sound field analysis and control for virtual acoustics in cars
TRANSCRIPT
Sound Field Analysis and Control for Virtual Acoustics in Cars
PHD STUDENT: MICHAEL STRAUSS
SUPERVISORS: PROF. ALOIS SONTACCHI, PROF. ROBERT HÖLDRICH, PROF. MIKAEL STERNAD (UPPSALA UNIVERSITY)
PHD-FORUM SOMMERSEMESTER 2021, KU GRAZ
2021-06-19
Introduction Big picture: Virtual Product Development PhD topic
Part #I: Summary and key outcome
Current research Focus on spatial audio perception
Part #II: State of the art, auditory models
Experimental work Working with auditory models
Ongoing and next steps
Content
19.06.2021 [email protected] 3
IntroductionStarting point: An incredible car audio system – WFS
19.06.2021 [email protected] 5
Wave Field Synthesis and Virtual Room Acoustics rendering system capabilities (Audi Sound Concept) [Strauss, Gleim, AES2012]
62 Loudspeakers (52 midrange, 5 tweeters, 4 woofers, 1 sub-woofer) Wave Field Synthesis rendering Virtual Room Acoustics processing Authoring and Sound Management GUI
IntroductionStarting point: An incredible car audio system – WFS
19.06.2021 [email protected] 6
Wave Field Synthesis and Virtual Room Acoustics rendering system capabilities (Audi Sound Concept) [Strauss, Gleim, AES2012]
IntroductionStarting point: An incredible car audio system – WFS
19.06.2021 [email protected] 7
Wave Field Synthesis and Virtual Room Acoustics rendering system capabilities (Audi Sound Concept) [Strauss, Gleim, AES2012]
IntroductionStarting point: An incredible car audio system – WFS
19.06.2021 [email protected] 8
Wave Field Synthesis and Virtual Room Acoustics rendering system capabilities (Audi Sound Concept) [Strauss, Gleim, AES2012]
IntroductionStarting point: An incredible car audio system …
… and a lot of question marks!
19.06.2021 [email protected] 9
Let‘s frame this research into the bigger picture of industrialproduct development … virtual product development …
IntroductionSound Field Analysis & Control for Virtual Acoustics in Cars
19.06.2021 [email protected] 10
Acquisition
Manipulation
Assessment
Simulation
Measurement
Subjective(Auralization, VR, …)
Objective(Quantification, …)
Virtual Tuning
Optimization
Data base Knowledge baseVirtual Model
(Car audio system)+ +
Listening evaluation
+ sound tuning
Numericaldescription
+ control
IntroductionVirtual Product Development applied to Acoustics (Industry 4.0)
19.06.2021 [email protected] 11
Acquisition
Manipulation
Assessment
Simulation
Measurement
Subjective(Auralization, VR, …)
Objective(Quantification, …)
Virtual Tuning
Optimization
Data base Knowledge baseVirtualModel
+ +
Sound Field Analysis & Control for Virtual Acoustics in Cars
IntroductionVirtual Product Development applied to Acoustics (Industry 4.0)
19.06.2021 [email protected] 12
Acquisition
Manipulation
Assessment
Simulation
Measurement
Subjective(Auralization, VR, …)
Objective(Quantification, …)
Virtual Tuning
Optimization
Data base Knowledge baseVirtualModel
+ +
Optimization
Sound Field Analysis & Control for Virtual Acoustics in Cars
IntroductionPart I: Overview
Tracing back to the origin of wave field artefacts How to measure/analyse the performance of a sound field reproduction method (e.g. Wave Field Synthesis)?
How to measure the quality or characteristics of virtual acoustics?
What is the car interior acoustics impact on the perceived (virtual) sound attributes?
Research topics Multi-channel data acquisition
Measurement error compensation
Multi-dimensional signal processing
Objective analysis of car interior acoustics
19.06.2021 [email protected] 13
IntroductionPart I: Building the data base
• In-situ analysis of a sound field reproduction method requires a dense grid ofdata points across the reproduction area (spatial sampling theorem!)
