Abstract | ||
---|---|---|
We address the closed-set problem of speaker identification by presenting a novel sparse representation classification algorithm. We propose to develop an over complete dictionary using the GMM mean super vector kernel for all the training utterances. A given test utterance corresponds to only a small fraction of the whole training database. We therefore propose to represent a given test utterance as a linear combination of all the training utterances, thereby generating a naturally sparse representation. Using this sparsity, the unknown vector of coefficients is computed via l1minimization which is also the sparsest solution [12]. Ideally, the vector of coefficients so obtained has nonzero entries representing the class index of the given test utterance. Experiments have been conducted on the standard TIMIT [14] database and a comparison with the state-of-art speaker identification algorithms yields a favorable performance index for the proposed algorithm. |
Year | DOI | Venue |
---|---|---|
2010 | 10.1109/ICPR.2010.1083 | Pattern Recognition |
Keywords | Field | DocType |
Gaussian processes,minimisation,performance index,signal classification,signal representation,speaker recognition,GMM mean super vector kernel,closed set problem,minimization,performance index,sparse representation classification algorithm,speaker identification,test utterance correspond | Kernel (linear algebra),TIMIT,Linear combination,Algorithm design,Pattern recognition,Computer science,Sparse approximation,Speech recognition,Speaker recognition,Artificial intelligence,Gaussian process,Hidden Markov model | Conference |
ISSN | ISBN | Citations |
1051-4651 | 978-1-4244-7542-1 | 34 |
PageRank | References | Authors |
1.27 | 11 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Imran Naseem | 1 | 142 | 13.51 |
Roberto Togneri | 2 | 814 | 48.33 |
M. Bennamoun | 3 | 3197 | 167.23 |