All Projects / Acoustic Room Geometry Reconstruction
Previous / Next
Acoustic Room Geometry Reconstruction image

Acoustic Room Geometry Reconstruction

Understanding the location of acoustically reflective surfaces in a room is a critical component in advanced sound processing. For example, intelligent speakers can use a room’s acoustic geometry to improve playback quality, source separation accuracy, and speech recognition. In this paper, we present Synesthesia, a system for capturing the acoustic properties of a room using a single fixed speaker and a mobile phone that records audio at multiple locations. Using the arrival time of echoes, the system is able to reconstruct the position of reflective surfaces like walls and then estimate properties like surface absorption.

In this project, we introduce a new approach that performs RIR imaging using a mobile phone that tracks its location with visual inertial odometry (VIO) to record a dense set of samples albeit with noise in their locations. We present a new approach that is able to relax several key assumptions on RIR and show through both experimentation and simulation that even with 20cm of uncertainty in the microphone locations provided by VIO, we are still able to reconstruct the room geometry with accurate shape and dimensions. We demonstrate this capability by prototyping a tool for acoustic engineers, that allows a user to view a room’s estimated geometry and absorption overlaid on the actual sensed space with augmented reality.

Latest News

  • Apr 15, 2019:
    Two new CPS Week papers at IPSN and ICCPS.
  • Mar 28, 2018:
    Congrats to Hsu-Chieh for defending!
  • Aug 31, 2018:
    Congrats to Max for defending!
  • Jun 29, 2018:
    Congrats to Oliver for defending!
  • Apr 13, 2018:
    Congrats to Niranjini, John, Krishna and Anh for winning best demo at IPSN for their AR work.
  • Apr 12, 2018:
    Congrats to Adwait, Revathy, Akshay, Anh and Artur for winning best paper at IPSN for Charm.
  • Apr 11, 2018:
    CMU and Yodel Labs (ALPS spinout) came in first and second at the Microsoft Indoor Localization Competition.
  • Apr 10, 2018:
    Congrats to Luis for defending!
  • Apr 10, 2018:
    Three new papers at IPSN in Porto!
  • Jan 15, 2018:
    The CONIX Research Center awarded $27.5M (press release)
  • Nov 26th, 2017:
    New ARPA-E project with Endeveo on people counting
  • Aug 11, 2017: Congrats to Patrick for defending!
  • Jul 24, 2017:
    Great NPR Morning Edition segment talking about our safe cyclist project
  • Jun 18, 2017:
    New NIST project on helping Firefighters navigate in buildings.
  • May 22, 2017:
    Congrats to Adwait for winning Best Presentation at RTAS this year for his Pulsar Paper.
  • Aug 15, 2016:
    New DOE Project on HVAC Control
  • Aug 10, 2016:
    New ARPA-E GridBallast project on Load Control with NRECA
  • May 12, 2016:
    Congrats to Niranjini for winning one of this year’s Samsung Fellowships
  • Apr 10, 2016:
    Initial public release of our CMU localization data set
  • May 12, 2016:
    Congrats to Niranjini for winning one of this year’s Samsung Fellowships
← Previous
Next →