Skip to content

Latest commit

 

History

History
98 lines (65 loc) · 7.42 KB

README.md

File metadata and controls

98 lines (65 loc) · 7.42 KB

I3DA-2021: Measurement of a car cabin binaural impulse response and auralization via convolution

GitHub release GitHub last commit GitHub issues LICENSE


Repository with support files for the I3DA 2021 paper "Measurement of a car cabin binaural impulse response and auralization via convolution".

Presented in International Conference on Immersive and 3D Audio · 8-10 September 2021, Bologna, Italy.

Description

This repository contains the resources necessary to reproduce the complete pipeline for recording and reproducing binaural signals, in the scope of the paper. The processing chain, sofa files, and binaural examples are also available.

Abstract

Auralization is a well-known method used to create virtual acoustic scenarios. This work elaborates techniques for the Binaural Room Impulse Response (BRIR) extraction of the interior of a passenger car cabin. As a novelty, this study developed a binaural recorder based upon digital MEMS (micro-electro-mechanical system) microphones combined with an Arduino compatible microcontroller (Teensy). To improve the recordings, spectral corrections to the acquired data were carried out. The system is compact and cost-effective, creating an advantage for those measuring different source-receiver positions. For this paper, several combinations of BRIRs were tested, including combinations of binaural recordings for ambiance. Finally, equipped with the BRIRs database, it was possible to auralize different situations in real-time, for example, a conversation between two people (with the possibility to switch/tune background music and engine sound on and off). Furthermore, it was also possible to combine sources and receivers simultaneously with a contactless camera head tracker. Listening tests revealed that the spatial impression was preserved, rendering exceptional results as a virtual environment.

Folder structure:

  • /Auralization: Contains the scripts to assemble, post process, and make auralization with .sofa files.

Research pipeline

This work has developed the following steps.

Measurement setup

The binaural recorder used for the measurements comprises a pair of I2S digital microphones connected to a Teensy 4.0 microcontroller (MCU), and has capabilities such as real-time filtering and plug-and-play audio interface behavior. Furthermore, in order to track the head orientation of the subjects, a contactless camera-based head tracker was implemented based on Google's MediaPipe library.

Auralization setup

Assemble SOFA files

  • generateSOFAfromMeasurements.m: From the measured files this function allows you to create the MultiSpeakerBRIRs generated for this paper (The source positions and listener head orientations are pretty much specific to the conditions of this experiment).
  • Janelamento.m: Time-windowing the generated SOFA file.
  • Frequency_filtering.m: Apply calibration and band filter the SOFA file.
  • filtro_para_correcao_dos_mics.m: Generates the calibration filters from the measured magnitude (relative calibration).

Real-time auralizations and rendered scenes

The code BRIR_auralization recreates the Scenarios 1, 2, and 3 described in the paper, just pick the scene by changing the parameter scene at the Config cell.

The rendering pipeline for these scenes follow the structure bellow:

The camera based head-tracker can be found here.

Cite us

William D’Andrea Fonseca; Felipe Ramos de Mello; Davi Rocha Carvalho; Paulo Henrique Mareze; Olavo M. Silva. “Measurement of car cabin binaural impulse responses and auralization via convolution,” in International Conference on Immersiveand 3D Audio — I3DA, Bologna, Italy, Sep. 2021, pp. 1–13. doi: 10.1109/I3DA48870.2021.9610834.

Bibtex:

@InProceedings{I3DA:binauralMEMS,
  author    = {William {\relax D’A}ndrea Fonseca and Felipe Ramos de Mello and Davi Rocha Carvalho and Paulo Henrique Mareze and Olavo M. Silva},
  booktitle = {{International Conference on Immersive and 3D Audio --- I3DA}},
  title     = {Measurement of car cabin binaural impulse responses and auralization via convolution},
  year      = {2021},
  address   = {Bologna, Italy},
  month     = {Sep.},
  pages     = {1--13},
  doi       = {10.1109/I3DA48870.2021.9610834},
  abstract  = {Auralization is a well-known method used to create virtual acoustic scenarios. This work elaborates techniques for the Binaural Room Impulse Response (BRIR) extraction of the interior of a passenger car cabin. As a novelty, this study developed a binaural recorder based upon digital MEMS (micro-electro-mechanical system) microphones combined with an Arduino compatible microcontroller (Teensy). To improve the recordings, spectral corrections to the acquired data were carried out. The system is compact and cost-effective, creating an advantage for those measuring different source-receiver positions. For this paper, several combinations of BRIRs were tested, including combinations of binaural recordings for ambiance. Finally, equipped with the BRIRs database, it was possible to auralize different situations in real-time, for example, a conversation between two people (with the possibility to switch/tune background music and engine sound on and off). Furthermore, it was also possible to combine sources and receivers simultaneously with a contactless camera head tracker. Listening tests revealed that the spatial impression was preserved, rendering exceptional results as a virtual environment.},
  comment   = {PACS: 43.10.Pr, 43.38.Gy, 43.55.Ka, 43.66.Pn, 43.55.Hy, 91.10.Lh },
  keywords  = {Binaural Room Impulse Response (BRIR), digital MEMS microphone, auralization, convolution, measurement, Finite Element Method (FEM), filtering, signal processing},
}