Group in Noise (GiN) data - 2023

This work is licensed under a Creative Commons Attribution Non-Commercial 4.0 International public license agreement.

This collection contains the Group in Noise (GiN) data presented in (d'Olne et al., 2023). A detailed description of the data and useful tools are available at https://github.com/ImperialCollegeLondon/sap-ic-gin.

The structure of the dataset is as follows

┬── array_irs.h5 (file): full acoustic impulse response for the glasses array.
├── array_irs_v2.h5 (file): cropped acoustic impulse response for the glasses array.
├── room_ID (directory): contains all the data for a given room.
│   ├── array_audio (directory): contains the array audio (glasses + binaural mics)
│   │   │  for the listener for all sessions in the room.
│   │   └── session_# (directory): the session directories for the array audio.
│   │       └── ##_##_###.wav (files): 7-channel audio wav files.
│   ├── close_talking_audio (directory): contains the close-talking microphone
│   │   │  recordings for the 6 participants (not including the waiter) for
│   │   │  all sesssions in the room.
│   │   └── session_# (directory): the session directories for the close-talking audio.
│   │       └── ##_##_###.wav (files): 6-channel audio wav files, with channels 1-6
│   │              corresponding to speakers 1-6.
│   ├── metadata (directory): contains relevant information about the room.
│   │   ├── loudspeakers_tascar.txt (file): loudspeaker position in right-hand, z-up
│   │   │      coordinate system used in Tascar.
│   │   └── session_# (directory): the session directories for metadata.
│   │       └── tasktimings.csv (file): csv file containing the timing information for
│   │              each of the task in the conversation.  
│   ├── reference_audio (directory): contains the audio from the reference microphone 
│   │   │  for all sessions in the room.
│   │   └── session_# (directory): the session directories for the reference microphone.
│   │       └── ##_##_###.wav (files): single-channel wav files from the reference 
│   │              microphone.
│   ├── tracked_data (directory): contains head pose data (position and orientation) for
│   │   │  all participants and all sessions in the room, in a compact format.
│   │   └── session_# (directory): the session directories for head pose data.
│   │       └── ##_##_###.json (files): JSON files containing the pose information
│   │              for each participant in Cartesian coordinates and Euler angles, with
│   │              the right-hand, z-up coordinate system used in Tascar.
│   ├── tracked_data_tascar (directory): contains head pose data for all participants
│   │   │  and all sessions in the room, in a format ready to use with Tascar.
│   │   └── session_# (directory): the session directories for head pose data.
│   │       ├── ##_##_###_speaker#_position.csv (files): csv files containing Cartesian
│   │       │      coordinates for each participant.         
│   │       └── ##_##_###_speaker#_rotation.csv (files): csv files containing Euler angles
│   │              for each participant.    
│   ├── vad_data (directory): contains automatically generated VAD data for the
│   │   │  close-talking microphones (including the waiter) for all sessions in the room.
│   │   └── session_# (directory): the session directories for VAD data.
│   │       └── ##_##_###.json (files): JSON files containing the start and end
│   │              frame indices of voice activity, as well as a labels indicating
│   │              the active speaker and the task being recorded.
└── └── waiter_audio (directory): contains the close-talking microphone recording
        │  for the waiter.
        └── session_# (directory): the session directories for waiter audio.
            └── ##_##_###.wav (files): single-channel audio wav files containing
                   close-talking microphone recordings for the waiter.