Learning speaker normalization using semisupervised manifold alignment
نویسندگان
چکیده
As a child acquires language, he or she: perceives acoustic information in his or her surrounding environment; identifies portions of the ambient acoustic information as languagerelated; and associates that language-related information with his or her perception of his or her own language-related acoustic productions. The present work models the third task. We use a semisupervised alignment algorithm based on manifold learning. We discuss the concepts behind this approach, and the application of the algorithm to this task. We present experimental evidence indicating the usefulness of manifold alignment in learning speaker normalization.
منابع مشابه
Semisupervised alignment of manifolds
In this paper, we study a family of semisupervised learning algorithms for “aligning” different data sets that are characterized by the same underlying manifold. The optimizations of these algorithms are based on graphs that provide a discretized approximation to the manifold. Partial alignments of the data sets—obtained from prior knowledge of their manifold structure or from pairwise correspo...
متن کاملAlignments of Manifold Sections of Different Dimensions in Manifold Learning
We consider an alignment algorithm for reconstructing global coordinates from local coordinates constructed for sections of manifolds. We show that, under certain conditions, the alignment algorithm can successfully recover global coordinates even when local neighborhoods have different dimensions. Our results generalize an earlier analysis to allow alignment of sections of different dimensions...
متن کاملManifold Alignment without Correspondence
Manifold alignment has been found to be useful in many areas of machine learning and data mining. In this paper we introduce a novel manifold alignment approach, which differs from “semisupervised alignment” and “Procrustes alignment” in that it does not require predetermining correspondences. Our approach learns a projection that maps data instances (from two different spaces) to a lower dimen...
متن کاملIndirect and Conditional Sensing in the Event Calculus
Manifold alignment has been found to be useful in many areas of machine learning and data mining. In this paper we introduce a novel manifold alignment approach, which differs from “semisupervised alignment” and “Procrustes alignment” in that it does not require predetermining correspondences. Our approach learns a projection that maps data instances (from two different spaces) to a lower dimen...
متن کاملNon-Linear I-vector Extraction for Speaker Recognition
We propose an algorithm for non-linear i-vector extraction. The algorithm is based on the manifold learning technique named Diffusion Maps (DM) and motivated by recent results that showed that the GMM supervectors reside on a low dimensional manifold. Our proposed method may further be processed using standard techniques such as Linear Discriminant Analysis (LDA), Within Class Covariance Normal...
متن کامل