Voice conversion, as considered in this paper, is defined as modifying the speech signal of one speaker (source speaker) so that it sounds as if it had been pronounced by a different speaker (target speaker). Our contribution includes the design of a new methodology for representing the relationship between two sets of spectral envelopes. The proposed method is based on the use of a Gaussian mixture model of the source speaker spectral envelopes. The conversion itself is represented by a continuous parametric function which takes into account the probabilistic classification provided by the mixture model. The parameters of the conversion function are estimated by least squares optimization on the training data. This conversion method is implemented in the context of the HNM (harmonic+noise model) system, which allows high-quality modifications of speech signals. Compared to earlier methods based on vector quantization, the proposed conversion scheme results in a much better match between the converted envelopes and the target envelopes. Evaluation by objective tests and formal listening tests shows that the proposed transform greatly improves the quality and naturalness of the converted speech signals compared with previous proposed conversion methods.

Continuous probabilistic transform for voice conversion

In this paper, we describe a novel spectral conversion method for voice conversion (VC). A Gaussian mixture model (GMM) of the joint probability density of source and target features is employed for performing spectral conversion between speakers. The conventional method converts spectral parameters frame by frame based on the minimum mean square error. Although it is reasonably effective, the deterioration of speech quality is caused by some problems: 1) appropriate spectral movements are not always caused by the frame-based conversion process, and 2) the converted spectra are excessively smoothed by statistical modeling. In order to address those problems, we propose a conversion method based on the maximum-likelihood estimation of a spectral parameter trajectory. Not only static but also dynamic feature statistics are used for realizing the appropriate converted spectrum sequence. Moreover, the oversmoothing effect is alleviated by considering a global variance feature of the converted spectra. Experimental results indicate that the performance of VC can be dramatically improved by the proposed method in view of both speech quality and conversion accuracy for speaker individuality.

/pdf/voice-conversion-based-on-maximum-likelihood-estimation-of-ysizn2pm2o.pdf

Voice Conversion Based on Maximum-Likelihood Estimation of Spectral Parameter Trajectory

INTERSPEECH2008: 9th Annual Conference of the International Speech Communication Association, September 22-26, 2008, Brisbane, Australia.

Low-delay voice conversion based on maximum likelihood estimation of spectral parameter trajectory.

A new voice conversion algorithm that modifies a source speaker's speech to sound as if produced by a target speaker is presented. It is applied to a residual-excited LPC text-to-speech diphone synthesizer. Spectral parameters are mapped using a locally linear transformation based on Gaussian mixture models whose parameters are trained by joint density estimation. The LPC residuals are adjusted to match the target speakers average pitch. To study effects of the amount of training on performance, data sets of varying sizes are created by automatically selecting subsets of all available diphones by a vector quantization method. In an objective evaluation, the proposed method is found to perform more reliably for small training sets than a previous approach. In perceptual tests, it was shown that nearly optimal spectral conversion performance was achieved, even with a small amount of training data. However, speech quality improved with increases in the training set size.

Spectral voice conversion for text-to-speech synthesis

Non-parametric techniques for pitch-scale and time-scale modification of speech

In this contribution, a new system for voice conversion is described. The proposed architecture combines a PSOLA (Pitch Synchronous Overlap and Add)-derived synthesizer and a module for spectral transformation. The synthesizer based on the classical source-filter decomposition allows prosodic and spectral transformations to be performed independently. Prosodic modifications are applied on the excitation signal using the TD-PSOLA scheme; converted speech is then synthesized using the transformed spectral parameters. Two different approaches to derive spectral transformations, borrowed from the speech-recognition domain, are compared: Linear Multivariate Regression (LMR) and Dynamic Frequency Warping (DFW). Vector-quantization is carried out as a preliminary stage to render the spectral transformations dependent of the acoustical realization of sounds. A formal listening test shows that the synthesizer produces a satisfyingly natural “transformed” voice. LMR proves yet to allow a slightly better conversion than DFW. Still there is room for improvement in the spectral transformation stage.

Voice transformation using PSOLA technique

Whereas speaker adaptation has received much attention for speech recognition few studies have been devoted to voice transformation for speech synthesis, despite the potential interests of such techniques. The authors propose a voice conversion system which combines the time-domain pitch synchronous overlap and add (TD-PSOLA) technique with a source-filter decomposition. The first technique allows prosodic modifications while the second enables spectral envelope transformations. Two approaches to learn spectral alteration are compared: the linear multivariate regression (LMR) and the dynamic frequency warping (DFW). >

The authors report results obtained in the framework of phonetic feature extraction using TDNNs (time-delay neural networks). Experiments were conducted on a set of 5270 phonemes manually segmented from natural continuous speech uttered by one male speaker. Nearly all scores on binary phonetic features range between 90% and 99%. Comments on the visualization of the weights in several networks are also presented. In particular, it is noted that there exist quantitative similarities between patterns of weights within the TDNN and acoustical characteristics of the phonetic feature. >

J. P. Tubach

Papers

Voice transformation using PSOLA technique

Voice transformation using PSOLA technique

TDNNs for phonetic features extraction: a visual exploration