ΝΕΑ ΜΕΘΟΔΟΣ ΑΥΤΟΜΑΤΗΣ ΑΝΑΓΝΩΡΙΣΗΣ ΟΜΙΛΗΤΗ

THE THESIS PRESENTS A NEW METHOD FOR AN AUTOMATIC SPEAKER RECOGNITION SYSTEM, WHICH COULD BE FAVORABLY IMPLEMENTED FOR IDENTIFICATION AS WELL AS FOR VERIFICATION PURPOSES. THIS METHOD IS BASED ON A SUBSTANTIALLY REDUCED NUMBER OF ELEMENTS OF THE PATTERN VECTOR IN COMPARISON TO SIMILAR METHODS. THE ELEMENTS ARE FREQUENCIES OF THE FIRST THREE FORMANTS WHICH CORRESPOND TO CHARACTERISTIC PEAKS OF THE ENERGY-TIME TRAJECTORY OF THE SPEECH SIGNAL AND ARE EXTRACTED AUTOMATICALLY WITHOUT PREVIOUS WARPING OF THE SPEECH SIGNAL. THE SPEECH EVENTS ARE LOCATED TAKING INTO ACCOUNT THE DISTANCE BETWEEN ADJACENT EXTREMES AND THEIR ENERGIES. IT SHOULD BE NOTICED THAT THE UTTERANCE IS SPECIALLY CHOSEN TO PRESENT THE DISTINCTIVE EXTREMES (SPEECH EVENTS).

Grantor dc:publisher

University of Patras

Year dc:date

1986

Authors dc:creator

• Φακωτάκης, Νικόλαος
• Fakotakis, Nikos

• Αναγνώριση ομιλητή
• ΑΥΤΟΜΑΤΗ ΜΕΘΟΔΟΣ ΑΝΑΓΝΩΡΙΣΕΩΣ ΟΜΙΛΗΤΗ
• ΕΞΑΚΡΙΒΩΣΗ ΟΜΙΛΗΤΗ
• Επιβεβαίωση ομιλητή
• AUTOMATIC SPEAKER RECOGNITION
• Speaker identification
• Speaker recognition
• Speaker verification
• Επιστήμες Μηχανικού και Τεχνολογία
• Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
• Engineering and Technology
• Electrical Engineering, Electronic Engineering, Information Engineering

Language dc:language

gre

Identifier

10.12681/eadd/0048

OAI identifier oai:identifier

oai:10442/0048