Combined i-Vector and Extreme Learning Machine Approach for Robust Speaker Identification and Evaluation with SITW 2016, NIST 2008, TIMIT Databases

In this article, a novel combined i-vector and an Extreme Learning Machine (ELM) is proposed for speaker identification. The ELM chosen because it fast to train has universal approximator property. Four combinations of features based on Mel Frequency Cepstral Coefficient Power Normalized are used. Besides, seven fusion methods exploited. system evaluated with three different databases, namely: the SITW 2006, NIST 2008, TIMIT database. This work employs 2016 database first time identification using integration between approach. From each database, 120 speakers 1200 speech utterances used (overall 360 3600 utterances). Furthermore, comprehensive evaluations exploited wide range realistic background noise types (Stationary AWGN Non-Stationary Noise types) handset effect. compared Gaussian Mixture Model-Universal Background Model (GMM-UBM) other states art approaches. results show that method outperforms GMM-UBM approach state- of-the-art under specific conditions, techniques can be improve robustness effects.