OldEvents2006
From SpeechWiki
(New page: == February 13, 2006 == === IFP Special Seminar 11:00am , BI 1215A=== "Did the Great Masters "Cheat" Using Optics? Computer Image Analysis Sheds Light on a Controversial Theory" David G. ...) |
|||
Line 1: | Line 1: | ||
+ | {| align="right" | ||
+ | | __TOC__ | ||
+ | |} | ||
+ | |||
== February 13, 2006 == | == February 13, 2006 == | ||
=== IFP Special Seminar 11:00am , BI 1215A=== | === IFP Special Seminar 11:00am , BI 1215A=== |
Revision as of 16:23, 5 January 2009
|
February 13, 2006
IFP Special Seminar 11:00am , BI 1215A
"Did the Great Masters "Cheat" Using Optics? Computer Image Analysis Sheds Light on a Controversial Theory"
David G. Stork, Ricoh Innovations
February 20, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic: The IFP/ISLE linux cluster (Bowon Lee)
You can access the IFP/ISLE linux cluster by typing
$ ssh ifp-32.ifp.uiuc.edu
using your user ID for the IFP network.
In order to use the cluster, please create your own directory under '/cworkspace/ifp-32-1/hasegawa/' and work in that directory because the slave nodes in the cluster cannot see any partition mounted through the network such as your home directory.
I've created documentation
Job Queuing in a Linux Cluster
wihch describes job queuing using the Sun Grid Engine (SGE)
and
Parallel Processing of the HTK Commands
which describes the Perl scripts for parallel processing of some of the HTK commands.
As an example, you can download another Perl script test.pl
which uses 'HVite.pl' and 'HResults.pl' for parallel processing of 'HVite' and 'HResults'.
This Perl script is a part of the Perl script 'train.pl' provided by Professor Hasegawa-Johnson and assumes that all the trained model exists under the directory 'mmf' that 'train.pl' generated.
- Slides
- Background
February 27, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic : Speech Recognition with Missing Features (Bowon Lee)
- Article
"'Robust automatic speech recognition with missing and unreliable acoustic data"
Martin Cooke, et. al.
Speech Communication
Volume 34, Issue 3, May 2001, Pages 267-285
This article can be obtained form here
The link should be accessible from any computers on campus.
Additional article: Raj and Stern 2005
- Slides
- Background
March 6, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic
- Slides
- Background
March 13, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic : Beamforming and postfiltering for speech enhancement
"An algorithm for linearly constrained adaptive array processing" by O. L. Frost, Proceedings of the IEEE, 1972
"An alternative approach to linearly constrained adaptive beamforming" by L. J. Griffiths and C. W. Jim, IEEE Trans. on Antennas and Propagation, 1982
"Speech enhancement using a minimum-mean square error short-time spectral amplitude estimator" by Ephraim and Malah, IEEE Trans on ASSP, 1984
"Speech enhancement using a minimum mean-square error log-spectral amplitude estimator" by Ephraim and Malah, IEEE Trans on ASSP, 1985
"Multichannel Post-filtering in Nonstationary Noise Environment" by Israel Cohen,IEEE Trans. on Signal Processing, Vol. 52, No. 5, May 2004, pp. 1149-1160
March 27, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic:
- Slides
- Background
April 3, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
April 3
10th April, 2006
ICSLP new deadline April 10
http://www.interspeech2006.org/
April 10, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Updated time for group meet April 10
The Powerpoint file for the presentation, Xiaodan
Weighted finite-state transducers in speech recognition, Mohri et al
Topic: Weighted finite-state transducers in speech recognition
Authors:Mehryar Mohri, Fernando Pereira and Michael Riley
Abstract:
We survey the use of weighted finite-state transducers (WFSTs) in speech recognition. We show that WFSTs provide a common and natural representation for hidden Markov models (HMMs), context-dependency, pronunciation dictionaries, grammars, and alternative recognition outputs. Furthermore, general transducer operations combine these representations flexibly and efficiently. Weighted determinization and minimization algorithms optimize their time and space requirements, and a weight pushing algorithm distributes the weights along the paths of a weighted transducer optimally for speech recognition.
Presenter: Xiaodan
April 17, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic: Analysis of Pitch Contours in Repetition-Disflueny using Stem-ML
Abstract
F0 analysis-by-synthesis methods are used in order to test the hypothesis that the pitch contour in the alteration segment of disfluency tends to mimic the pitch contour in the reparandum segment of that disfluency. This hypothesis is the same as the hypothesis addressed in the Cole et al., DISS 2005 paper but a different speech syntheses model ie Stem-ML is used to test the hypothesis rather than ANOVA. The results show that this hypothesis is not supported by experimental analysis.
by Rajiv
April 24, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic: Continued talk from the last topic "Beamforming and postfiltering for speech enhancement" (Lae-Hoon Kim)
May 1, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
Topic
- Slides
- Background
May 10, 2006
ISLE/LVCSR group meeting 11:00am , BI 3169
(Arthur) Topic Graphical Models and Junction Tree algorithm
- Slides coming soon
- Background
Probabilistic Modelling and Reasoning The Junction Tree Algorithm by David Barber
The Junction Tree Algorithms by Mark Paskin
Optimal Junction Trees by Jensen and Jensen
Exact Inference by Junction/Join/Clique Trees by David Page
Search Graphical Models for the slide Art used CS 498 slides
May 11, 2006
ISLE/LVCSR group meeting 1:00pm , BI 3169
(Arthur again) Topic: continuing Graphical Models and Junction Tree algorithm