Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Simulating Room Acoustics

Sascha Spors
September 21, 2015

Simulating Room Acoustics

Talk given at the summer school of the FET-Open project Two!Ears on the generation of ears signals using the binaural simulator.

Sascha Spors

September 21, 2015
Tweet

More Decks by Sascha Spors

Other Decks in Science

Transcript

  1. Simulating Room Acoustics
    Sascha Spors et al.
    Universität Rostock
    Institut für Nachrichtentechnik
    Two!Ears Summer School
    21st September 2015, Toulouse

    View Slide

  2. Model Architecture
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Introduction 1 / 21

    View Slide

  3. Scenario-Based Development and Evaluation Process
    real environment
    observer
    source
    model
    ear signals vision
    feedback
    Real World
    virtual environment
    virtual observer
    virtual
    source
    model
    ear signals vision
    feedback
    Virtual World
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Introduction 2 / 21

    View Slide

  4. Scenario-Based Development and Evaluation Process
    real environment
    HATS
    source
    model
    ear signals vision
    feedback
    Real World
    simulation
    framework
    model
    ear signals vision
    feedback
    scenario database
    Virtual World
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Introduction 2 / 21

    View Slide

  5. Scenario-Based Development and Evaluation Process
    real-world input less controllable
    but potentially richer
    interactive and exploratory nature
    of model
    signals depend on feedback
    from upper stages
    block/frame-based computation
    of signals
    simulation
    framework
    model
    ear signals vision
    feedback
    scenario database
    Virtual World
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Introduction 2 / 21

    View Slide

  6. Techniques for Binaural Synthesis
    Degrees of realism
    capture of real-environments
    model/data-based synthesis of virtual environments
    numeric simulation of virtual environments
    Ear signals captured with
    individuals
    head and torso simulator (HATS)
    Underlying signal processing techniques
    recording/playback of ear-signals
    capture/convolution of source signals with impulse responses
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Introduction 3 / 21

    View Slide

  7. Pre-Recorded Binaural Signals
    Recording
    defined position in the (blocked) ear canal
    Playback
    headphone compensation
    Pros/Cons
    + complex scenarios
    – fixed head orientation
    – fixed scenarios
    Applications
    diffuse background noise
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 4 / 21

    View Slide

  8. Head-Related Impulse Responses (HRIRs)
    Recording
    identify impulse responses from a source to the left/right ears → HRIRs
    database of HRIRs for different source positions
    head/torso rotation equivalent to source at fixed distance
    Playback
    convolution of dry source signal with HRIRs
    Pros/Cons
    + copes for head orientation
    + variable position of source/receiver
    – no diffuse sources
    Applications
    (anechoic) scenarios with limited number of
    (static) sources
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 5 / 21

    View Slide

  9. Example – QU KEMAR with 3m Source Distance
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 6 / 21

    View Slide

  10. Example – QU KEMAR with 3m Source Distance
    Left/right HRIR for source 60o to the left side
    0 1 2 3 4 5
    time (ms)
    −1.0
    −0.5
    0.0
    0.5
    1.0
    left ear
    right ear
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 6 / 21

    View Slide

  11. Example – QU KEMAR with 3m Source Distance
    Left HRIRs for varying source angle
    −180 −135 −90 −45 0 45 90 135
    angle (deg)
    0
    1
    2
    3
    4
    5
    time (ms)
    0
    −10
    −20
    −30
    −40
    −50
    −60
    −70
    relative level (dB)
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 6 / 21

    View Slide

  12. Example – QU KEMAR with 3m Source Distance
    Left HRTFs for varying source angle
    −180 −135 −90 −45 0 45 90 135
    angle (deg)
    0
    5
    10
    15
    20
    frequency (kHz)
    20
    10
    0
    −10
    −20
    −30
    −40
    −50
    relative level (dB)
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 6 / 21

    View Slide

  13. The SoundScape Renderer (SSR)
    A versatile framework for spatial sound reproduction
    techniques: WFS, HOA, VBAP, binaural synthesis, ...
    Linux and OS X
    open source software (GPLv2)
    http://spatialaudio.net/ssr
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 7 / 21

    View Slide

  14. Binaural Room Impulse Responses (BRIRs)
    Recording
    identify room impulse responses from a source to the left/right ears → BRIRs
    database of BRIRs for different head-orientations
    Playback
    convolution of dry source signal with BRIRs
    Pros/Cons
    + copes for head rotation
    + includes room acoustics
    – fixed position of source/receiver
    Applications
    reverberant scenarios with low number of static
    sources
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 8 / 21

    View Slide

  15. Example – QU KEMAR in Auditorium
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 9 / 21

    View Slide

  16. Example – QU KEMAR in Auditorium
    Left/right ear HRIR for head rotated 60o to the right side
    0 50 100 150 200
    time (ms)
    −1.0
    −0.5
    0.0
    0.5
    1.0
    left ear
    right ear
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 9 / 21

    View Slide

  17. Example – QU KEMAR in Auditorium
    Left HRIRs for varying head rotation
    −90 −45 0 45
    angle (deg)
    10
    20
    30
    40
    50
    60
    70
    80
    90
    100
    time (ms)
    −60
    −54
    −48
    −42
    −36
    −30
    −24
    −18
    −12
    −6
    0
    relative level (dB)
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 9 / 21

    View Slide

  18. Example – QU KEMAR in Auditorium
    Left HRTFs for varying head rotation
    −90 −45 0 45
    angle (deg)
    0
    5
    10
    15
    20
    frequency (kHz)
    25
    20
    15
    10
    5
    0
    −5
    −10
    −15
    −20
    −25
    −30
    relative level (dB)
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 9 / 21

