This paper proposes a new mapping method combining GMM and codebook mapping methods to transform spectral envelope for voice conversion system. After analyzing overly smoothing problem of GMM mapping method in detail, we propose to convert the basic spectral envelope by GMM method and convert envelope-subtracted spectral details by GMM and phone-tied codebook mapping method. Objective evaluatio...

This paper presents a new approach to modeling speech spectra and pitch for text-independent speaker identification using Gaussian mixture models based on multi-space probability distribution (MSD-GMM). The MSD-GMM allows us to model continuous pitch values for voiced frames and discrete symbols representing unvoiced frames in a unified framework. Spectral and pitch features are jointly modeled...

Impressive progress has been recently witnessed on deep unsupervised clustering and feature disentanglement. In this paper, we propose a novel method top of one recent architecture with explanation Gaussian mixture model (GMM) membership, accompanied by GMM loss to enhance the clustering. The is optimized explicitly computed parameters under our coupled inspired framework. Specifically, takes a...

The Gaussian mixture modeling (GMM) techniques are increasingly being used for both speaker identification and verification. Most of these models assume diagonal covariance matrices. Although empirically any distribution can be approximated with a diagonal GMM, a large number of mixture components are usually needed to obtain a good approximation. A consequence of using a large GMM is that its ...

The generalized method of moments (GMM) is a very popular estimation and inference procedure based on moment conditions. When likelihood-based methods are difficult to implement, one can often derive various moment conditions and construct the GMM objective function. However, minimization of the objective function in the GMM may be challenging, especially over a large parameter space. Due to th...

The recently proposed “generalized min-max” (GMM) kernel [9] can be efficiently linearized, with direct applications in large-scale statistical learning and fast near neighbor search. The linearized GMM kernel was extensively compared in [9] with linearized radial basis function (RBF) kernel. On a large number of classification tasks, the tuning-free GMM kernel performs (surprisingly) well comp...

This paper demonstrates that modelling session variability during GMM training can improve the performance of a GMM supervector SVM speaker verification system. Recently, a method of modelling session variability in GMM-UBM systems has led to significant improvements when the training and testing conditions are subject to session effects. In this work, session variability modelling is applied d...

We describe a GMM-UBM-based evaluation system for pathologic voices that uses standard cepstral features. Per speaker one GMM is created and its components are used to create a so-called GMM supervector. The supervector of each speaker is labeled with the intelligibility values obtained by human evaluation and is used to train an SVR. We studied different GMM supervectors containing different G...