• Measurement grid 57x60 positions, 2cm sampling distance
3420 RIRs per loudspeaker (62)
3 BRIR sets per loudspeaker
19.06.2021 [email protected] 14
Processing steps to get a valid data set
array position
IntroductionPart I: Wave Field Analysis Toolbox
Multi-dimensional Fourier Transform Spatial decomposition of sound events
Analysis in spatial-frequency domain
Attenuation of specific directions of arrival
Extrapolation from 2D plane into 3D space
19.06.2021 [email protected] 15
Fourier Domain – Spatial Filtering
• 2D mapping onto kx-ky cone• 3D mapping onto kxkykz-sphere
IntroductionPart I: Wave Field Analysis Toolbox
Application example Sound field of right side loudspeaker 𝑝(𝑥, 𝑦, 𝑡)
Wave front is invoking a significant artefact
Where does this origin from?
Processing 3D-FT into spatial-frequecy domain 𝑃(𝑘𝑥, 𝑘𝑦, 𝑤)
Apply spatial filter to suppress horizontal waves
Back-transformation into original domain
Now, just components arriving from above are included
Artefact origin identified as upper door handle (roof)
19.06.2021 [email protected] 16
Current researchOverview Part#2
19.06.2021 [email protected] 18
Perceptual attributes & closing the optimization loop Which cues are supportive to create the illusion of an acoustic space,
larger than the car interior cabin?
Which sound field parameters are linked to such perceptual cues?
What are the objective and subjective analysis methods, suitable to assess the criteria mentioned above?
How to influence those properties on a target system (optimization-loop)?
Workflow What sounds good? target definition, parameter selection
Link perceptual and physical domain stability, repeatability
Test specific method(s) to reproduce/control transferability
focus = virtual room acoustics
Current researchRoom acoustics spatial perception
Searching for:
• (virtual) room acoustics
• room size/extent
• envelopment
• immersion
• perception of room boundaries,
• dissolving (reproduction) room boundaries,
• room-in-room,
• auditory model,
…
19.06.2021 [email protected] 19
Concert hall acousticsmusical performance
Artistical aspects
Virtual AcousticsVAE
VR/AR, Games, Entertainment, Education
Maths/Physicssignal processing, numerical
simulation, optimization, machine learning, …
Psycho-acoustics, Auditory modeling, Hearing,
Audiology, Medical aspects
Car audio / interioracoustics, Industrial aspects,
measurement based, subjective/sensory
evaluation
Current researchRoom acoustics spatial perception
Most prominent measures for spaciousness are based on: Interaural time difference (ITD)
Interaural level difference (ILD)
Interaural cross correlation (IACC)
Lateral engery fraction (LF)
Corresponding perceptual attributes: Apparent source width (ASW)
Listener envelopment (LEV)
Additional factors of influence: room-in-room perception, proximity/echolocation, visual context
19.06.2021 [email protected] 20
room-in-room acoustics situation [Strauss]
Map of perceptual attributes [Kaplanis2017]
Current researchRoom acoustics spatial perception
Apparent Source Width (ASW)
» Source perceived wider than visual/physicalsize of source [Keet1968]
» Spatially and teporally fused auditory image of the original sound and early reflections [Kaplanis2017]
» Related physical parameters:
(early) Lateral energy fraction (LF) and
Early Interaural CC (IACC_e) [ISO3382]
» Relevant f-bands and reliability of IACC under discussion [Witew2010, Beranek2008]
» In concert halls, high ASW seen as beneficial [Beranek2008]
» Listener Envelopment (LEV)
» Feeling of beeing surrounded by sound
» Mainly related to late lateral energy[Bradley and Soulodre, 1995]
» Related physical parameters
Late lateral sound level (GLL) and
Late Interaural CC (IACC_l) [ISO3382]
» Principal cue are ITD fluctuations (asymetriclateral and median room modes, fluctuation freq. ~ 3 - 20Hz) [Griesinger1997]
