Title Speaker identification accuracy improvement using BLSTM neural network /
Authors Dovydaitis, Laurynas ; Rudžionis, Vytautas
Full Text Download
Is Part of Indian journal of computer science and engineering.. Vandalur, Chennai : Engg Journals Publications. 2018, vol. 9, iss. 2, p. 31-37.. ISSN 2231-3850. eISSN 0976-5166
Keywords [eng] neural network ; speaker identification ; LSTM ; spoken language
Abstract [eng] In this work we analyze speaker identification accuracy on Lithuanian speaker dataset LIEPA. This dataset consists of 370 Lithuanian speakers reading given text samples. We preform speaker identification with HMM classification and then repeat the same test with different types of LSTM and BLSTM neural networks. On the given dataset we experimentally observe speaker identification accuracy improvement from 3% to 6% compared to best HMM implementation.
Published Vandalur, Chennai : Engg Journals Publications
Type Journal article
Language English
Publication date 2018