Automatic speaker recognition of spanish siblings: (monozygotic and dizygotic) twins and non-twin brothers

Abstract

The performance of the automatic speaker recognition (ASR) system Batvox TM (Version 4.1) has been tested with a male population of 24 monozygotic (MZ) twins, 10 dizygotic (DZ) twins, 8 non-twin siblings and 12 unrelated speakers (aged 18-52 with Standard Peninsular Spanish as their mother tongue). Since the cepstral features in which this ASR system is based depend largely on anatomical-physiological foundations, we hypothesized that such features ought to be gene-dependent. Therefore, higher similarity values should be found in MZ twins (100% shared genes) than in DZ twins, in brothers (B) or in a reference population of unrelated speakers (US). Results corroborated the expected decreasing scale MZ > DZ > B > US since the similarity coefficients yielded by the automatic system for these speakers decreased exactly in the same direction as the kinship degree of the four speaker groups diminishes. This suggests that the system features are to a great extent genetically conditioned and that they are hence useful and robust for comparing speech samples of known and unknown origin, as found in legal cases. Furthermore, the 9.9% EER (Equal Error Rate) obtained when testing MZ pairs lies around the same value (11% EER) found in Künzel (2010) with German twins.

A Long Short-term Memory Neural Network for Improved Twins' Voice Differentiation

“…The results from these experiments show that identical twins consistently pose a greater problem to the distinguishing capabilities of voice matching algorithms than non-related individuals as corroborated by Segundo et al [42]. Due to the nature of the present data, it was not possible to replicate all previous works' experiments.…”

“…The only study to investigate not only MZ twins but also DZ twins and their siblings was conducted by San Segundo and Kunzel [42]. The same set of matching experiments as in [40] [43], adjustment of speaker models [44] or score normalization [45].…”

Arabic Speech Recognition Based on Knn, J48, and LVQ

“…The purpose of automatic speaker recognition is to identify a person from pronounced speech, this work can be further turned to the process of the identification or verification where the first matching is 1: 1 (claiming an identity and matcher task is to accept or reject the individual), or the latter, which is a 1: N match (the identification of speaker depends on making comparison to N registered speakers) [12]. Fig.…”

“…Seemingly, text-independent is seen much more difficult task to achieve in speaker recognition system. Generally, choosing the features of any Speaker Recognition system is regarded a key matter in order to obtain an accurate performance, where this system must have high variability between speakers and low variability within each person,in addition to other standards [12].…”

“…The traits are described of being environmentally-influenced that anyone can acquire them by education, place of living, and the social or family environment, socio-economic status. Remarkably, these features could Figure 1: Speaker verification diagram adapted from [12] have many advantages and the most striking one is robustness to channel effects and noise and. However, they entail advanced extraction techniques [14].…”