» Strength of late arriving energy plays a role [Furuya2005, Soulodre2003a]
» Fully de-correlated reverberation causes maximum envelopment [Blauert1996, Griesinger1997]
19.06.2021 [email protected] 21
Current researchAuditory models
Auditory Modeling for Assessing Room Acoustics
- Described in [van Dorp Schuitman2010]- based on [Dau1996], extended by [Breebaart2001]
Model features:- foreground stream: direct sound- background stream: reverberant sound- monoaural output:
room attributes reverberance and clarity (S_REV, S_CLA)- binauraul output:
spaciousness related (S_ASW, S_LEV)
Stage #1: Peripheral processor:- Ear canal = bandpass, Cochlea = Gammatone filters- Hair cells = half-wave rectification- Neural adaptors (5 feedback loops)
Stage #2: Binaural and Central processor (+ Parameter calculation)- Split into direct and reverberant stream (peak detection, per channel)- Binaural: ITD calculated for each f-band, late/early values calculated using direct/reverberant streams- Central: combines peripheral and binaural processor (running ITD) outputs + calculates the 4 objective parameters
19.06.2021 [email protected] 22
AMARA, processing blocks of auditory model [van Dorp Schuitman2010]
Current researchAuditory models
Room Acoustical Perception
- Described in [Klockgether2014]- Psychoacoustic experiments showed that listeners may have difficulties
to segregate between source/room stream- Model takes into account human insensitivity to changes in IACC [Klockgether2013]
Stage #1: Extract binaural cues from stereo input signalPeripheral processing- separately for left/right ear channel, frequency processing of inner ear- Gammatone filter bank (42 filters, 100Hz - 10kHz), half-wave-rectification,
low-pass filter (4th order Butterworth)Binaural processing- partitioning (40ms-blocks), ICC for each block and filter (weighted around zero with
Gaussian distribution, sd of 250us)- ILD calculation for each block and filter, then accumulation using random weighting factors- ICC are accumulated, point of gravity is ITD estimate [Stern&Shear1996]
Stage #2: Predict perceptual attributes (ASW, LEV)- Energy ratio determines direct and diffuse part- Modified ICC (power of 4) to model human insensitivity at low ICC values!- ICC weighted by Energy ratio
19.06.2021 [email protected] 23
RAP model [Klockgether2014]
ExperimentsVirtual room acoustics – Test conditions
- Car audio system model = Binaural database from Part #1
- Virtual room acoustics rendered by Wave Field Synthesis audio system
- Virtual Stereo configuration, playback of Stereo audio material
19.06.2021 [email protected] 25
Condition Number ofoptions
Description
Seating position 2 positions Driver (front left)Backseat (back right)
Room acoustics 4 rooms Big hall (T60 > 1.2s)Large room (T60 > 0.6s)Small room ( T60 > 0.3s)No-room (T60 ~ 0.05s)
Loudspeakers 2 configurations Midrange + tweeters onlyFull system (incl. sub-woofers)
Stimuli 4 items Acoustic guitar solo (dry recording)Pop song1 (strong rhythm section)Pop song2 (synthesizers, bass)Song (female voice and piano)
ExperimentsAMARA
Virtual room acoustics in car compartment
• S_ASW: relatively constant across conditions
• S_CLA: significant impact of sub-woofers
• S_LEV: closely following room size, also detecting the sub-woofer condition
• S_REV: decreasing with room size, but „no-room“ detected to be more reverberantthan „small room“
Need to study in-depth!
19.06.2021 [email protected] 26
AMARA parameters @Driver seat: All conditions, 1 stimulus (pop-synth)
ExperimentsAMARA
Virtual room acoustics in car compartment
• Comparison between driver and back seatfor 1 single stimulus (pop song2)
• Relative difference calculated as
𝑣𝑎𝑙𝑢𝑒@𝑏𝑎𝑐𝑘 − 𝑣𝑎𝑙𝑢𝑒@𝑑𝑟𝑖𝑣𝑒𝑟
𝑣𝑎𝑙𝑢𝑒@𝑑𝑟𝑖𝑣𝑒𝑟
• S_CLA: significantly higher on back seat
• All other parameters lower at back seat
Need to study in-depth!
