Title:  AUDIO DIARIZATION FOR LENA DATA AND ITS APPLICATION TO COMPUTING LANGUAGE BEHAVIOR STATISTICS FOR INDIVIDUALS WITH AUTISM

Committee: 

Dr. Clements, Advisor

Dr. Anderson, Co-Advisor        

Dr. Fekri, Chair

Dr. Moore

Abstract:

The objective of the proposed research is to develop algorithms for diarization of audio data collected using a device called LENA and apply them to compute language behavior statistics for individuals with autism. I-vector based methods are to be developed to diarize a LENA audio recording with limited labeled data available for training. The proposed research aims are 1) Development of semi-supervised and unsupervised i-vector methods to diarize audio recordings of infants and older children. 2) Use diarized output to compute language behavior statistics and study their role in early autism detection, and monitoring treatment changes.