Echolalia_detection

Repository for my Master's Thesis 'Automatic detection of echolalic speech in children with Autism Spectrum Disorder'

Contents

1.A: Transcription models

Adapted from Whisper Timestamped (https://github.com/linto-ai/whisper-timestamped) (with and without prior speaker diarization)

1.B: Transcription-based echolalia and self-repetition detection models

1.C: Audio-based echolalia and self-repetition detection models

• Training and testing procedures of both models, on the basis of extracted features (for procedure of feature extraction: see final models)
• Final models, that include feature extraction, classification by the trained classifier and output of predictions in pandas DataFrame + .TextGrid