Speaker recognition based on multilevel speech signal analysis on Polish corpus

Szymon Drgas,Adam Dabrowski

doi:10.1007/s11042-013-1502-0

Szymon Drgas, Adam Dabrowski

Open Access

https://doi.org/10.1007/s11042-013-1502-0

Copy DOI

Abstract

This article deals with a new approach to the text-independent speaker verification task. It is namely proposed to combine spectral and the so-called high-level features (prosodic, articulatory, and lexical) in order to increase accuracy of speaker verification. The presented experiments were performed using a Polish language corpus developed by the authors, the so-called PUEPS corpus. It contains semi-spontaneous telephone conversations (acted emergency telephone notifications) recorded in laboratory conditions. As the Polish language is under resourced and the PUEPS corpus is relatively small, in this case a new approach is needed, other than these well known from NIST (National Institute of Standards and Technology) evaluations. The authors proposed to use the fast scoring instead of more complex classifiers and the AdaBoost (adaptive boosting) algorithm for features combination. Combination of features resulted in the equal error rate (EER) reduction for various SNR (signal-to-noise ratio) conditions. Additionally, score normalization methods were evaluated. It was shown that significant benefits can be obtained using the z-norm2 method.

Highlights

In this article the text-independent speaker verification task is considered in the context of emergency telephone conversations
In this paper a method is proposed that is a combination of algorithms known from the literature—cosine similarity system with scoring methods z-norm and z-norm2 (z-norm modified by authors), where features combination is performed by means of the AdaBoost algorithm
The about two times higher equal error rate (EER) has been brought for prosodic features (21.72 %)

Summary

Introduction

In this article the text-independent speaker verification task is considered in the context of emergency telephone conversations. Dabrowski et al [9] developed Polish language corpus called PUEPS. It contains semi-spontaneous telephone conversations recorded in laboratory conditions. In this paper a method is proposed that is a combination of algorithms known from the literature—cosine similarity system with scoring methods z-norm and z-norm (z-norm modified by authors), where features combination is performed by means of the AdaBoost algorithm. This method turned out to be effective for the corpus with Polish speech. The use of various scoring methods has been investigated

Objectives

Methods

Results

Conclusion