Minimum error rate training for designing tree-structured probability density function
نویسنده
چکیده
In this paper, we propose a signal prototype classi cation and evaluation framework in acoustic modeling. Based on this framework, a new tree-structured likelihood function is derived. It uses a designated cluster kernel f m for signal prototype classi cation and a designated cluster kernel f m for likelihood evaluation of outlier or tail events of the cluster. A minimum classi cation error (MCE) rate training approach is described for designing tree-structured likelihood function. Experimental results indicate that the new tree-structured likelihood function signi cantly improves the acoustic resolution of the model. It has a more signi cant speedup in decoding than the one obtained from the conventional approach.
منابع مشابه
Optimized Wavelet Packet decomposition based on Minimum Probability of Error Signal Representation
This work addresses the problem of optimal Wavelet packet (WP) filter bank decomposition based on the minimum probability of error signal representation (MPE-SR) principle. The problem is formulated as a complexity regularized optimization, where the tree-indexed structure of the WP family is explored to find conditions for reducing this problem to a type of minimum cost tree pruning, a method ...
متن کاملMinimum risk acoustic clustering for multilingual acoustic model combination
In this paper we describe procedures for combining multiple acoustic models, obtained using training corpora from different languages, in order to improve ASR performance in languages for which large amounts of training data are not available. We treat these models as multiple sources of information whose scores are combined in a log-linear model to compute the hypothesis likelihood. The model ...
متن کاملAdaptive nonlinear least bit error-rate detection for symmetrical RBF beamforming
A powerful symmetrical radial basis function (RBF) aided detector is proposed for nonlinear detection in so-called rank-deficient multiple-antenna assisted beamforming systems. By exploiting the inherent symmetry of the optimal Bayesian detection solution, the proposed RBF detector becomes capable of approaching the optimal Bayesian detection performance using channel-impaired training data. A ...
متن کاملAdaptive Minimum Bit-Error-Rate Filtering
Adaptive filtering has traditionally been developed based on the minimum mean square error (MMSE) principle and has found ever-increasing applications in communications. The paper develops adaptive filtering based on an alternative minimum bit error rate (MBER) criterion for communication applications. It is shown that the MBER filtering exploits the non-Gaussian distribution of filter output e...
متن کاملمطالعات درخت تصمیم در برآورد ریسک ابتلا به سرطان سینه با استفاده از چند شکلیهای تک نوکلوئیدی
Abstract Introduction: Decision tree is the data mining tools to collect, accurate prediction and sift information from massive amounts of data that are used widely in the field of computational biology and bioinformatics. In bioinformatics can be predict on diseases, including breast cancer. The use of genomic data including single nucleotide polymorphisms is a very important ...
متن کامل