Voice pattern recognition system

• Main Authors: Azahari, Azhar, Ferry, Darryl Paul R., Tirona, John
• Format: text
• Language: English
• Published: Animo Repository 1991
• Subjects: Voice; Pattern recognition systems; Computer vision; Coding theory; Speech processing systems
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/16391
• Institution: De La Salle University

Voice Pattern Recognition System (VoPRecSys), is a text independent speaker recognition system that is capable of recognizing the identity of the speakers from their input voices. The system first creates templates of the input signal, the resulting templates are then compared with the reference templates in the systems vocabulary library. The acoustic signals in the library are those previously trained by the trainer or trainers. If the input voice from the trainer exists in the vocabulary, then the system can identify the correct identification of the speaker. The criteria used in selecting the correct identity of the speaker is by the use of density features and the smallest Mahalanobis distance criteria. The system, though it is text-independent, will be able to recognize clear, discreet, and distinct input voices which are pronounced correctly and properly. Reliability of the system's performance depends on the way the user speaks the input voice. Also provided in the system are useful utilities that would help in the study of the data gathered during the recognition process. These utilities are in form of graphic displays of the input voice and the word listings uttered by different speakers. Though no specific application has been interfaced to the system, the possibility that it can be interfaced to any desired application is never impossible.