19.06.2021 [email protected] 27
AMARA parameters: relative difference due to seating positions, 1 stimulus
ExperimentsAMARA
Virtual room acoustics in car compartment
19.06.2021 [email protected] 28
Frequency [Hz]
Frequency [Hz]Frequency [Hz]
Subwoofer: 1st cabin mode @46Hz
SPL across car cabin:
front
back
front row
back row
right
left
right
left
Impact on decay time (T10) across cabin:
right
left
T10 [s]
0.2 s
0.4 s
ASW (coloured) and S_ASW (grey) on driver seat (front-left) LEV (coloured) and S_LEV (grey) on driver seat
ExperimentsRAP vs. AMARA
Comparison:ASW and LEV spaciousness attributes @driver seat
19.06.2021 [email protected] 29
Audio examples for download!! Link shared via Zoom chat
ExperimentsStimulus
Any way to standardize the input signal? • Comparison of music signal (pop song) vs. pink burst signal (500ms, windowed)
• Virtual room acoustics, different virtual source positions (left/right Stereo source position)
• Binaural recording at different seating positions
Can „pink burst“ stimuli deliver more stable results (smaller variance)?
19.06.2021 [email protected] 30
music pink burst
vQ: left right left right
------- driver seat ------- -------- back seat ---------
music pink burst
left right left rightvQ:
------- driver seat ------- -------- back seat ---------
Used in automotive audio qualityassessment [e.g. Azzali, Farina]
ExperimentsOpen: Design of the control loop
How to integrate perceptual attributes into the control loop?
• Estimation of actual perceptual attributes (e.g. ASW, LEV)
• a) Binaural optimization Target sound Filters
• b) Use microphone signals
• c) Binaural to microphone processing
• Manipulation of IACC [Grosse2014]
• Control frameworks [Brännmark2013, 2015]
19.06.2021 [email protected] 31
Control loop concept is currently at the stage of „thoughts on paper“
ExperimentsNext steps
• Define test conditions for a structured analysis of auditory model performance
• Identify model deficiencies and accompany by in-depth analysis of acousticparameters• Influence of stimuli, adjustment of model parameters
• Go in-depth with further aspects of very small reproduction rooms:• Room-in-room perception
• Acoustic proximity/echolocation
• Integration of auditory assessment into control-loop• binaural-to-microphone, usage of non-binaural microphone measurements
• If required: tuning of auditory models or find other perceptual attributes
• Prepare for a subjective test (validate perceptual relevance of findings)
19.06.2021 [email protected] 32
ReferencesPhase II
19.06.2021 [email protected] 34
[Keet 1968] Keet W.d.V., “The influence of early lateral reflections on spatial impression". In Proceedings of the 6th International Congress on Acoustics (ICA 68), Tokyo, Japan, pages E53-E56.
[Kaplanis2014] Kaplanis N., Bech S., Jensen S. H. & van Waterschoot T., “Perception of reverberation in small rooms: a literature study”. Audio Engineering Society, 55th International Conference, 2014.
[ISO 3382] ISO (2009). \ISO 3382-1:2009: Acoustics - Measurement of room acoustic parameters - Part 1: Performance spaces". International Organization for Standardization.
[Witew2010] Witew I.B., Lindau A., Van Dorp Schuitman J., Vorländer M., Weinzierl S., De Vries D., “Uncertainties of IACC related to dummy head orientation". In Proceedings of DAGA 2010, Berlin, Germany
[Beranek2008] Beranek L. L., “Concert hall acoustics - 2008". Journal of the Audio Engineering Society, 56(7/8), pp. 532-544.
[Bradley and Soulodre1995] Bradley J. S., Soulodre G. A., “Objective measures of listener envelopment". Journal of the Acoustical Society of America, 98(5), pp. 2590-2597.
[Griesinger1997] Griesinger D., “The psychoacoustics of apparent source width, spaciousness and envelopment in performance spaces". Acta Acustica united with Acustica, 83(4), pp. 721-731.
