Gaussian distributions are usually parameterized with their natural parameters: the mean μ and the covariance Σ. They can also be re-parameterized as exponential models with canonical parameters P = Σ and ψ = Pμ. In this paper we consider modeling acoustics with mixtures of Gaussians parameterized with canonical parameters where the parameters are constrained to lie in a shared affine subspace....

Information retrieval researchers have studied passage retrieval extensively, yet there is no consensus within the community about how to evaluate the results of passage retrieval experiments. This paper describes five character-level passage evaluation measures and tasks for which they may be appropriate. In the second half of the paper we compare several passage retrieval models, including a ...

This paper addresses the problem of real-time speaker change detection in TV news broadcast, in which no prior knowledge on speakers is assumed. To remove the unreliable frames and background frames in the speech stream, we propose a new approach for feature categorization based on Gaussian Mixture Model Universal Background Model (GMM-UBM). The feature vectors are categorized into three sets, ...

In this paper, the mixture of support vector machines is proposed and applied to text-independent speaker recognition. The mixture of experts is used and is implemented by the divide-and-conquer approach. The purpose of adopting this idea is to deal with the large scale speech data and improve the performance of speaker recognition. The principle is to train several parallel SVMs on the subsets...

Traditional approaches such as Gaussian mixture model (GMM), Otsu’s and moment preserving (MP) methods are developed for segmentation of opaque objects. For semi-opaque objects like flame and smoke the result is cluttered, due to inappropriate threshold, especially if one dominates the other. Besides, rapidly changing environments like foggy and rainy scenes increase the difficulty in foregroun...

This paper proposes a new approach to multi-object tracking by semantic topic discovery. We dynamically cluster frame-by-frame detections and treat objects as topics, allowing the application of the Dirichlet Process Mixture Model (DPMM). The tracking problem is cast as a topic-discovery task where the video sequence is treated analogously to a document. This formulation addresses tracking issu...

This paper presents a flexible topic model based on the nested Indian buffet process (nIBP). The flexibility is achieved by relaxing three constraints: (1) number of topics is fixed, (2) topics are independent, and (3) topic hierarchy for a document is limited by a single tree path. Bayesian nonparametric learning is conducted to build a tree model where the number of topics and the topic hiera...

In this article we introduce how to put vague hyperprior on Dirichlet distribution, and we update the parameter of it by adaptive rejection sampling (ARS). Finally we analyze this hyperprior in an over-fitted mixture model by some synthetic experiments.

We report on the progress with respect to an emotion-aware voice portal concerning several directions. A comprehensive new data collection has been carried out and gives new insight on the nature of real life data. The labeling process and the structure of the data will be discussed. Experiments with the anger detector on that data indicate that the acoustic features based on voicing don’t play...

We present a method for discovering patterns of activation observed through fMIRI in experiments with multiple stimuli/tasks. We introduce an explicit parameterization for the profiles of activation and represent fMRI time courses as such profiles using linear regression estimates. Working in the space of activation profiles, we design a mixture model that finds the major activation patterns al...