Title:
Blind Source Separation for Real-World Speech Applications
Speaker:
Hiroshi Saruwatari
Nara Institute of Science and Technology
Abstract:
In this talk, I review blind source separation technology and its recent current. Blind source separation (BSS) is an approach taken to estimate original source signals using only observed signals without knowing a priori information. Owing to the attractive feature of BSS, much attention has been paid to BSS in many fields of signal processing. In this talk, first, the basic principle of speech/acoustic signal separation based on independent component analysis is explained, and the recent research and development trends are shown.
Next, I can demonstrate our recently developed real-time BSS unit which is one of the world's smallest BSS microphone miniaturized into pocket-size hardware.
Finally, I show a hands-free spoken dialogue system with real-time BSS used for real-world speech applications.