Özkanca, Yasin Serdar2019-11-262019-11-262018-12-19http://hdl.handle.net/10679/6285http://discover.ozyegin.edu.tr/iii/encore/record/C__Rb3101891?lang=enghttps://tez.yok.gov.tr/Depression is a common mental health problem around the world with a large burden on economies, well-being, hence productivity, of individuals. Early diagnosis and detection of depression can aid treatment, but diagnosis typically requires an interview with a health provider or structured diagnostic questionnaire. Thus, unobtrusive measures that might be able to monitor depression symptoms in daily life could have great utility in monitoring depression for clinical treatment. Vocal biomarkers of depression are a potentially e ective method of assessing depression symptoms in daily life, which is the focus of the current research. Although there have been e orts to automatically assess depression levels from audiovisual features, use of transcriptions along with the acoustic features has emerged as a more recent research venue. Moreover, di culty in data collection and the limited amounts of data available for research are also challenges that are hampering the success of the algorithms. One of the novel contributions in this thesis is to exploit the databases from multiple languages for feature selection. Since a large number of features can be extracted from speech, and given the small amounts of training data available, e ective data selection is critical for success. Our proposed multi-lingual method was e ective at selecting better features and signi cantly improved depression assessment accuracy. In addition, text-based features were used for assessment and a novel strategy to fuse the text- and speech-based classi ers were proposed, which further boosted the performance.Depresyon, bireylere ekonomik, refah duzeyi, dolay s yla uretgenlik a c s ndan buyuk bir yuk olan, yayg n bir zihinsel sa gl k sorunudur. Erken tan ve depresyonun tespiti tedaviye yard mc olabilir, ancak tan genellikle bir sa gl k kurulu su ile ileti sim veya yap land r lm s tan sal bir anket gerektirir. Bu nedenle, gunluk hayatta depresyon belirtilerini izleyebilecek goze batmayan onlemler, klinik tedavi i cin depresyonun izlenmesinde buyuk yarar sa glayabilir. Depresyonun vokal biyobelirte cleri, guncel ara st rmalar n oda g olan, gunluk hayatta depresyon belirtilerini de gerlendirmede potansiyel olarak kullan labilecek etkili bir ara ct r. Gorsel-i sitsel ozelliklerden depresyon duzeylerini otomatik olarak de gerlendirme cabalar na ra gmen, akustik ozellikler ile birlikte yaz l metin kullan m daha yeni bir ara st rma alan olarak ortaya c km st r. Ek olarak, veri toplanmas ndaki zorluk ve ara st rmaya a c k s n rl miktarda veri de algoritmalar n ba sar s n engelleyen zorluklardand r. Bu makalenin sundu gu katk lardan biri, oznitelik se cimi i cin veritaban olarak birden cok dil kullanmakt r. Etkili bir oznitelik se cimi, az say da konu sma verisinden cok say da oznitelik elde edilebildi ginden dolay , ba sar l bir cozum i cin cok onemlidir. Onerilen cok dilli yontemimizin daha iyi oznitelikler se cmede etkili oldu gu ve depresyon de gerlendirme do grulu gunu onemli ol cude geli stirdi gi gozlemlendi. Ayrıca, de gerlendirme i cin metin tabanl oznitelikler de kullan ld ve performans artt rmas ad na metin ve konu sma temelli s n and r c lar birle stiren bir strateji onerildi.engrestrictedAccessMaster's thesisDemiroğlu, CenkFeature selectionMachine-learningSpeech processingSignal processing3101891