Open AccessSimplification of I‐Vector Extraction for Speaker Identification
Author(s) -
Xu Longting,
Yang Zhen,
Sun Linhui
Publication year - 2016
Publication title -
chinese journal of electronics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.267
H-Index - 25
eISSN - 2075-5597
pISSN - 1022-4653
DOI - 10.1049/cje.2016.10.016
Subject(s) - identification (biology) , vector space , computer science , qr decomposition , matrix decomposition , pattern recognition (psychology) , series (stratigraphy) , feature vector , algorithm , matrix (chemical analysis) , diagonal , identity (music) , eigenvalues and eigenvectors , decomposition , artificial intelligence , mathematics , pure mathematics , botany , biology , paleontology , ecology , physics , materials science , geometry , quantum mechanics , acoustics , composite material
The identity vector (i‐vector) approach has been the state‐of‐the‐art for text‐independent speaker recognition, both identification and verification in recent years. An i‐vector is a low‐dimensional vector in the socalled total variability space represented with a thin and tall rectangular matrix. This paper introduces a novel algorithm to improve the computational and memory requirements for the application. In our method, the series of symmetric matrices can be represented by diagonal expression, sharing the same dictionary, which to some extent is analogous to eigen decomposition, and we name this algorithm Eigen decomposition like factorization (EDLF). Similar algorithms are listed for comparison, in the same condition, our method shows no disadvantages in identification accuracy.