Difference between revisions of "Publications/dehak.09.interspeechb"
From LRDE
Line 7: | Line 7: | ||
| note = Open Performance Sub-Challenge Prize |
| note = Open Performance Sub-Challenge Prize |
||
| urllrde = 200909-INTERSPEECH-A |
| urllrde = 200909-INTERSPEECH-A |
||
− | | abstract = In this paper, we describe systems that were developed for the Open Performance Sub-Challenge of the INTERSPEECH 2009 Emotion Challenge. We participate to both two-class and five-class emotion detection. For the two-class problemthe best performance is obtained by logistic regression fusion of three systems. Theses systems use short- and long-term speech features. This fusion achieved an absolute improvement of 2,6% on the unweighted recall value compared with [6]. For the five-class problem, we submitted two individual systems: cepstral GMM vs. long-term GMM-UBM. |
+ | | abstract = In this paper, we describe systems that were developed for the Open Performance Sub-Challenge of the INTERSPEECH 2009 Emotion Challenge. We participate to both two-class and five-class emotion detection. For the two-class problemthe best performance is obtained by logistic regression fusion of three systems. Theses systems use short- and long-term speech features. This fusion achieved an absolute improvement of 2,6% on the unweighted recall value compared with [6]. For the five-class problem, we submitted two individual systems: cepstral GMM vs. long-term GMM-UBM. The best result comes from a cepstral GMM and produced an absolute improvement of 3,5% compared to [6]. |
| lrdenewsdate = 2009-06-22 |
| lrdenewsdate = 2009-06-22 |
||
| type = inproceedings |
| type = inproceedings |
Revision as of 09:00, 10 September 2014
- Authors
- Pierre Dumouchel, Najim Dehak, Yazid Attabi, Réda Dehak, Narjès Boufaden
- Where
- Interspeech
- Type
- inproceedings
- Date
- 2009-06-22
Abstract
In this paper, we describe systems that were developed for the Open Performance Sub-Challenge of the INTERSPEECH 2009 Emotion Challenge. We participate to both two-class and five-class emotion detection. For the two-class problemthe best performance is obtained by logistic regression fusion of three systems. Theses systems use short- and long-term speech features. This fusion achieved an absolute improvement of 2,6% on the unweighted recall value compared with [6]. For the five-class problem, we submitted two individual systems: cepstral GMM vs. long-term GMM-UBM. The best result comes from a cepstral GMM and produced an absolute improvement of 3,5% compared to [6].
Bibtex (lrde.bib)
@InProceedings{ dehak.09.interspeechb, author = {Pierre Dumouchel and Najim Dehak and Yazid Attabi and R\'eda Dehak and Narj\`es Boufaden}, title = {Cepstral and Long-Term Features for Emotion Recognition}, booktitle = {Interspeech}, year = 2009, month = sep, note = {Open Performance Sub-Challenge Prize}, abstract = {In this paper, we describe systems that were developed for the Open Performance Sub-Challenge of the INTERSPEECH 2009 Emotion Challenge. We participate to both two-class and five-class emotion detection. For the two-class problem, the best performance is obtained by logistic regression fusion of three systems. Theses systems use short- and long-term speech features. This fusion achieved an absolute improvement of 2,6\% on the unweighted recall value compared with [6]. For the five-class problem, we submitted two individual systems: cepstral GMM vs. long-term GMM-UBM. The best result comes from a cepstral GMM and produced an absolute improvement of 3,5\% compared to [6].} }