Pamukkale Univ Muh Bilim Derg. Baskýdaki Makaleler: PAJES-66263 | DOI: 10.5505/pajes.2024.66263

EEG tabanlý haberleþme sisteminde Levenshtein mesafe algoritmasýnýn doðruluk performansýna etkisinin incelenmesi

Güneþ Ekim¹, Nuri Ýkizler¹, Ayten Atasoy^2
1Karadeniz Teknik Üniversitesi Trabzon Meslek Yüksekokulu, Elektronik ve Otomasyon Bölümü, Trabzon
²Karadeniz Teknik Üniversitesi Elektrik-Elektronik Bölümü, Trabzon

Çalýþmamýzda, amyotrofik lateral skleroz, felç ve kilitli kalma sendromu gibi motor engeli olan kiþiler için, EEG sinyalleri kullanarak geliþtirdiðimiz göz-kýrpma iletiþim sistemi üzerinde Levenshtein mesafe algoritmasýnýn etkilerini araþtýrdýk. Sistem, göz-kýrpma sinyallerini analiz ederek bilgi çýkarmakta ve seslendirmektedir. EEG sinyalleri, kablosuz NeuroSky MindWave cihazý kullanýlarak sol gözün üzerine yerleþtirilmiþ bir elektrottan elde edilmiþtir. Göz-kýrpmalarla oluþturulan Mors kodlu kelimeler sisteme giriþ olarak verilmiþ ve Dalgacýk Dönüþümü kullanýlarak öznitelik vektörleri çýkarýlmýþtýr. Bu vektörlerle Destek Vektör Makineleri eðitilmiþ ve sýnýflandýrma hatalarýný azaltmak için Levenshtein mesafe algoritmasý kullanýlmýþtýr. Son olarak, Metin-Konuþma sentezi algoritmasý ile sistem tamamlanmýþtýr. Kendini ifade etmek için 20 kelimenin kullanýldýðý deneyler oldukça baþarýlý sonuçlar vermiþtir.

Anahtar Kelimeler: EEG, Beyin-bilgisayar arayüzü, Dalgacýk dönüþümü, Destek vektör makineleri, Levenshtein mesafe algoritmasý, göz-kýrpma tespiti

---

Investigating effects of Levenshtein distance algorithm on accuracy performance in EEG based communication system

Güneþ Ekim¹, Nuri Ýkizler¹, Ayten Atasoy^2
1Electronics and Automation Department, Trabzon Vocational School, Karadeniz Technical University, Trabzon
²Electrical-Electronics Engineering Department, Engineering Faculty, Karadeniz Technical University, Trabzon

In our study, we investigated the effects of the Levenshtein distance algorithm on the eye-blink communication system that we developed based on EEG signals for people with severe motor disabilities, such as Amyotrophic Lateral Sclerosis, stroke, and locked-in syndrome. The developed system analyzes eye-blink signals to extract information and vocalize it. EEG signals were obtained from an electrode above the left eye using a NeuroSky MindWave Mobile device. Morse-coded eye-blink words were input to the system and feature vectors were extracted using the Wavelet Transform method. Vector support machines were trained with these vectors and the Levenshtein distance algorithm was used to reduce classification errors. Finally, the system was completed with a text-to-speech synthesis algorithm. The experiments, which used 20 words for self-expression, obtained highly successful results.

Keywords: EEG, Brain-computer interface, Wavelet transform, Support vector machine, Levenshtein distance algorithm, Eye-blink detection

---

ATIF KOPYALA