then, writing. |Ci| =D. j=1. λ(i)j , (20). ... 12] K. K. Sung and T. Poggio. Example-based learning for view-basedhuman face detection.PAMI, 20(1), 1998. -
Augmented Statistical Models for SpeechRecognition Mark Gales &…
mi.eng.cam.ac.uk/~mjfg/Edin_talk.pdf5 Jul 2006: 25. 20. 15. 10. 5. 0. 5 sh ow dh ax g r ih dd l iy z. ... Cambridge UniversityEngineering Department. Trajectory Models For Speech Processing Workshop 20. Augmented Statistical Models for Speech Recognition. -
C:/SFWDoc/Academic/Publications/2006/ICPR_2006/Final_ContGest/icpr_200…
mi.eng.cam.ac.uk/reports/svr-ftp/sfwong_icpr06a.pdf21 Sep 2006: 20.2 fps).Figure 3 illustrates the recognition process on a typical test-ing clip. ... Pentland. Real-time ameri-can sign language recognition using desk and wearable com-puter based video.PAMI, 20(12):1371–1375, 1998. -
paper.dvi
mi.eng.cam.ac.uk/~mjfg/liao_INTER06.pdf22 Nov 2006: Thefront-end uncertainty schemes used diagonal transformations. SNR(dB)System 20 15 10 5. ... Number of SNR(dB)System Transforms 20 15 10 5. Diagonal Transformations1 3.33 5.92 13.35 31.96. -
A New Look at Filtering Techniques for Illumination Invariance ...
mi.eng.cam.ac.uk/reports/svr-ftp/arandjelovic_AFG06.pdf30 Jan 2006: 10. 15. 20. 25. 30. 35. 40. 45. Method. Err. or r. ... 5. 10. 15. 20. 25. Method. Err. or r. ate,. std. (%. ). -
Modelling Dependencies in SequenceClassification: Augmented…
mi.eng.cam.ac.uk/~mjfg/gales_UEA06.pdf22 Nov 2006: Frames from phrase:SHOW THE GRIDLEY’S. Legend. • True• HMM• SLDS. 20 40 60 80 10035. ... 0.5. 1. 2. 5. 10. 20. 40. False Alarm probability (in %)M. -
Learning Discriminative Canonical Correlationsfor Object Recognition…
mi.eng.cam.ac.uk/reports/svr-ftp/kim_eccv06.pdf21 Sep 2006: Learning Discriminative Canonical Correlationsfor Object Recognition with Image Sets. Tae-Kyun Kim1, Josef Kittler2, and Roberto Cipolla1. 1 Department of Engineering, University of CambridgeCambridge, CB2 1PZ, UK. {tkk22,cipolla}@eng.cam.ac.uk2 -
THE CU-HTK MANDARIN BROADCAST NEWS TRANSCRIPTION SYSTEM R. Sinha, ...
mi.eng.cam.ac.uk/~mjfg/sinha_ICASSP06.pdf22 Nov 2006: S1HLDA 12.4 6.6 21.1GAUSS 11.9 6.2 20.3. Table 2. Unadapted performance of baseline systems using HLDAor GAUSS front-ends, V1 segmentation and lm1.0. ... System LMCER (%). dev04f eval03m eval04. S2 GAUSS. lm1.0 12.2 5.6 20.0lm2.0 12.3 5.2 20.2lm2.1 11.5 5 -
johnson06stable.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/johnson_stable06.pdf18 Sep 2006: 20].Several comparison studies of the various interest points and descriptors havebeen carried out by Mikolajczyk and Schmid, of which the most recent [14] isan excellent survey of the field. ... 20] T. Tuytelaars and Luc Van Gool. Matching widely -
Explicitly Generating Complementary Systems for Large…
mi.eng.cam.ac.uk/~mjfg/breslin_INTER06.pdf22 Nov 2006: babi. lity. Figure 3: Training set word posteriors, G0. 0 10 20 30 40 50 60 70 80 90 100104. ... oCptqu=s [ , corresponding to32% of words is also marked. Approximately 20% of words havezero posterior probability with respect to G0. -
EUROGRAPHICS 2006 / E. Gröller and L. Szirmay-Kalos(Guest Editors) ...
