Group in Noise (GiN) dataset

This page provides some illustrative examples of the recordings in the GiN database. Full information about the database can be found in the paper and the full data can be access through this link.

The examples here are best presented through headphones.

Task comparison

This example shows the difference in sound scene between the different tasks performed in each session, as measured at the reference microphone.

• Task 1: reading task. Speakers must read a passage from Alice in Wonderland one at a time, such that there should not be any overlapping speech. Listen
• Task 2: group task. The group must decide on a holiday destination that suits everyone’s secret constraint. There should be mild levels of overlapping speech. Listen
• Task 3: pair task. Participants must play a game of I spy while using synonyms to yes and no. There should be almost continuous levels of overlapping speech. Listen
• Task 4: group task. The group must play a game of Go fish with clear turn taking. There should be limited levels of overlapping speech. Listen

These excerpts are taken from the room_faraday/session_1/reference_audio/* files.

Channel comparison

This example shows the difference in sound scene capture between the different close-talking microphones, the binaural microphones worn by ‘Participant 3’, and the reference microphone. The baseline enhancement is also provided.

Click on the buttons to listen to the microphone recordings.

	BASELINE

These excerpts are taken from the room_faraday/session_1/*/02_00_000 files.

Room comparison

This example shows the difference in sound scene between rooms in the database, as measured at the reference microphone.

Click on each picture to listen to the recordings.

Room 403	Room 807	Room Faraday

These excerpts are taken from the room_403/session_2/reference_audio/00_00_000 and room_*/session_1/reference_audio/00_00_000 files.