Search
Search Funnelback University
- Refined by:
- Date: 2009
1 -
10 of
21
search results for katalk:za33 24 |u:mi.eng.cam.ac.uk
where 0
match all words and 21
match some words.
Results that match 1 of 2 words
-
3D ULTRASONIC STRAIN IMAGINGUSING FREEHAND SCANNING ANDA…
mi.eng.cam.ac.uk/reports/svr-ftp/housden_tr623.pdf28 Jan 2009: 22, 24, 25, 26]. ... Ultrasound in Medicine and Biology, 20(1):27–33,1994. 16. [24] H. Ponnekanti, J. -
CAMBRIDGE UNIVERSITY ENGINEERING DEPARTMENT NOISY CMLLR FOR…
mi.eng.cam.ac.uk/~mjfg/kim_tr611.pdf18 Feb 2009: 24). where γ(m)o,t is the posterior probability of component m given the observation sequence O, bias. ... parameter Mb, and model set M. Differentiating equation 24 with respect to µ̂(rm)b and Σ̂(rm). -
is2008.dvi
mi.eng.cam.ac.uk/~mjfg/raut_INTER08.pdf2 Mar 2009: SI — MLLR 24.3 27.0MLLR MLLR 23.6 26.4. DSAT DLT DLT 18.4 28.1MLLRDMT MLLRDMT 22.5 25.3. -
Context Dependent Language Model Adaptation X. Liu, M. J. ...
mi.eng.cam.ac.uk/~mjfg/liu_INTER08.pdf2 Mar 2009: The largest two corpora, giga-cna(0.46) and giga-xin (0.24) were heavily weighted. -
paper.2col.dvi
mi.eng.cam.ac.uk/~mjfg/kai_ASP09.pdf2 Mar 2009: is introduced to map, for example, ML-trained trans-forms into discriminative transforms [24]. ... This had a gain of 0.24%. k. yu et al.: UNSUPERVISED ADAPTATION WITH DISCRIMINATIVE MAPPING TRANSFORMS 9. -
IEEE TRANSACTIONS ON AUDIO, SPEECH AND LANGUAGE PROCESSING, VOL. ...
mi.eng.cam.ac.uk/~mjfg/cl336_ASP09.pdf2 Mar 2009: The primal weight vectorisgiven by. wk =. N. i=1. αiyiβkφk(Oi; λ) (24). ... At each iterationJ(s, β) can be estimated using astandard efficient SVM implementation such as [24]. -
Uniform precision ultrasound strain imaging G.M. Treece, J.E. Lindop, …
mi.eng.cam.ac.uk/reports/svr-ftp/treece_tr624.pdf9 Mar 2009: 6 3 ANALYSIS. 20 40 60 80 100sample number. r2=a. r2=24 a. -
Model-Based Approaches to Speaker andEnvironment Adaptation Mark…
mi.eng.cam.ac.uk/~mjfg/mjfg_china09.pdf28 Apr 2009: Cambridge UniversityEngineering Department. Tsinghua University April 2009 24. Model-Based Approaches to Speaker and Environment Adaptation. ... Ratio of corrupted speech magnitudeto clean speech magnitude. • magnitude (α = 1, γ = 1)• power (α = 0, -
Sequence Kernels for Speaker and SpeechRecognition Mark Gales - ...
mi.eng.cam.ac.uk/~mjfg/jhu09.pdf16 Jul 2009: Cambridge UniversityEngineering Department. JHU Workshop 2009 24. Sequence Kernels for Speaker and Speech Recognition. -
Learning to Track with Multiple Observers Björn StengerComputer…
mi.eng.cam.ac.uk/~cipolla/publications/inproceedings/2009-CVPR-hand-trackingpdf28 Jul 2009: discriminativepower trade-off was explored [24]. It is possible to use each observer as a single compo-nent in a tracking algorithm. ... InFourth Asian Conference onComputer Vision, Taipei, Taiwan, January 2000.1. [24] M.
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.