mi.eng.cam.ac.uk/reports/svr-ftp/hernandez_eg06.pdf19 Sep 2006: These two steps are interleaved until convergence whichtakes about 20 steps for the sequences we experimentedwith. -
Discriminative Adaptation for Speaker Verification C. Longworth and…
mi.eng.cam.ac.uk/~mjfg/longworth_INTER06.pdf22 Nov 2006: 0.1 0.2 0.5 1 2 5 10 20 40. 0.1. 0.2. ... 20, pp. 210–229, 2005. [5] D. Povey, M.J.F. Gales, D.Y. Kim, and P.C. -
Multi-Sensory Face Biometric Fusion (for Personal Identification)…
mi.eng.cam.ac.uk/reports/svr-ftp/arandjelovic_OTCBVS06.pdf19 Mar 2006: a) Glasses ON. 0 20 40 60 80 100 1200.85. 0.9. ... IEEE Trans. Pattern Analysis and Machine Intelli-gence, 20(12):1295–1307, 1998. [21] T. -
JOURNAL OF IEEE TRANS. ACOUST., SPEECH, SIGNAL PROCESSING, JULY ...
mi.eng.cam.ac.uk/~mjfg/sim_SAP06.pdf22 Nov 2006: 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,. -
THE CU-HTK MANDARIN BROADCAST NEWS TRANSCRIPTION SYSTEM R. Sinha, ...
mi.eng.cam.ac.uk/research/projects/AGILE/publications/rs_ICASSP06.pdf23 Feb 2006: S1HLDA 12.4 6.6 21.1GAUSS 11.9 6.2 20.3. Table 2. Unadapted performance of baseline systems using HLDAor GAUSS front-ends, V1 segmentation and lm1.0. ... System LMCER (%). dev04f eval03m eval04. S2 GAUSS. lm1.0 12.2 5.6 20.0lm2.0 12.3 5.2 20.2lm2.1 11.5 5 -
pami04.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/hernandez_pami06.pdf19 Sep 2006: 17]–[20]. In particular, the works of [18] and [19] use onlythe two outermost epipolar tangents,. ... to match them across different views and handle their visibility, as proposed in [16] and [20]. -
Face Recognition from Video using the GenericShape-Illumination…
mi.eng.cam.ac.uk/reports/svr-ftp/arandjelovic_ECCV06.pdf17 Feb 2006: a) Histograms. 0 5 10 15 20 25 30 35 40 450.965. ... 20. D. M. Gavrila. Pedestrian detection from a moving vehicle. ECCV, 2:37–49, 2000.21. -
yu.dvi
mi.eng.cam.ac.uk/research/projects/AGILE/publications/ky_ICASSP06.pdf23 Feb 2006: Bayesian ML Train MPE TrainApprox. SI SAT SI SAT. — 32.83 — 29.20 —FI —- 32.90 — 29.74. ... Bayesian ML Train MPE TrainApprox SI SAT SI SAT. — 32.83 — 29.20 —MLthresh 31.23 — 27.81 —. ML 32.23 31.84 — 28.72MAP 30.92 30.40 — -
ESTIMATION OF DISPLACEMENTLOCATION FOR ENHANCED STRAIN IMAGING J. E.…
mi.eng.cam.ac.uk/reports/svr-ftp/lindop_tr550.pdf30 Mar 2006: We also present strain images for qualitative assessment. 13. 0 5 10 15 20 25 300. ... 20 dB data at0.5% strain as a function of c, the compression factor. -
The Layout Consistent Random Field for Recognizing and Segmenting ...
mi.eng.cam.ac.uk/reports/svr-ftp/shotton_cvpr06.pdf3 Apr 2006: the car), and also a subsetof 20 images from the UIUC car database [11], containingone completely visible car instance, which were segmentedby hand. ... Segmentation Accuracy: We evaluated the segmentationaccuracy on a randomly chosen subset of 20 of the -
PHASE-BASED ULTRASONICDEFORMATION ESTIMATION J. E. Lindop, G. M.…
mi.eng.cam.ac.uk/reports/svr-ftp/lindop_tr555.pdf25 May 2006: 0 20 40 60 80 100 1200. 0.5. 1. 1.5. 2. ... EPZSEPZS_A1EPZS_A2. 0 10 20 30 40 50 60 70 800. 5. -
stenger_imavis06.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/stenger_imavis06.pdf21 Sep 2006: One option that is commonlyfollowed is to independently train a classifier for each ob-ject [20]. ... InProc. 7th Int. Conf. on Computer Vi-sion, volume I, pages 20–27, Corfu, Greece, September1999. -
EUROGRAPHICS 2006 / E. Gröller and L. Szirmay-Kalos(Guest Editors) ...
mi.eng.cam.ac.uk/reports/svr-ftp/johnson_semantic06.pdf1 Jun 2006: IEEE transac-tions on PAMI 20, 12 (November 2001), 1222–1239. [CDM02] CUTLER B., DORSEY J., MCMILLAN L., MÜLLERM., JAGNOW R.: A procedural approach to authoring solid mod-els. ... IEEE Trans. Pattern Analysisand Machine Intelligence 20, 1 (1998), -
IEEE TRANS. ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2007 ...
