Gaussian mixture modelling is a powerful tool for data analysis. However, the selection of number of Gaussians in the mixture, i.e., the mixture model or scale selection, remains a difficult problem. In this paper, we propose a new kind of dynamic merge-or-split learning (DMOSL) algorithm on Gaussian mixture such that the number of Gaussians can be determined automatically with a dynamic merge-...

Most techniques for speaker verification today use Gaussian Mixture Models (GMMs) and make the decision by comparing the likelihood of the speaker model to the likelihood of a universal background model (UBM). The paper proposes to replace the UBM by an individual background model (IBM) that is generated for each speaker. The IBM is created using the K-nearest cohort models and the UBM by a sim...

This paper deals with the tracking of speech segments in audio documents. We use a cepstral-based acoustic analysis and gaussian mixture models for the representation of the training data. Three ways of scoring an audio document based on a frame-level likelihood calculation are proposed and compared. Our experiments are done on a database composed of television programs including news reports, ...

This paper presents a 3-stage adaptation framework based on speaker selection training. First a subset of cohort speakers is selected for test speaker using Gaussian mixture model, which is more reliable given very limited adaptation data. Then cohort models are linearly transformed closer to each test speaker. Finally the adapted model for the test speaker is obtained by combining these transf...

In this paper a solution to detect wrong way drivers on highways is presented. The proposed solution is based on three main stages: Learning, Detection and Validation. Firstly, the orientation pattern of vehicles motion flow is learned and modelled by a mixture of gaussians. The second stage (Detection and Temporal Validation) applies the learned orientation model in order to detect objects mov...

Because of the limitations of matrix factorization, such as losing spatial structure information, the concept of low-rank tensor factorization (LRTF) has been applied for the recovery of a low dimensional subspace from high dimensional visual data. The low-rank tensor recovery is generally achieved by minimizing the loss function between the observed data and the factorization representation. T...

This paper describes a forensic application scenario in the speaker recognition domain and explains the techniques required to develop a complete forensic analysis system based on speaker identification. It is presented a standard manner of how Gaussian Mixture Models (GMMs) could be used in order to compute and express identification results in a judicial system. A review of possible technique...

This paper presents a voice conversion technique using Deep Belief Nets (DBNs) to build high-order eigen spaces of the source/target speakers, where it is easier to convert the source speech to the target speech than in the traditional cepstrum space. DBNs have a deep architecture that automatically discovers abstractions to maximally express the original input features. If we train the DBNs us...

In this paper, three speaker verification procedures are tested. All the procedures are based on Gaussian mixture models (GMM), however, they differ in the way, in which they use particular feature vectors of an utterance for speaker verification. A lot of experiments have been performed in a group of 329 speakers. The results showed that there is a procedure that enables to achieve better resu...

In certain cases, images of different scenes can be mixed to produce an image of a novel scene. For example, an image of a pink sphere can be additively mixed from suitable images of a red and a white sphere. Three ways in which scenes can differ are considered: in the spectral composition of the illuminant and in the spectral and the geometric reflectance of scene objects. Sufficient condition...