Automatic Speaker Recognition System Based on Gaussian Mixture Models, Cepstral Analysis, and Genetic Selection of Distinctive Features

Abstract

This article presents the Automatic Speaker Recognition System (ASR System), which successfully resolves problems such as identification within an open set of speakers and the verification of speakers in difficult recording conditions similar to telephone transmission conditions. The article provides complete information on the architecture of the various internal processing modules of the ASR System. The speaker recognition system proposed in the article, has been compared very closely to other competing systems, achieving improved speaker identification and verification results, on known certified voice dataset. The ASR System owes this to the dual use of genetic algorithms both in the feature selection process and in the optimization of the system’s internal parameters. This was also influenced by the proprietary feature generation and corresponding classification process using Gaussian mixture models. This allowed the development of a system that makes an important contribution to the current state of the art in speaker recognition systems for telephone transmission applications with known speech coding standards.

Behavioral features of the speech signal as part of improving the effectiveness of the automatic speaker recognition system

“…At the same time, for the same base and the times of the training and testing segments, the system based on physical features achieved results equal to 86.26% and 98.21%. Such significant differences in the performance of the two systems prompted the authors to perform a solution fusion to maximize the identification performance of the physical feature-based solution (Kamiński , Dobrowolski, 2022) . The next step of testing a behavioral ASR system is to use it as a selector that initially narrows down the number of cases closest to a given voice collected in the voice database.…”

“…The next step of testing a behavioral ASR system is to use it as a selector that initially narrows down the number of cases closest to a given voice collected in the voice database. Then -within this limited set -the physical featurebased system searches for the right object using a classifier based on Gaussian mixtures (Kamiński , Dobrowolski, 2022). The results of these tests are shown in Tables 1 and 2.…”

The Effect of Speaker Variability on Distinguishability of Bonafide and Synthetized Speech

“…Для моделирования распределений вероятности голосовых признаков в данной работе используется модель смеси гауссовых распределений [17] с количеством компонентов равным 512, по аналогии с конфигурацией базовой системы обнаружения спуфинга в конкурсе ASVspoof 2021 [4]. Мы выбрали данную вероятностную модель в связи с тем, что она наиболее широко используется для аппроксимации вероятностных распределений различных голосовых признаков [17].…”

“…В качестве множества X выступает пространство возможных значений коэффициентов LFCC. Универсальное распределение голосовых признаков всех дикторов P (x) моделируется путём обучения модели смеси гауссовых распределений при помощи алгоритма максимизации ожидания [17]. Модели смеси гауссовых распределений конкретных дикторов формируются путём MAP-адаптации [17] модели смеси гауссовых распределений, соответствующей универсальному распределению всех дикторов.…”

Evaluating the Efficacy of Traditional Machine Learning Models in Speaker Recognition: A Comparative Study Using the LibriSpeech Dataset

“…Pioneering studies in this domain employed basic linear models, which laid the groundwork for more complex approaches. The introduction of Gaussian Mixture Models (GMM) marked a significant advancement, as exemplified in the research by (Barai, Chakraborty, Das, Basu, & Nasipuri, 2022;Kamiński & Dobrowolski, 2022;Sisman, Yamagishi, King, & Li, 2020), which provided a robust method for modeling voice characteristics.…”