Smart glasses audio dataset
This webinar notebook generates a labeled audio dataset across 10 restaurant environments, including the wearer's own voice, conversations, background music, and ambient noise. It walks through room simulations, impulse response collection, audio scene generation, and exporting audio and metadata for algorithm development.
The notebook uses webinar-fall26-pp4-glasses, which is included in the standard device
library. No device import or device simulation is required before running this example.
Running the room simulations requires approximately 22 Treble tokens for 10 rooms.
Want to try your own smart glasses design? Follow Make your own smart glasses device to create and register it, then update the notebook's device lookup to use your device.
For a complementary home-like test scene, see Recreate ETSI Annex B.