Discriminative Training Algorithm Research Articles

Hidden Markov models-based 3D MRI brain segmentation

This paper introduces a 3D MRI segmentation algorithm based on Hidden Markov Models (HMMs). The mathematical models for the HMM that forms the basis of the segmentation algorithm for both the continuous and discrete cases are developed and contrasted with Hidden Markov Random Field in terms of complexity and extensibility to larger fields. The presented algorithm clearly demonstrates the capacity of HMM to tackle multi-dimensional classification problems. The HMM-based segmentation algorithm was evaluated through application to simulated brain images from the McConnell Brain Imaging Centre, Montréal Neurological Institute, McGill University as well as real brain images from the Internet Brain Segmentation Repository (IBSR), Harvard University. The HMM model exhibited high accuracy in segmenting the simulated brain data and an even higher accuracy when compared to other techniques applied to the IBSR 3D MRI data sets. The achieved accuracy of the segmentation results is attributed to the HMM foundation and the utilization of the 3D model of the data. The IBSR 3D MRI data sets encompass various levels of difficulty and artifacts that were chosen to pose a wide range of challenges, which required handling of sudden intensity variations and the need for global intensity level correction and 3D anisotropic filtering. During segmentation, each class of MR tissue was assigned to a separate HMM and all of the models were trained using the discriminative MCE training algorithm. The results were numerically assessed and compared to those reported using other techniques applied to the same data sets, including manual segmentations establishing the ground truth for real MR brain data. The results obtained using the HMM-based algorithm were the closest to the manual segmentation ground truth in terms of an objective measure of overlap compared to other methods.

Statistical Classification of Abnormal Blood Profiles in Athletes

Blood doping has been challenging the scientific community since the early 1970's, where it was demonstrated that blood transfusion significantly improves physical performance. Here, we present through 3 applications how statistical classification techniques can assist the implementation of effective tests to deter blood doping in elite sports. In particular, we developed a new indirect and universal test of blood doping, called Abnormal Blood Profile Score (ABPS), based on the statistical classification of indirect biomarkers of altered erythropoiesis. Up to 601 hematological profiles have been compiled in a reference database. Twenty-one of them were obtained from blood samples withdrawn from professional athletes convicted of blood doping by other direct tests. Discriminative training algorithms were used jointly with cross-validation techniques to map these labeled reference profiles to target outputs. The strict cross-validation procedure facilitates the adherence to medico-legal standards mandated by the World Anti Doping Agency (WADA). The test has a sensitivity to recombinant erythropoietin (rhEPO) abuse up to 3 times better than current generative models, independently whether the athlete is currently taking rhEPO or has stopped the treatment. The test is also sensitive to any form of blood transfusion, autologous transfusion included. We finally conclude why a probabilistic approach should be encouraged for the evaluation of evidence in anti-doping area of investigation.

Lattice segmentation and minimum Bayes risk discriminative training for large vocabulary continuous speech recognition

Lattice segmentation techniques developed for Minimum Bayes Risk decoding in large vocabulary speech recognition tasks are used to compute the statistics needed for discriminative training algorithms that estimate HMM parameters so as to reduce the overall risk over the training data. New estimation procedures are developed and evaluated for both small and large vocabulary recognition tasks, and additive performance improvements are shown relative to maximum mutual information estimation. These relative gains are explained through a detailed analysis of individual word recognition errors.

A Discriminative Training Algorithm for Hidden Markov Models

We introduce a discriminative training algorithm for the estimation of hidden Markov model (HMM) parameters. This algorithm is based on an approximation of the maximum mutual information (MMI) objective function and its maximization in a technique similar to the expectation-maximization (EM) algorithm. The algorithm is implemented by a simple modification of the standard Baum-Welch algorithm, and can be applied to speech recognition as well as to word-spotting systems. Three tasks were tested: isolated digit recognition in a noisy environment, connected digit recognition in a noisy environment and word-spotting. In all tasks a significant improvement over maximum likelihood (ML) estimation was observed. We also compared the new algorithm to the commonly used extended Baum-Welch MMI algorithm. In our tests the algorithm showed advantages in terms of both performance and computational complexity.

A discriminative training algorithm for VQ-based speaker identification

A novel method, referred to as group vector quantization (GVQ), is proposed to train VQ codebooks for closed-set speaker identification. In GVQ training, speaker codebooks are optimized for vector groups rather than for individual vectors. An evaluation experiment has been conducted to compare the codebooks trained by the Linde-Buzo-Grey (LBG), the learning vector quantization (LVQ), and the GVQ algorithms. It is shown that the frame scores from the GVQ trained codebooks are less correlated, therefore, the sentence level speaker identification rate increases more quickly with the length of test sentences.

Context-dependent word duration modelling for Korean connected digit recognition

In a Korean connected digit recogniser, duration modelling is necessary to reduce insertion and deletion errors due to monophonemic digits, which cannot usually be corrected even by discriminative training algorithms. In the Letter the authors incorporate context-dependent word duration modelling directly in a decoding algorithm to reduce those errors. While incorporating duration information in the postprocessing stage shows little improvements over a baseline system, the proposed method reduces word error rates by ~10% for unknown length decoding when both maximum likelihood estimation and generalised probabilistic descent training are used.

A minimum-distortion segmentation/LVQ hybrid algorithm for speech recognition.

The quick and simple training of learning vector quantization (LVQ) can produce classi ficationpower at least as high as that by a powerful, but complex classifier based on artificial neural networks. However, LVQ is a discriminative training algorithm for a distance classifier handling static (fixed-dimensional) patterns. Thus, an innovative process is required to apply this algorithm to dynamic (variable-durational) speech patterns. To meet this requirement, an HHM/LVQ hybrid algorithm was proposed which integrated HMM (Viterbi) segmentation with LVQ classification. However, this algorithm, using all the possible HMM models for segmentation, produces an enormous number of training tokens, making it difficult to apply to large-scale continuous speech recognition tasks. In this light, we present a new minimum-distortion segmentation (MDS)/discriminative classification hybrid algorithm. The MDS algorithm produces one segmentation and this is used in place of the many HMM segmentations. To make a proper comparison between the two methods we used as our discriminative classifier the same LVQ formulation. For clarity, we refer to this proposed algorithm as an MDS/LVQ hybrid algorithm. Results on the E-set task show that MDS/LVQ, with its significantly reduced training, can achieve discriminative power at least as high as HMM/LVQ.