DocumentCode :
1928818
Title :
Discovering hierarchical speech features using convolutional non-negative matrix factorization
Author :
Behnke, Sven
Author_Institution :
Int. Comput. Sci. Inst., Berkeley, CA, USA
Volume :
4
fYear :
2003
fDate :
20-24 July 2003
Firstpage :
2758
Abstract :
Discovering a representation that reflects the structure of a dataset is a first step for many inference and learning methods. This paper aims at finding a hierarchy of localized speech features that can be interpreted as parts. Non-negative matrix factorization (NMF) has been proposed recently for the discovery of parts-based localized additive representations. The author proposes a variant of this method, convolutional NMF, that enforces a particular local connectivity with shared weights. Analysis starts from a spectrogram. The hidden representations produced by convolutional NMF are input to the same analysis method at the next higher level. Repeated application of convolutional NMF yields a sequence of increasingly abstract representations. These speech representations are parts-based, where complex higher-level parts are defined in terms of less complex lower-level ones.
Keywords :
feature extraction; matrix decomposition; speech recognition; convolutional NMF; hierarchical speech feature discovery; inference methods; learning methods; local connectivity; localized speech features; nonnegative matrix factorization; parts-based localized additive representations; parts-based speech representations; shared weights; spectrogram; Analysis of variance; Automatic speech recognition; Computer science; Convolution; Feature extraction; Independent component analysis; Mel frequency cepstral coefficient; Pattern recognition; Psychoacoustic models; Unsupervised learning;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Neural Networks, 2003. Proceedings of the International Joint Conference on
ISSN :
1098-7576
Print_ISBN :
0-7803-7898-9
Type :
conf
DOI :
10.1109/IJCNN.2003.1224004
Filename :
1224004
Link To Document :
بازگشت