Data-driven voice source waveform analysis and synthesis

Abstract

The paper presents a voice source waveform modeling techniques based on principal component analysis (PCA) and Gaussian mixture modeling (GMM). The voice source is obtained by inverse-filteirng speech with the estimated vocal tract filter. This decomposition is useful in speech analysis, synthesis, recognition and coding. Existing models of the voice source signal are based on function-fitting or physically motivated assumptions and although they are well defined, estimation of their parameters is not well understood and few are capable of reproducing the large variety of voice source waveforms. Here, a data-driven approach is presented for signal decomposition and classification based on the principal components of the voice source. The principal components are analyzed and the 'prototype' voice source signals corresponding to the Gaussian mixture means are examined. We show how an unknown signal can be decomposed into its components and/or prototypes and resynthesized. We show how the techniques are suited for both low bitrate or high quality analysis/synthesis schemes.

Voice source characterization using pitch synchronous discrete cosine transform for speaker identification

“…This behaviour is similar to that of the principal components of the VS proposed in Ref. 16. The gross shape of the ILPR is captured by only 12 coefficients.…”

Breathy, Resonant, Pressed – Automatic Detection of Phonation Mode from Audio Recordings of Singing

“…This process is called inverse filtering [Fritzell 1992, Walker and Murphy 2007, Drugman et al 2012, Gudnason et al 2012. The vocal tract (throat, mouth and in some cases nose) forms the tube, which is characterized by its resonances.…”

Glottal source processing: From analysis to applications

“…A set of the first PCA components were then modelled with Gaussian Mixture Models (GMMs). The obtained GMMs were shown by (Gudnason et al (2012)) to enable parameterizing source features, such as non-flatness of the closed phase, that traditional approaches fail to model. Finally, a new parameterization method of the voice source was recently proposed by (Kane and Gobl (2013a)) by adopting dynamic programming using which the settings originating from a manual voice source analysis could be combined into an automatic, machine-based analysis.…”

“…Instead of fitting the computed glottal flow with a pre-defined function, a parameterization scheme based on a data-driven approach was recently proposed by (Gudnason et al (2012)). More specifically, a glottal flow estimate was first computed with IAIF and the obtained waveform was then processed with Principal Component Analysis (PCA) in order to achieve dimensionality reduction.…”