Multiple kernel learning for speaker verification

Longworth, Chris; Gales, Mark J. F.

doi:10.1109/icassp.2008.4517926

Cited by 15 publications

(8 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2, middle panel). Longworth [18] combines various dynamic kernels, including derivative kernel and parametric kernels, using MKL for speaker verification. In recent image recognition research, however, the MKL approach begun to be used for the purpose of feature vector selection or weighting.…”

Section: Feature Selection and Classification Using Mkl-svmmentioning

confidence: 99%

Acoustic feature selection utilizing multiple kernel learning for classification of children with autism spectrum and typically developing children

Kakihara

Takiguchi

Ariki

et al. 2013

Proceedings of the 2013 IEEE/SICE International Symposium on System Integration

View full text Add to dashboard Cite

This paper reports the result of a classification experiment carried out using acoustic features for children with autism spectrum, where a new featureweighting method using a multiple kernel learning (MKL) algorithm is proposed for classification between children with autism spectrum and typically developing children. Our MKL-SVM simultaneously estimates both the classification boundary and weight of each acoustic feature, where 484 acoustic features are used in our experiments. The estimated weight indicates how acoustic features are useful for classification. Our results show the large weight acoustic features mainly for line spectral frequencies in the classification experiment using acoustic features for children with autism spectrum.

show abstract

Section: Feature Selection and Classification Using Mkl-svmmentioning

confidence: 99%

Acoustic feature selection utilizing multiple kernel learning for classification of children with autism spectrum and typically developing children

Kakihara

Takiguchi

Ariki

et al. 2013

Proceedings of the 2013 IEEE/SICE International Symposium on System Integration

View full text Add to dashboard Cite

show abstract

“…where 0 < α (s t ) ≤ 1 denotes the learning step size and ∂û (s t )/∂σ (s t ) can be calculated from (32). Under the framework of MKL, the MMLC algorithm is proposed as an improved approach of RL in two aspects.…”

Section: The Mmlc Algorithmmentioning

confidence: 99%

“…The existing MKL algorithms can be divided into five major groups [26]: fixed rules, heuristic approaches, optimization approaches, Bayesian approaches and boosting approaches. In recent literature, MKL has been widely applied in many fields, such as visual object recognition [30], visual search [31], speaker verification [32], dimensionality reduction [33], structured prediction [34] and braincomputer interfacing [35].…”

Section: Introductionmentioning

confidence: 99%

Model-Free Multi-Kernel Learning Control for Nonlinear Discrete-Time Systems

Liu¹,

Xu²,

Huang³

et al. 2017

International Journal of Robotics and Automation

View full text Add to dashboard Cite

Reinforcement learning (RL) has become an important research topic to solve learning control problems of nonlinear dynamic systems. In RL, feature representation is a critical factor for improving the performance of online or offline learning controllers. Although multikernel learning has been studied in supervised learning problems, there is little work on multi-kernel-based feature representation in RL algorithms. In this paper, a model-free multi-kernel learning control (MMLC) approach is proposed for a class of nonlinear discrete-time systems. MMLC has advantages over other singlekernel-based RL algorithms in that the parameters in the kernel functions can be learned adaptively. Furthermore, MMLC uses a model-free actor-critic learning structure, where the critic is designed to approximate the derivatives of value functions. Different from the popularly studied dual heuristical programming algorithm, the proposed MMLC approach can learn the dynamics of the nonlinear system in a data-driven way. To evaluate the performance of MMLC, single-link and double-link inverted pendulums are employed as two benchmarks. The effectiveness of the MMLC algorithm has been demonstrated in simulation. It is shown that MMLC can achieve better performance than previous kernel-based dual heuristic programming with partial model information.

show abstract

“…Clearly, if the chosen kernel is a linear transformation of concatenation x nÀ1 ¼ fx i nÀ1 ; x iÀ1 nÀ1 g, then the solution is equivalent to using the common KT, which is a particular case of the proposed machine. The so-called summation kernel (SK) has been extensively used in the literature as a straightforward way of fusing heterogeneous information [9,8].…”

Section: Article In Pressmentioning

confidence: 99%

“…Differences with previous composite kernels: Composite kernels have been paid attention in the kernel methods literature [6,15,21,8]. Note, however, that our emphasis is on the difference between applying the KT and developing specific signal-based kernels.…”

Section: Article In Pressmentioning

confidence: 99%

Learning non-linear time-scales with kernel -filters

et al. 2009

View full text Add to dashboard Cite

A family of kernel methods, based on the g-filter structure, is presented for non-linear system identification and time series prediction. The kernel trick allows us to develop the natural non-linear extension of the (linear) support vector machine (SVM) g-filter [G. Camps-Valls, M. Martínez-Ramó n, J.L. Rojo-Á lvarez, E. Soria-Olivas, Robust g-filter using support vector machines, Neurocomput. J. 62(12) (2004) 493-499.], but this approach yields a rigid system model without non-linear cross relation between time-scales. Several functional analysis properties allow us to develop a full, principled family of kernel g-filters. The improved performance in several application examples suggests that a more appropriate representation of signal states is achieved.

show abstract

Multiple kernel learning for speaker verification

Cited by 15 publications

References 8 publications

Acoustic feature selection utilizing multiple kernel learning for classification of children with autism spectrum and typically developing children

Acoustic feature selection utilizing multiple kernel learning for classification of children with autism spectrum and typically developing children

Model-Free Multi-Kernel Learning Control for Nonlinear Discrete-Time Systems

Learning non-linear time-scales with kernel -filters

Contact Info

Product

Resources

About