An Evaluation of DTW, AA and ARVM for Fixed-Text Speaker Identification
نویسنده
چکیده
Three different methodologies for automatic speaker identification have been evaluated in the paper, namely the well known Dynamic Time Warping (DTW), the Auto-Regressive Vector Models (ARVM) and an Algebraic Approach (AA). The aim of our study is to examine the effectiveness of these approaches in the fixed-text speaker identification task with short phrases in Bulgarian language collected over noisy telephone channels. Furthermore, two well-known speech features, namely the Linear Predictive Coding derived Cepstrum (LPCC) and the Mel-Frequency Cepstral Coefficients (MFCC) were evaluated. As experimental results shown the joint work of the ARVM and the MFCC outperforms the all others approaches used in this study.
منابع مشابه
Field Evaluation of Text-Dependent Speaker Recognition in an Access Control Application
Vector quantization (VQ) is a widely used matching algorithm for text-independent speaker recognition. In this paper we study the use of text-dependent speaker recognition in practical access control application. We compared dynamic time warping (DTW) to VQ-based matching using textdependent pass phrases. Our goal was to find out, how fixed phrase affects speaker recognition performance. We col...
متن کاملGMM and ARVM cooperation and competition for text-independent speaker recognition on telephone speech
We develop a cooperation and a competition of two different natures modelizations. The first one, the GMM [1], is a modelization of the parametrisation distribution of the speaker speech. The second, the ARVM [2, 3], is a modelization of the speaker speech spectral evolution. To allow cooperation and competition between different modelizations we use a classical measure normalization. We invest...
متن کاملA DTW-based DAG technique for speech and speaker feature analysis
A DTW-based directed acyclic graph (DAG) optimization method is proposed to exploit the interaction information of speech and speaker in feature component. We introduce the DAG representation of intra-class samples based on dynamic time warping (DTW) measure and propose two criteria based on in-degree of DAG. Combined with (l − r) optimization algorithm, the DTW-based DAG model is applied to di...
متن کاملComparative study of GMM, DTW, and ANN on Thai speaker identification system
This paper proposes a new investigation on Gaussian mixture model (GMM) by comparing it with some preliminary experiments on multilayered perceptron network (MLP) with backpropagation learning algorithm (BKP) and dynamic time warping (DTW) techniques on Thai text-dependent speaker identification system. Three major identification engines are conducted on 50 speakers with isolated digits 0-9. Tr...
متن کاملOn the use of nearest feature line for speaker identification
As a new pattern classification method, Nearest Feature Line (NFL) provides an effective way to tackle the sort of pattern recognition problems where only limited data are available for training. In this paper, we explore the use of NFL for speaker identification in terms of limited data and examine how the NFL performs in such a vexing problem of various mismatches between training and test. I...
متن کامل