In this paper we propose several methodologies for the use of feature integration and evaluate them in a low-latency classification framework. These general methodologies are based on three key aspects that will be assessed in this study: the selection of the features that have to be temporally integrated, the choice of the integration techniques, i.e., how the temporal information is extracted, and the size of the integration window. The experiments carried out for the speech/music/mix classification task show that the different methodologies have a significant impact on the global performance. Compared to the state of the art procedures, the methodologies we proposed achieved the best performance, even with the low-latency constraints.
Authors:
Flocon-Cholet, Joachim; Faure, Julien; Guérin, Alexandre; Scalart, Pascal
Affiliations:
Orange Labs, Lannion, France; INRIA/IRISA, Université de Rennes, Rennes, France(See document for exact affiliation information.)
AES Convention:
137 (October 2014)
Paper Number:
9180
Publication Date:
October 8, 2014
Subject:
Signal Processing
Click to purchase paper as a non-member or you can login as an AES member to see more options.
No AES members have commented on this paper yet.
To be notified of new comments on this paper you can subscribe to this RSS feed. Forum users should login to see additional options.
If you are not yet an AES member and have something important to say about this paper then we urge you to join the AES today and make your voice heard. You can join online today by clicking here.