KENDİ KENDİNİ DÜZENLEYEN HARİTALAR YÖNTEMİYLE TÜRKÇE SESLİ HARFLERİN SINIFLANDIRILMASI VE TANINMASI

Ozet: Lokal dinamik modelleme teknikleri kullanilarak zaman serilerini modellemek ozellikle son yillarda oldukca basarili sonuclar vermektedir. Kohonen'in 1990 yilinda sundugu 'Kendi Kendini Duzenleyen Haritalar' yontemi ile lokal dinamik modelleme teknigine farkli bir bakis acisi kazandirilmistir. Bu yontem ile, zaman serilerinden turetilen lokal dinamik modeller, sinyalin tum dinamikleri oldukca basarili ve kolay bir yontemle gosterebilmektedir. Zamanla bu teknik pek cok alanda kendine uygulama alani bulmus, gerek Kohonen, gerekse diger uzmanlar tarafindan pek cok farkli versiyonu turetilmistir. Yapilan calismada SOM yontemi kisaca aciklanmis ve bu yontem yardimiyla Turkce sesli harfler icin siniflandirma ve tanima uygulamasi yapilmis ve sonuclari tartisilmistir. Anahtar Kelimeler: Kendi Kendini Duzenleyen Haritalar, Lokal Dinamik Modelleme, Ses Tanima. The Classification and Recognition of Turkish Vowels with Self-Organizing Maps Abstract: The easiness of putting the model into practice, and making signal or system dynamics and structure observable has made dynamic modeling for time series very popular for the last years. By the years, new versions and approaches of dynamic modeling have been developed and applied to different kind of signals and systems. Kohonen's suggestion was a new approach to local dynamic modeling which he had offered in 1990. The new technique's name was 'Self Organizing Maps'. The innovation of the new approach was its needless of the memory for saving the history of time series. It was because the whole model is updated with the new sample of time series. New versions of this technique are introduced in a lot of different kinds of applications by the years. In this work, 'Self Organizing Maps' technique is applied to Turkish vowels and worked on the advantages of the technique.