Voice Recognition for Smart Homes Applications

This video demonstrates the integration of microphone arrays, machine learning techniques, and wireless sensor/actuator networks for the design and development of a robust, hands-free speech, multilingual control interface for smart home systems. The platform has been developed by the LISTEN project consortium, and with the collaboration of the TNL laboratory, FORTH-ICS.

Multiple Sound Sources Separation

These videos demonstrate how the sound sources can be isolated based on the previous step of sound localization using the microphone array. In this demo, 3 persons are speaking simultaneously and a music clip is also playing at the same time. Although all sound sources are close to each other, the separation performed by our method achieves an excellent isolation of the sound sources. Using our method, a listener can also interactively control the sound level of the sound he/she wishes to isolate compared to the background sounds, so that the source can be totally or partly isolated, depending on the application.


Sound Sources Localization

This video demonstrates the real-time localization of multiple simultaneous sound sources using our methods. On the conference table, on top of the microphone array, we project a graphical representation of the localization result using our algorithm. Each circle corresponds to a sound source localized by the method, and its color is indicating the relevant sound source energy at each time instant.