Score distribution scaling for speaker recognition
نویسندگان
چکیده
In this study, we transform the verification scores of a speaker recognition system in order to standardize the imposter score distribution, this facilitates setting of a speaker-independent threshold at desired False Alarm (FA) rates. Impostor score distributions are estimated using GMMs, and a univariate Gaussianization [1] transform (which is a monotonically increasing mapping) is applied on the scores. It is shown that if a monotonically increasing mapping is used, the Probability of correct detection for a given setting of the FA is maintained as before. Hence, the proposed technique performs distribution scaling without affecting the False Alarm to False Reject relationship of the original test statistic. The maximum (relative) mismatch between the obtained and desired False Alarm rates is less than 10% for a wide range of False Alarm rates. When compared to modeling the imposter score distributions using a single Gaussian (Z-norm case), the overall relative mismatch is reduced by an average of 30%. While the application focus is on speaker recognition, the proposed technique can be used for other binary speech classification tasks as well.
منابع مشابه
Text-independent speaker recognition using graph matching
Technical mismatches between the training and matching conditions adversely affect the performance of a speaker recognition system. In this paper, we present a matching scheme which is invariant to feature rotation, translation and uniform scaling. The proposed approach uses a neighborhood graph to represent the global shape of the feature distribution. The reference and test graphs are aligned...
متن کاملThe distribution of calibrated likelihood-ratios in speaker recognition
This paper studies properties of the score distributions of calibrated log-likelihood-ratios that are used in automatic speaker recognition. We derive the essential condition for calibration that the log likelihood ratio of the log-likelihood-ratio is the log-likelihood-ratio. We then investigate what the consequence of this condition is to the probability density functions (PDFs) of the loglik...
متن کاملA Review of Various Score Normalization Techniques for Speaker Identification System
This paper presents an overview of a state-of-the-art text-independent speaker verification system using score normalization. First, an introduction proposes a modular scheme of the training and test phases of a speaker verification system. Then, the most commonly speech parameterization used in speaker verification, namely, cepstral analysis, is detailed. Normalization of scores is then explai...
متن کاملSpeaker Independent Speech Recognition Using Hidden Markov Models for Persian Isolated Words
متن کامل
Speaker Adaptation in Continuous Speech Recognition Using MLLR-Based MAP Estimation
A variety of methods are used for speaker adaptation in speech recognition. In some techniques, such as MAP estimation, only the models with available training data are updated. Hence, large amounts of training data are required in order to have significant recognition improvements. In some others, such as MLLR, where several general transformations are applied to model clusters, the results ar...
متن کامل