Online Non-Negative Convolutive Pattern Learning for Speech Signals

The unsupervised learning of spectro-temporal patterns within speech signals is of interest in a broad range of applications. Where patterns are non-negative and convolutive in nature, relevant learning algorithms include convolutive non-negative matrix factorization (CNMF) and its sparse alternativ...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:IEEE transactions on signal processing Ročník 61; číslo 1; s. 44 - 56
Hlavní autori: Dong Wang, Vipperla, R., Evans, N., Zheng, T. F.
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: New York, NY IEEE 01.01.2013
Institute of Electrical and Electronics Engineers
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Predmet:
Algorithms
Applied sciences
Coding, codes
Computing costs
convolutive NMF
Cost function
Detection, estimation, filtering, equalization, prediction
Distance learning
Exact sciences and technology
Image reconstruction
Information, signal and communications theory
Learning
Non-negative matrix factorization
On-line systems
Online
online pattern learning
Signal and communications theory
Signal processing
Signal, noise
sparse coding
Speech
Speech processing
speech recognition
Statistics
Studies
Telecommunications and information theory
Training
Training data
E-learning
Acoustic signal
Non-negative matrix factorization
Updating
online pattern learning
Factorization
sparse coding
Remote teaching
Unsupervised learning
Implementation
Non negative matrix
Vocal signal
Coding
Volume
Batch process
Speech recognition
On line processing
Signal processing
Cost function
Objective function
Learning algorithm
Speech processing
convolutive NMF
ISSN:1053-587X, 1941-0476
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:The unsupervised learning of spectro-temporal patterns within speech signals is of interest in a broad range of applications. Where patterns are non-negative and convolutive in nature, relevant learning algorithms include convolutive non-negative matrix factorization (CNMF) and its sparse alternative, convolutive non-negative sparse coding (CNSC). Both algorithms, however, place unrealistic demands on computing power and memory which prohibit their application in large scale tasks. This paper proposes a new online implementation of CNMF and CNSC which processes input data piece-by-piece and updates learned patterns gradually with accumulated statistics. The proposed approach facilitates pattern learning with huge volumes of training data that are beyond the capability of existing alternatives. We show that, with unlimited data and computing resources, the new online learning algorithm almost surely converges to a local minimum of the objective cost function. In more realistic situations, where the amount of data is large and computing power is limited, online learning tends to obtain lower empirical cost than conventional batch learning.
Bibliografia:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Feature-1
content type line 14
content type line 23
ISSN:1053-587X
1941-0476
DOI:10.1109/TSP.2012.2222381