Search
Search Funnelback University
- Refined by:
- Date: 2004
31 -
39 of
39
search results for news |u:mi.eng.cam.ac.uk
Fully-matching results
-
3D ULTRASOUND PROBECALIBRATION WITHOUT A POSITION SENSOR A. H. ...
mi.eng.cam.ac.uk/reports/svr-ftp/gee_tr488.pdf4 Sep 2004: The results for the new phantom are in the bottom two blocks of Table 3. ... Unusually, for this new technique we are able to assess the accuracy of the calibration processalone. -
paper.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/liu_icassp2004.pdf29 May 2004: When varying the number of Gaussian components ina state it is impractical to obtain new statistics for each number ofcomponents, even if the state alignments are fixed. -
CAMBRIDGE UNIVERSITYENGINEERING DEPARTMENT SWITCHINGLINEAR DYNAMICAL…
mi.eng.cam.ac.uk/reports/svr-ftp/rosti_tr461.pdf29 Jan 2004: In both models,the new discrete state, qt1, is conditionally independent on the history of the discrete statesgiven the state at time t. ... 16. 5 Conclusions. This paper has introduced a new method to train and evaluate switching linear dynamical -
Cluster Voting for Speaker Diarisation S.E.…
mi.eng.cam.ac.uk/reports/svr-ftp/tranter_tr476.pdf13 May 2004: Experiments presented using two inputsconfirm that the diarisation error rate can be reduced using this new method. ... 32.2 The Broadcast News Data for Diarisation. 32.3 Diarisation Scoring. 4. -
GENERATING AND EVALUATING SEGMENTATIONS FOR AUTOMATIC…
mi.eng.cam.ac.uk/reports/svr-ftp/tranter_icassp04.pdf25 May 2004: This suggeststhat predicting the WER on the eval02 superset of data using justthe dry03 subset, can be done with as much confidence using thediarisation score as the WER, so new segmentations -
CAMBRIDGE UNIVERSITYENGINEERING DEPARTMENT GENERALISEDLINEAR GAUSSIAN …
mi.eng.cam.ac.uk/reports/svr-ftp/rosti_tr420.pdf26 Jan 2004: However it requires some manipulation to obtain the. new parameters using convenient matrix operations. ... The new model parameters can be obtainedby the following two re-estimation formulae. -
RAO-BLACKWELLISED GIBBS SAMPLING FOR SWITCHING LINEAR…
mi.eng.cam.ac.uk/reports/svr-ftp/rosti_icassp2004.pdf22 May 2004: Further results and analysis are presented in [6]. 5. CONCLUSIONS. This paper has introduced a new method to train and evaluateswitching linear dynamical systems. ... The new scheme is based onMCMC simulation of the discrete state space and takes -
Sparse Finite Elements for Geodesic Contours with Level-Sets? Martin…
mi.eng.cam.ac.uk/reports/svr-ftp/weber_eccv2004.pdf18 Jun 2004: 2 Efficient representation with ‘banded’ finite elements. The new numerical representation u consists of a global and a local component:. – ... 6 Conclusions and future work. We have proposed a new level-set scheme and algorithm for automatically -
Linear Gaussian Models for Speech Recognition
mi.eng.cam.ac.uk/reports/svr-ftp/rosti_thesis.pdf8 Oct 2004: Using this approximation,. a new acoustic model called the factor analysed HMM (FAHMM) is presented. ... that it can be divided into short frames, often 10ms. In each frame period a new observation vec-.
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.