Audio Content Search

Dr. Joachim Köhler showed several techniques and applications in the area of audio content search. Robust segmentation algorithms were applied to detect speech and non/speech events and to perform a speaker segmentation and clustering task. Further speech and music alignment techniques were demonstrated to generate a time code to existing textual and score descriptions. The current state-of-the-art technology for spoken document retrieval was given and alternative approaches and innovative applications were presented to avoid the Out-Of-Vocabulary problem in speech indexing tasks.