mi.eng.cam.ac.uk/~mjfg/gales_ASL.pdf22 Nov 2006: This is generallydone in two separate stages, although it is possible to havemore integrated schemes which alter the segmentation duringthe clustering process [19], [20]. ... 0.16 0.18 0.2 0.22 0.24 0.2612. 14. 16. 18. 20. 22. -
pami04.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/stenger_pami06.pdf21 Sep 2006: 20). where in this equation the imageI is indexed by itsx andy-coordinates. ... 20. 30. 40. Frame. RM. S e. rror. (d). Figure 9: Error performance. -
article.dvi
mi.eng.cam.ac.uk/~mjfg/rosti_CSL04.pdf22 Nov 2006: Nine full iterations of embedded training were used with 20 withiniterations and 20 row by row transform iterations (Gales, 1999). ... matrix. Nine full iterationsof embedded training were used, each with 20 within iterations. -
Hole Filling Through Photomontage Marta Wilczkowiak∗, Gabriel J.…
mi.eng.cam.ac.uk/reports/svr-ftp/brostow_HoleFillingBMVC05.pdf14 Sep 2006: Real-time texturesynthesis by patch-based sampling. ACM Trans. Graph., 20(3):127–150, 2001. [11] D. -
TextonBoost: Joint Appearance, Shape andContext Modeling for…
mi.eng.cam.ac.uk/reports/svr-ftp/shotton_eccv06.pdf15 Feb 2006: The parameters were set as M = 5000, K = 400, θφ = [20, 2]T , and wλ = 4. ... In: CVPR01. (2001) I:511–518. 20. Belongie, S., Malik, J., Puzicha, J.: Shape matching and object recognition usingshape contexts. -
C:/SFWDoc/Academic/Publications/2006/ICPR_2006/Final_AppTrack/icpr_200…
mi.eng.cam.ac.uk/reports/svr-ftp/sfwong_icpr06b.pdf21 Sep 2006: D. Hager and P. N. Belhumeur. Efficient region trackingwith parametric models of geometry and illumination.PAMI,20(10):1025–1039, 1998. -
Ultrasound compounding withautomatic attenuation compensation using…
mi.eng.cam.ac.uk/reports/svr-ftp/treece_tr558.pdf22 Jun 2006: Typicallythis subset of data needs to be large, say 10 to 20 percent of the height of the image. ... 1998. Multi-angle compound imaging. UltrasonicImaging 20 (2), 81–102. Knipp, B. S., Zagzebski, J. -
Unsupervised Bayesian Detection of Independent Motion in Crowds…
mi.eng.cam.ac.uk/reports/svr-ftp/brostow_MotionInCrowdsCVPR06.pdf14 Sep 2006: 70. 60. 50. 40. 30. 20. 10. 0. 10. Combinations of pairing Ci C. ... 20] A. Shashua, Y. Gdalyahu, and G. Hayun. Pedestrian detection for driving as-sistance systems: Single frame classification and system level performance. -
1 Model-Based Hand Tracking Using a HierarchicalBayesian Filter…
mi.eng.cam.ac.uk/reports/svr-ftp/thayananthan_pami06.pdf14 Sep 2006: Asthe search proceeds, these regions are progressively. 0 10 20 30 40 50 60 70 80 900. ... 10. 20. 30. 40. Frame. RM. S e. rror. (a). 0 20 40 60 80 100 120 140 1600. -
SENSORLESS RECONSTRUCTIONOF UNCONSTRAINED FREEHAND 3D ULTRASOUND DATA …
mi.eng.cam.ac.uk/reports/svr-ftp/housden_tr553.pdf22 May 2006: 0 20 40 60 80 100 1200.5. 0. 0.5. 1. 1.5. ... et (. mm. ). 0 20 40 60 80 100 1200.5. -
AUGMENTED STATISTICAL MODELS FOR SPEECH RECOGNITION M.I. Layton and…
mi.eng.cam.ac.uk/~mjfg/layton_ICASSP06.pdf22 Nov 2006: Nodata or feature whitening was performed. ClassifierCriterion Componentsλ α 10 20. ... A point of particular interest is that despite poorer statesegmentation—the sufficient statistics fix the state segmentation—C-Aug models with ML statistics -
MODEL-BASED TECHNIQUES FORNOISE ROBUST SPEECH RECOGNITION Mark John…
mi.eng.cam.ac.uk/~mjfg/thesis.pdf5 Jun 2006: 91. 8.20 Word error rates (%) of iterative DPMC-compensated model sets with theExtended covariance approximation on Lynx Helicopter additive and con-volutional noise-corrupted RM at 10dB, indicates that tilt ... L(YT |M) = αN (T ) = β1(0) =N. j=1. αj -
C:/SFWDoc/Academic/Publications/2005/BMVC_2005/FinalPaper/bmvc_05_sfwo…
mi.eng.cam.ac.uk/reports/svr-ftp/sfwong_bmvc05.pdf21 Sep 2006: Thepercentage of test cases that cannot be mapped into any classis 20.3%. ... 7] T. Starner, J. Weaver, and A.P. Pentland. Real-time american signlanguage recognition usingdesk and wearable computer based video.PAMI, 20(12):1371–1375, December 1998. -
INCREMENTAL BAYESIAN ADAPTATION K. Yu and M.J.F. Gales Engineering ...
