Güvenlik Sistemleri Için Mel Frekans Kepstrum Katsayilari ve Yapay Sinir Aglari Kullanilarak Konusmaci Tanima

Year 2007, Volume: 3 Issue: 1, 8 - 15, 01.04.2007

Ahmet Küçüker Can Yüzkollar Ahmet Sansli Fatih Sen

Abstract

Bu çalismada konusmaci ses örnekleri bilgisayara alinarak bu konusmacilarin kimliklerinin belirlenmesi
saglanmistir. Çalisma üç asamadan olusmaktadir. Ilk asama konusmaci sesinin dijital ortamda veriye dönüstürülmesi
islenebilir hale getirilerek özellik çikartilmasi ikinci asama elimizdeki konusmaci ses örnekleriyle yapay sinir aginin
egitilmesi ve son asama da egitilmis sinir agina test verileri göndererek istatistik bilgiler elde edilmesidir.

Keywords

Sinir Aglari, Konusmaci kimliklendirme, Konusmaci tanima, Mel Kepstrum Katsayilar

Speaker Recognition Using Mel-Frequency Cepstrum Coefficients and Artificial Neural Network for Security Systems

Abstract

In this study; the sample sounds of the speakers are transfered to computer for identifing the speaker.
Study is based on three steps; In first step analog sound that taken is converted to data to be processed and determine
own characteristic of each sound. in second step converted sounds called data are used to train neural network. at
last step trained neural network is test by the test sounds and obtained istatistical informations.

Keywords

Neurel Network, Speaker Identification, Speech Recognition, Mel Cepstrum Coefficients