Thesis (M.Sc.) - Cairo University - Faculty of Engineering - Department of Electronics and Communications

Speaker Diarization is known as the task that answers the question, who spoke, when in an audio {uFB01}le or a set of audio {uFB01}les that contain unknown number of speakers. The determination of speaker segments is done in an unsupervised manner. Our Speaker Diarization system composed of two main blocks; Speech Activity Detector and Speaker Clustering. In speech activity detection we propose several solutions including; Phoneme Recognition system, SVMHMM system and i-vector based system. In speaker clustering area we propose an enhancement over state of the art techniques as cosine based Hierarchal Agglomerative Clustering. Such enhancement including enhancing clustering by classi{uFB01}cation methods as SVM, DNN and Random Forrest. Finally we investigated enhancing the i-vector representation via extracting them from a DNN based background model

Issued also as CD