On Word Boundary Detection in Digit-based Speaker Verification
نویسنده
چکیده
In an automatic speaker verification (ASV) system with prompted passwords, we use vocabulary-dependent hidden Markov models and rely on the ability to explicitly locate the corresponding words and their boundaries in the speech signal. In an experiment on 41 speakers in a Swedish telephone speech database, we compare the use of utterance segmentation produced by automatic and manual methods, and conclude that not much is lost in ASV performance with the automatic method compared to the manual.
منابع مشابه
A connected digit recognizer based on dynamic time warping and isolated digit templates
A connected digit recognizer is proposed in which a set of isolated word templates is used as reference patterns and an unconstrained dynamic time warping algorithm is used to literally "spot the digits in the string. Segmentation boundaries between digits are obtained as the termination point of the dynamic path from the previous time warp. A region around the boundary is searched for the opti...
متن کاملPerformance improvement of connected digit recognition using unsupervised fast speaker adaptation
In this paper, we investigate unsupervised fast speaker adaptation based on eigenvoice to improve the performance of Korean connected digit recognition over the telephone channel. In addition, utterance verification is introduced into speaker adaptation to examine whether input utterance is appropriate to adaptation or not. Performance evaluation showed that the proposed method yielded performa...
متن کاملDeep CNN based feature extractor for text-prompted speaker recognition
Deep learning is still not a very common tool in speaker verification field. We study deep convolutional neural network performance in the text-prompted speaker verification task. The prompted passphrase is segmented into word states — i.e. digits — to test each digit utterance separately. We train a single high-level feature extractor for all states and use cosine similarity metric for scoring...
متن کاملDetermination of threshold for speaker verification using speaker adaptation gain in likelihood during training
This paper describes methods to determine thresholds for speaker verification. Setting an appropriate threshold a priori is difficult because likelihood verification covers a wide range and the appropriate threshold for each speaker is different. We propose new methods to determine the speaker verification threshold depending on the "adaptation degree" for each speaker. We use the gain in likel...
متن کاملPrompting of passwords in speaker verification systems
The problem of how to prompt a client with a password in an automatic prompted speaker verification system is addressed. Text-prompting of four-digit sequences is compared to speech-prompting of the same sequences, and speech-prompting of fourdigit sequences is compared to speech-prompting of five-digit sequences. Speech recordings are analyzed by comparing speaker verification performance and ...
متن کامل