mi.eng.cam.ac.uk/~mjfg/yu_ICASSP06.pdf22 Nov 2006: The baseline performance is shown in table 1. The. Incremental Adaptation ML-SI MPE-SI— 32.83 29.20. ... 3. 5 10 15 20 25 30. 29. 30. 31. 32. -
IEEE TRANS. ON SAP, VOL. ?, NO. ??, ????? ...
mi.eng.cam.ac.uk/research/projects/AGILE/publications/mjfg_ASL.pdf23 Feb 2006: This is generallydone in two separate stages, although it is possible to havemore integrated schemes which alter the segmentation duringthe clustering process [19], [20]. ... 0.16 0.18 0.2 0.22 0.24 0.2612. 14. 16. 18. 20. 22. -
Incremental Learning of Temporally-CoherentGaussian Mixture Models…
mi.eng.cam.ac.uk/reports/svr-ftp/arandjelovic_SME06.pdf14 Mar 2006: Briefly, Zwolinski and Yang [20], and Figueredo and Jain [6] overestimate thecomplexity of the model and reduce it by discarding “improbable” components. ... 20] M. Zwolinski and Z. R. Yang. Mutual information theory for adaptive mixture -
Temporally Varying Model Parameters forLarge Vocabulary Continuous…
mi.eng.cam.ac.uk/~mjfg/sim_INTER05.pdf22 Nov 2006: atj(20). µmjaij. =2hitγ. mlm (t)(otj µmtj ). β̃mlmj(21). σ2mjaij. =2hitatj γ. -
thesis.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/nock_thesis.pdf14 Jun 2006: Speaking Style and Its Effects 20. (F1) than for the more formal, planned studio speech condition (F0) [126]. -
Semi-supervised Learning of Joint DensityModels for Human Pose…
mi.eng.cam.ac.uk/reports/svr-ftp/navaratnam_semi_supervised.pdf14 Sep 2006: ln = 19.99RMS= 20.83. ln = 31.49RMS = 11.61. ln = 71.82RMS= 20.81. ... ln = 98.76RMS= 34.40. ln = 18.46RMS= 10.85. ln = 20.00RMS= 31.02. -
Joint Uncertainty Decoding for Noise Robust Speech Recognition H. ...
mi.eng.cam.ac.uk/~mjfg/liao_INTER05.pdf19 Dec 2006: Uncertainty 1 4 16 256. Clean — 33.2. SPLICENo. 24.6 20.7 17.0 12.3FE-CMLLR 16.3 15.3 12.8 13.5. -
C:/SFWDoc/Academic/Publications/2005/ICCV_HCI_2005/FinalPaper/iccv_hci…
mi.eng.cam.ac.uk/reports/svr-ftp/sfwong_iccv_hci05.pdf21 Sep 2006: We used different amount of trainingdata per each pair to re-train the classifier (sample sizes used are 10, 20, 50, 100,300). ... PAMI 20 (1998) 1371–1375. 2. Vogler, C., Metaxas, D.: A framework for recognizing the simultaneous aspects ofamerican -
DEVELOPMENT OF THE CUHTK 2004 MANDARIN CONVERSATIONAL TELEPHONESPEECH …
mi.eng.cam.ac.uk/~mjfg/gales_ICASSP05.pdf22 Nov 2006: ldc04swm03S2 12 36.3 48.2S3 16 36.1 47.9S4 20 36.0 47.2. Table 3. -
Hierarchical Part-Based Human Body PoseEstimation R. Navaratnam∗ A.…
mi.eng.cam.ac.uk/reports/svr-ftp/navaratnam_hierarchical.pdf14 Sep 2006: to [20]. The dimensions of this model are generic and not adopted to fit each person. ... 20 highest rankedposes (in 98% of such instances) as illustrated in figure 4.
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.