Skip to content

Supporting code for the paper "A study on more realistic room simulation for far-field keyword spotting".

Notifications You must be signed in to change notification settings

ebezzam/room-simulation

Repository files navigation

Room impulse response generation code

Supporting code for the paper "A study on more realistic room simulation for far-field keyword spotting".

Measured RIR dataset can be downloaded here.

Installation

First install Python dependencies:

pip install -r requirements.txt

Then install code for room impulse response generation:

# pyroomacoustics
pip install pyroomacoustics
# pygsound
pip install git+https://github.com/RoyJames/pygsound

For pygsound, please refer to their project page for additional dependencies to install.

Scripts

Compare computation time between pyroomacoustics and pygsound:

  • Varying number of rays: profile_nrays.py
  • Varying number of specular reflections / ISM order: profile_ism_order.py

Preparing RIR datasets:

  • Split original data into train and dev sets: create_measured_room_data_split.py
  • Simulate measured rooms based on metadata: simulate_measured_room_dataset.py

About

Supporting code for the paper "A study on more realistic room simulation for far-field keyword spotting".

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published