Phone classification via manifold learning based dimensionality reduction algorithms

Huang, Heyun; Bosch, L.F.M. ten; Cranen, B.; Boves, L.W.J.

doi:10.1016/j.specom.2015.10.005

Cited by 3 publications

(2 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are quite a number of studies on vowel classification in the literature. Most of them are based on frequency domain analysis using features such as formant frequencies [2,3], linear predictive coding coefficients (LPCC), perceptual linear prediction (PLP) coefficients [4], mel frequency cepstral coefficients (MFCC) [5,6,7,8], wavelets [9], spectro-temporal features [10], and spectral decomposition [11]. However, there are fewer studies using time domain analysis [12,13].…”

Section: Introductionmentioning

confidence: 99%

Vowel Classification Based on Waveform Shapes

Tora¹,

Karacor²,

Uslu³

2019

Adv. sci. technol. eng. syst. j.

View full text Add to dashboard Cite

Vowel classification is an essential part of speech recognition. In classical studies, this problem is mostly handled by using spectral domain features. In this study, a novel approach is proposed for vowel classification based on the visual features of speech waveforms. In sound vocalizing, the position of certain organs of the human vocal system such as tongue, lips and jaw is very effective on the waveform shapes of the produced sound. The motivation to employ visual features instead of classical frequency domain features is its potential usage in specific applications like language education. Even though this study is confined to Turkish vowels, the developed method can be applied to other languages as well since the shapes of the vowels show similar patterns. Turkish vowels are grouped into five categories. For each vowel group, a time domain speech waveform with an interval of two pitch periods is handled as an image. A series of morphological operations is performed on this speech waveform image to obtain the geometric characteristics representing the shape of each class. The extracted visual features are then fed into three different classifiers. The classification performances of these features are compared with classical methods. It is observed that the proposed visual features achieve promising classification rates.

show abstract

Section: Introductionmentioning

confidence: 99%

Vowel Classification Based on Waveform Shapes

Tora¹,

Karacor²,

Uslu³

2019

Adv. sci. technol. eng. syst. j.

View full text Add to dashboard Cite

show abstract

“…The computations used in the demonstration are reviewed in Appendix A (hence, readers unfamiliar with graph-based methods may want to read material therein before proceeding with Section 2). In general, graph-based manifold methods have been usefully applied across a range of speech science modeling problems including phonetic category learning (e.g., Jansen & Niyogi, 2007; Plummer et al, 2010; Plummer, 2014) and automatic speech recognition (e.g., Errity & McKenna, 2006; Jafari & Almasganj, 2010; Zhao & Zhang, 2012; Norouzian et al, 2013; Tomar & Rose, 2014; Huang et al, 2016b,a). Section 3 discusses broader motivation for the manifold assumption as well as how our framework is situated with respect to recent advances in representation learning within the machine learning community.…”

Section: Introductionmentioning

confidence: 99%

Computing low-dimensional representations of speech from socio-auditory structures for phonetic analyses

Plummer

Reidy

2018

Journal of Phonetics

View full text Add to dashboard Cite

Low-dimensional representations of speech data, such as formant values extracted by linear predictive coding analysis or spectral moments computed from whole spectra viewed as probability distributions, have been instrumental in both phonetic and phonological analyses over the last few decades. In this paper, we present a framework for computing low-dimensional representations of speech data based on two assumptions: that speech data represented in high-dimensional data spaces lie on shapes called manifolds that can be used to map speech data to low-dimensional coordinate spaces, and that manifolds underlying speech data are generated from a combination of language-specific lexical, phonological, and phonetic information as well as culture-specific socio-indexical information that is expressed by talkers of a given speech community. We demonstrate the basic mechanics of the framework by carrying out an analysis of children’s productions of sibilant fricatives relative to those of adults in their speech community using the phoneigen package – a publicly available implementation of the framework. We focus the demonstration on enumerating the steps for constructing manifolds from data and then using them to map the data to a low-dimensional space, explicating how manifold structure affects the learned low-dimensional representations, and comparing the use of these representations against standard acoustic features in a phonetic analysis. We conclude with a discussion of the framework’s underlying assumptions, its broader modeling potential, and its position relative to recent advances in the field of representation learning.

show abstract

The Dynamics of Deforming Manifold: A Mathematical Model

Zhuang

Mastorakis

2022

International Journal of Mathematical Models and Methods in Applied Sciences

View full text Add to dashboard Cite

In order to meet the requirements of practical applications, a model of deforming manifold in the embedded space is proposed. The deforming vector and deforming field are presented to precisely describe the deforming process, which have clear physical meanings. The proposed model is a modification of the general differential dynamic model, with constraints of spatial and temporal continuity on the deforming field. The deformation integral and derivative are presented as compact expressions of manifold deforming process. Moreover, a specific autonomous deforming field with flattening effect is defined, which provides a novel geometric viewpoint on data dimension reduction. The effectiveness of this autonomous deforming field is proved by numerical computation simulations, which indicate the promising potential of the proposed model in practical dimension reduction tasks.

show abstract

Phone classification via manifold learning based dimensionality reduction algorithms

Cited by 3 publications

References 45 publications

Vowel Classification Based on Waveform Shapes

Vowel Classification Based on Waveform Shapes

Computing low-dimensional representations of speech from socio-auditory structures for phonetic analyses

The Dynamics of Deforming Manifold: A Mathematical Model

Contact Info

Product

Resources

About