Stochastic Speaker Recognition Model
Július Zimmermann and Július Zimmermann, Jr.
The task of speech processing is being spread over last years with speaker recognition, which consists of speaker identification and speaker verification. In the paper there is an explained method which uses the stochastic model to solve this task. The experiment has been done using hidden Markov models by means of HTK tool software. Testing speech materials consist of 200 speakers from corpus SpeechDat-Sk. Applied experiments reached the efficiency between 92,86% and 100%.
1. Introduction Human voice timbre and quality belong to unique speaker qualities. Already in the past they have served for distinguishing different people. Acoustic feature differences of speech signal of an individual are results of an anatomic difference combination in the articulatory tract and learnt individual speaking habits. In present rapid development and during putting automated systems into existence they may be applied also in the devices for speaker recognition, for instance authorised check in, banking via telephone, and in the telephone credit cards. Speaker recognition involves verification and identification. At verification the voiceprint is compared with the speaker voice model registered in the database that we want to verificate. The result of comparison is a measure of the similarity (score) from which rejection or acceptance of the verified speaker follows. At identification the voiceprint is compared with model voices of all speakers in the database. The comparison results are measures of the similarity from which the maximal quality is chosen. Following factors influence errors at speaker recognition: an incorrect reading or incorrect expressing of the required phrase, extreme emotional state, the changed microphone position within individual recognition or between individual recognitions, a different microphone at the speaker enrolment into the database and at the recognition, room... [continues]
Július Zimmermann and Július Zimmermann, Jr.
The task of speech processing is being spread over last years with speaker recognition, which consists of speaker identification and speaker verification. In the paper there is an explained method which uses the stochastic model to solve this task. The experiment has been done using hidden Markov models by means of HTK tool software. Testing speech materials consist of 200 speakers from corpus SpeechDat-Sk. Applied experiments reached the efficiency between 92,86% and 100%.
1. Introduction Human voice timbre and quality belong to unique speaker qualities. Already in the past they have served for distinguishing different people. Acoustic feature differences of speech signal of an individual are results of an anatomic difference combination in the articulatory tract and learnt individual speaking habits. In present rapid development and during putting automated systems into existence they may be applied also in the devices for speaker recognition, for instance authorised check in, banking via telephone, and in the telephone credit cards. Speaker recognition involves verification and identification. At verification the voiceprint is compared with the speaker voice model registered in the database that we want to verificate. The result of comparison is a measure of the similarity (score) from which rejection or acceptance of the verified speaker follows. At identification the voiceprint is compared with model voices of all speakers in the database. The comparison results are measures of the similarity from which the maximal quality is chosen. Following factors influence errors at speaker recognition: an incorrect reading or incorrect expressing of the required phrase, extreme emotional state, the changed microphone position within individual recognition or between individual recognitions, a different microphone at the speaker enrolment into the database and at the recognition, room... [continues]
Cite This Essay
- APA
-
(2011, 04). Speaker Identification. StudyMode.com. Retrieved 04, 2011, from http://www.studymode.com/essays/Speaker-Identification-663931.html
- MLA
-
"Speaker Identification" StudyMode.com. 04 2011. 04 2011 <http://www.studymode.com/essays/Speaker-Identification-663931.html>.
- CHICAGO
-
"Speaker Identification." StudyMode.com. 04, 2011. Accessed 04, 2011. http://www.studymode.com/essays/Speaker-Identification-663931.html.