    View Slide

  19. Dynamic Binaural Room Impulse Responses
    Recording
    time-variant room impulse responses from a moving source to the left/right ears
    Playback
    time-variant convolution of dry source signal
    Pros/Cons
    + includes acoustic properties of room
    – fixed trajectory of source/receiver
    – fixed head orientation
    Applications
    reverberant scenario with low number of dynamic
    sources
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 10 / 21

    View Slide

  20. Data-based Binaural Synthesis
    Recording
    identify impulse responses from a source to a microphone array
    decomposition of captured sound field into plane waves
    Playback
    filtering of plane waves with HRIRs → BRIRs, convolution of source signal
    Pros/Cons
    + copes for small translatory movements
    + head rotation possible
    + individual HRIRs
    – nearly fixed position of source/receiver
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Binaural Synthesis Techniques 11 / 21

    View Slide

  21. Room Acoustic Models
    Numerical/Wave-based
    (numerical) solution of the wave equation
    includes all effects due to the wave nature of sound
    complex for large rooms/small wavelegths
    Geometrical
    decomposes sound propagation into rays of sound
    effects due to wave nature are not fully covered
    applicable for large rooms/small wavelengths
    Statistical
    statistical properties of perceived room impression
    underlying assumption: diffuse sound field
    buildup and decay of sound, stationary sound field in rooms
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Numeric Simulation 12 / 21

    View Slide

  22. Numerical Simulation of Room Acoustics
    Various simulation environments compute the sound field in a room numerically
    State of the art
    hybrid model: mirror image model, raytracing
    closed (and coupled) rooms with arbitrary geometry
    frequency depend boundary conditions, directive sources/receivers
    estimation of various room acoustic measures, auralization
    Selected frameworks
    ODEON – Room Acoustics Software, http://www.odeon.dk/
    EASE – Enhanced Acoustic Simulator for Engineers, http://ease.afmg.eu/
    CATT – Acoustic, http://www.catt.se/
    RAVEN – Room Acoustics for Virtual ENvironments, RWTH Aachen
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Numeric Simulation 13 / 21

    View Slide

  23. Measurement vs Simulation of Acoustic Environments
    Measurement/Recording
    + fine structure of responses
    + capture of diffuse sound fields
    – effort for multiple source/listener positions and environments
    Numeric Simulation
    + arbitrary position of source/receiver
    + individual HRIRs
    – limited natural fine structure
    – high computational complexity for complex/large environments
    Caution: perceptual equivalence does not guarantee physical equivalence!
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Numeric Simulation 14 / 21

    View Slide

  24. Summary – Techniques for Binaural Synthesis
    technique diffuse sound
    field
    moving
    sources
    head-
    orientation
    head-
    translation
    realism application
    pre-recorded
    binaural signals
    yes yes no no high background noise
    static HRIR/BRIR limited limited yes limited high static
    sound sources
    dynamic BRIR no yes no no high moving
    sound sources
    data-based
    synthesis
    yes no yes small-scale high static
    sound sources
    numerical
    simulation
    limited yes yes yes limited acoustic
    navigation
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Numeric Simulation 15 / 21

    View Slide

  25. Database of Audio-Visual Data
    synthesis of ear/eye signals requires data characterizing the environment
    evaluation requires physical and perceptual labels
    central database of audio-visual data and perceptual labels
    Infrastructure
    public and project-internal database
    software interface for seamless access
    common data formats
    https://github.com/TWOEARS/data
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Database 16 / 21

    View Slide

  26. Data Formats
    Binary Data
    available datasets are stored in various formats
    sounds converted to Waveform Audio File Format (WAV)
    IRs converted to Spatially Oriented Format for Acoustics (SOFA)
    Audio-Visual Scene Description
    XML-based description of scene
    references to binary data
    supported by binaural simulator
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Database 17 / 21

    View Slide

  27. Current Contents of Database
    Head-related transfer functions
    (SCUT, MIT, KEMAR, KU100, ...)
    Binaural room impulse responses
    (various rooms at TUB, URO, Surrey, ...)
    Sounds (monaural, binaural)
    (speech, environmental noise, ...)
    Perceptual labels
    (on-/offsets, head movements, localisation/coloration)
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Database 18 / 21

    View Slide

  28. The Binaural Simulator
    Acoustic Scene
    Listener
    Sources &
    Environment
    XML
    Parser
    Scene
    Description
    File
    MEX
    SoundScape
    Renderer
    Robot Interface
    Knowledge Sources
    scene objects
    frame size, sample rate, ...
    ear signals
    listener &
    sources
    head
    rotation
    &
    translatory
    movement
    current head
    orientation
    control ear signals
    https://github.com/TWOEARS/binaural-simulator
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Implementation 19 / 21

    View Slide

  29. Features
    Virtual scenes composed from
    pre-recorded binaural signals
    virtual sources synthesized by HRIRs/BRIRs
    moving virtual sources by concatenation of static positions
    spatial room response/ambience via plane wave decompositions
    virtual rooms by two-dimensional mirror image source model
    Outlook
    integration of numerical room acoustic simulation (RAVEN)
    three-dimensional scenes and HRIRs/BRIRs
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Implementation 20 / 21

    View Slide

  30. Conclusions/Outlook
    framework for scenario-based research and development
    open source, systematic evaluation ⇒ reproducible research
    best practice for other research projects
    Spors et al. | 21.9.2015 | Two!Ears – Simulating Room Acoustics | Conclusions 21 / 21

    View Slide