Automatically understanding the content of audio data is useful for audio indexing, target-based distribution of media, and intelligent remixing. It includes audio segmentation, which divides an audio signal into homogenous segments like music or speech. This talk explains how to develop a state-of-the-art segmentation algorithm by combining deep learning techniques and artificial training set synthesis.
In addition, we present a novel idea for real-time radio remixing. It is a customised radio that plays diary reminders and music from the user’s playlist. Using audio segmentation, we ensure that transitions between live radio and the remixed content are seamless.
This talk is part of the RadioMe project
Bio
Satvik holds a Bachelor of Technology in Information and Communication Technology from SASTRA University, India and a Master of Research in Computer Music from ICCMR, University of Plymouth, UK. He currently is studying for a PhD in ICCMR on the topic of audio segmentation and intelligent mixing for live radio broadcast. His research interests include Deep Learning, Brain-Computer Interfaces, and Unconventional Computing for music. Satvik is also an accomplished musician and performer.
