What is it about?
ConceptBeam is a system that extracts the speech of speakers speaking about a concept, i.e., a topic of interest, using a concept conveyed by speech and images as a clue. This system can be applied to innovative applications such as listening AIs that focus on a particular topic discussed in a conversation.
Featured Image
Photo by Priscilla Du Preez on Unsplash
Why is it important?
The clues used for sound source separation have been the location of the sound sources, harmonic relationships, statistical independence of the signals, and properties of the sound sources such as the voice characteristics of the target speaker. All of these clues have been based on the "appearance" of the signal. We have developed the first technology in history to extract signals based on the semantic content or concept conveyed by the signal, rather than on the appearance of the signal.
Read the Original
This page is a summary of: ConceptBeam, October 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3503161.3548397.
You can read the full text:
Resources
Contributors
The following have contributed to this page
