Room acoustics: How to characterise and improve acoustics of a classroom

Benoit Keryer - Théo Vidal

🇬🇧 Project file (soon) — 🇫🇷 Dossier projet — 🇫🇷 Transparents

This repository contains the source code used in a study projects to characterise rooms in terms of acoustic metrics (R60, EDT, C50...). Based on audio records of claps and sineswepts, this code calculates all metrics and outputs the results in a .csv file for investigation.

• 🚀 How to use this project
  • Step 1 - Apply the experimental protocol
  • Step 2 - Run the script and extract metrics
  • Step 3 - Create a numerical simulation
• 🖼️ Sample results
• 📜 Credits
• 🔐 License

🚀 How to use this project

Step 1 - Apply the experimental protocol

The simplest protocol:

• Place a microphone in a room
• Produce a short impulsion in the room (get close as posible to a Dirac impulsion), for instance by clapping your hands or making balloons explode
• Record the output

A more refined protocol:

• Place as well a speaker in the room
• Play a sweptsine signal in the speaker. This type of signal is defined using a central frequency, and allows for a more precise calculation.
  • A tool script is included to generate sweptsine signals
• Record the response of the room

Illustration of the experimental protocol used for the experiments

Step 2 - Run the script and extract metrics

• Place all your data in a folder using the following structure:

  data/
      room_1/
          audio_1.wav
          audio_2.wav
          ...
          audio_n.wav
      room_2/
          ...
  

  • If an audio is the room response to a sweptsine, the filename must have a format that can be set in the script using the following named groups:
    • f1 and f2 and the minimum and maximum frequencies of the signal
    • Fs is the sampling rate
    • T is the duration of the signal (not the audio) The default format is: sweptsine_f1-f2_Hz_-_Fs_T_s.wav
• Set up the parameters in the main MATLAB script main.m
• Run the script to extract all data in a .csv file
  • The script will ask for each audio signal to set a maximum treatment duration. Graphs of the signal energy are shown in log-scale, it is common to stop the treatment when the main .. has been observed

Metrics include:

• The reverberation time calculated using the T30
• The reverberation time calculated using the T20
• The Early Decay Time (EDT)
• Definitions and speech clarity at 50ms (D50, C50)
• Speech clarity at 80ms

Step 3 - Create a numerical simulation

You can optionally try to recreate the experiments in a numerical simulation. We used CATT-Acoustics v8, examples can be found in the corresponding folder.

🖼️ Sample results

Experimental data

Simulations

Numerical reproduction of a classroom

Numerical reproduction of an amphitheater

📜 Credits

• Main contributors: Théo Vidal, Benoit Keryer
• Supervisor: Karsten Plamann

🔐 License

GNU GPL v3