Speaker Verification from Short Utterance Perspective: A Review

摘要

Speaker recognition has emerged as an important field over the past several decades and has evolved over time. This work is attempted to investigate speaker verification (SV), particularly focusing on short utterances for the purpose of person authentication. The significance of short utterance-based SV is highlighted from its potential for practical deployment in person authentication-based application. Although the general view of short utterances is towards the text-dependent SV, this work focuses on the scope of text-independent SV using short utterances in relation to develop a robust SV framework in application-oriented scenario. First, it includes brief review of different features, modelling techniques, and classifiers that are widely used in the area of text-independent SV. The work then focuses on highlighting probable directions suitable from the outlook of short utterance-based SV. Three different directions are categorized based on their similar aspects that can be seen as, use of different/complementary features, utilization of acoustic-phonetic information and patten recognition and normalization techniques useful for handling short utterances. We bring the end to this work by bringing out a framework involving the directions discussed into a common platform for having a deployable system using short utterance-based SV.