[Furuya2005] Furuya H., Fujimoto K., Wakuda A. and Nakano Y., “The influence of total and directional energy of late sound on listener envelopment". Acoustic Science and Technology, 26(2), pp. 208-211.
[Soulodre2003] Soulodre G. A., Lavoie M. C. and Norcross S. G. “Objective measures of listener envelopment in multichannel surround systems". Journal of the Audio Engineering Society, 51(9), pp. 826-840.
[Blauert1996] Blauert, J. (1996). Spatial hearing. The MIT Press, Cambridge, Massachusetts, revised edition.
[Griesinger1997a]Griesinger D., “Spatial impression and envelopment in small rooms”. Presented at AES 103rd Convention, New York, USA. Preprint 4638, 1997.
[Dau1996] Dau, T., P•uschel, D. and Kohlrausch, A. (1996a). “A quantitative model of the “effective" signal processing in the auditory system. I. Model structure". JASA, 99(6), pp. 3615-3622.
[Breebaart2001] Breebaart D. J., Modeling binaural signal detection. PhD thesis, Eindhoven University of Technology, Eindhoven. 2001.
[Klockgether2014] Klockgether, S. & van de Par, S. A Model for the Prediction of Room Acoustical Perception Based on the Just Noticeable Differences of Spatial Perception Acta Acustica united with Acustica, 2014, 100, 964-971
[Brannmark2015] Brännmark L.-J., Sternad M., Controlling the impulse responses and the spatial variability in digital loudspeaker-room correction, International Symposium on Electro Acoustic Technologies – ISEAT, Shenzen, China, 2015
[Brännmark2013]Brännmark L.-J., Bahne A., Sternad M., Compensation of loudspeaker-room responses in a robust MIMO control framework. IEEE Trans. Audio, Speech, and Lang. Process., 21(6):1201-1216, 2013.
ReferencesPhase I
19.06.2021 [email protected] 35
[2007 ICA 19] Strauß M., Munderloh M., "Influence of Loudspeaker Displacement on the Reproduction Quality of Wave Field Synthesis Systems", 19th Int. Congress on Acoustics, Madrid, 2007.
[2008 DAGA 33] Strauß M., Zhykhar A., Heeg S., "Einsatz von Mikrofonarrays zur Schallfeldanalyse in Fahrzeugen", 34. Jahrestagung für Akustik DAGA, Dresden, Germany, 2008.
[2008 AES 125Conv] Strauß M., de Vries D., "Application of Multichannel Impulse Response Measurement and Analysis to Automotive Audio", AES 125th Convention, San Francisco, USA, 2008.
[2009 DAGA 35] Strauß M., Nowak H., de Vries D., "Analysis of Enclosed Sound Fields using Multichannel Impulse Response Measurement", invited paper 35. Jahrestagung für Akustik DAGA, Rotterdam, The Netherlands, 2009.
[2009 AES 36Conf] Strauß M., Nowak H., de Vries D., "Approach to Sound Field Analysis and Simulation inside a Car Cabin”, AES 36th Int. Conference, Dearborn, USA, 2009
[2011 AES 131Conv] Strauß M., Gleim P., "Application of Wave Field Synthesis and Analysis on Automotive Audio", AES 131th Convention, New York, 2011.
[2012 AES 48Conf] Strauß M., Nowak H., " Sound Field Reproduction Analysis in a Car Cabin Based on Microphone Array Measurements”, AES 48th Int. Conference, Munich, Germany, 2012.
[2016 IEEE Spectrum] Strauß, M. et al, Carver L., "Virtually Tuning an Automotive Audio System”, Multiphysics Simulation, Insert to IEEE Spectrum, September 2016.
[2019 IEEE Spectrum] Strauß, M. et al, Forrister T., "Interactive Product Development – Simulation Applications Help Shape the Design of Car Audio Systems”, Multiphysics Simulation, Insert to IEEE Spectrum, December 2019.
[2019 AES AutoConf] Strauß M., Malbos F., Bogdanski M., "Virtual Reality Experience for the Optimization of a Car Audio System”, AES Int. Conference on Automotive Audio, Ingolstadt, Germany, 2019.