Back to search

University of Crete (UOC)

Τεχνικές αραιής και χαμηλής τάξης αναπαράστασης για εύρωστη αναγνώριση ομιλητή και ανακατασκευή ελλιπών χαρακτηριστικών

Abstract

dc:description

Speaker recognition is the process for recognizing a speaker automatically, based on specific features extracted from the speech signal. It is divided in two distinct categories, namely, speaker identification and speaker verification. A broad range of applications exploits at its core the process of speaker recognition, where usually the presence of environmental noise in the speech signal impedes the inference of correct decisions. An additional factor, which contributes to the difficulty of recognizing a speaker correctly, is the limited amount of available training and evaluation data. This can be due to either a practical difficulty in obtaining a large volume of training data, or to the need to reduce the overall computational cost by using limited, yet reliable, evaluation data. Focusing on overcoming the above limitations, while achieving high rates of successful recognition, this dissertation is divided in two main parts. In the first part, the problem of speaker recognition is reduced in an equivalent classification problem. To this end, we develop and study the performance of classification techniques, which are based on the framework of sparse representations, where we focus on the task of speaker identification by employing highly limited amounts of training and evaluation data, in environments with high levels of noise. The main assumption that governs these techniques is that the identified speech signal, and specifically the features that have been extracted from this signal, can be expressed as a sparse linear combination in terms of the columns of an overcomplete matrix, which is often referred in the literature with the term dictionary. This dictionary is constructed appropriately from the available training data, while the computation of the sparse linear combinations is achieved via the solution of an optimization problem based on lp-norms (p = 1 or 2). The optimally estimated sparse weights of the linear combinations, the so-called sparse codes, which are obtained as the solutions of the optimization problem, are then employed for the final identification of the speaker based on a minimum reconstruction error criterion. Extending our previous classification method based on sparse representations, we study the efficiency of a method for discriminative dictionary learning. This method estimates jointly the dictionary comprising of the training data in conjunction with an appropriate linear classifier. The advantage of this approach is that it results in sparse codes, which are characterized by enhanced discriminative capability. For this experimental evaluation of the performance of our proposed method, through extensive simulations, a relatively small-sized database was used. The corresponding data were corrupted by several distinct types of environmental noise, for a wide range of signal-to-noise ratio values. Extensive comparisons with probabilistic models, which are based on the hypothesis that the extracted speech features follow a generalized Gaussian distribution, as well as with some of the state-of-the-art classification methods, such as Gaussian mixture models and joint factor analysis, revealed the superiority of our proposed method in terms of achieving higher correct recognition rates in noisy environments combined with the use of short training and testing speech data. The second part of this dissertation focuses on the use of low-rank techniques as a powerful tool for extracting reliable features from a speech signal. More specifically, a technique for recovering a low-rank matrix is designed, which is employed for the reconstruction of those spectral regions of a speech signal, which are unreliable due to the presence of noise. The discrimination of the spectral regions is achieved by means of a reliability mask, which discriminates the regions characterized by the presence of noise from the regions which are dominated by the speech signal information. The completion of the empty spectral regions is performed based on the assumption that the logarithmic magnitude representation of a speech signal in the time-frequency domain, obtained via the short-time Fourier transform (STFT), is of low rank. Then, the Singular Value Thresholding (SVT) algorithm is exploited for the completion of those regions of the STFT representation, which are considered to be unreliable. The experimental evaluation of the proposed method reveals its power in extracting reliable features, which yield high rates of correct speaker identification in cases of high noise levels. The comparison against the widely used method of sparse imputation, which is based on sparse representations, reveals the superiority of our proposed approach in terms of achieving accurate speaker identification, especially for low levels of signal-to-noise ratios. The above method does not take into account the existing prior knowledge with respect to the available training data, constituting essentially an unsupervised method. Motivated by this observation, we propose an extension of the matrix completion method, which exploits the prior knowledge that the data matrix is low rank, as well as the knowledge that the data can be represented efficiently in terms of a dictionary. In particular, we proposed an algorithm for joint low-rank representation and matrix completion (J-SVT). J-SVT is superior when compared with the standard SVT with respect to the computation of the low-rank representation of a data matrix in terms of a given dictionary, by employing a small number of observations from the original matrix. Through extensive simulations, we observed an improvement of the reconstruction error achieved by the J-SVT, in contrast to the typical SVT, for several distinct experimental scenarios.

Degree

thesis:*
Grantor dc:publisher
University of Crete (UOC)
Year dc:date
2014

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Tzagkarakis, Christos
  • Τζαγκαράκης, Χρήστος

Subjects

dc:subject × 12

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/34770

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Tzagkarakis, Christos; Τζαγκαράκης, Χρήστος. Τεχνικές αραιής και χαμηλής τάξης αναπαράστασης για εύρωστη αναγνώριση ομιλητή και ανακατασκευή ελλιπών χαρακτηριστικών. University of Crete (UOC), 2014. http://hdl.handle.net/10442/hedi/34770