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

H. Wierstorf - Talk at the 129th AES Convention

H. Wierstorf - Talk at the 129th AES Convention

spatial_audio_research

March 06, 2012
Tweet

More Decks by spatial_audio_research

Other Decks in Science

Transcript

  1. Reducing Artifacts of Focused Sources in WFS Hagen Wierstorf, Matthias

    Geier and Sascha Spors Quality and Usability Lab Deutsche Telekom Laboratories Technische Universität Berlin 129th AES Convention November 06, 2010
  2. Wave Field Synthesis Huygens-Fresnel principle quantitative formulation by Rayleigh integrals

    sound field synthesis in a half-space V Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 1 / 15
  3. Wave Field Synthesis Huygens-Fresnel principle quantitative formulation by Rayleigh integrals

    sound field synthesis in a half-space V Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 1 / 15
  4. Focused Sources 0 1 2 3 -2 -1 0 1

    2 y (m) x (m) -1 -0.5 0 0.5 1 acoustic focusing based on time-reversal principle acoustic point sink as source model for WFS converging field from secondary sources to focus point diverging field from the focus point (field of a point source) Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 2 / 15
  5. Focused Sources Sampling Artifacts 1000 Hz 0 1 2 3

    -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 loudspeaker spacing 0.15 m spatial sampling artifacts occur above approx. 1100 Hz focal point without artifacts due to spatial sampling Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 3 / 15
  6. Focused Sources Sampling Artifacts 2000 Hz 0 1 2 3

    -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 loudspeaker spacing 0.15 m spatial sampling artifacts occur above approx. 1100 Hz focal point without artifacts due to spatial sampling Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 3 / 15
  7. Focused Sources Sampling Artifacts 5000 Hz 0 1 2 3

    -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 loudspeaker spacing 0.15 m spatial sampling artifacts occur above approx. 1100 Hz focal point without artifacts due to spatial sampling Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 3 / 15
  8. Focused Sources Pre-Echoes spatial sampling causes pre-echoes in time domain

    pre-echoes very critical for perception (Spors et al. 2009, Geier et al. 2010) number of pre-echoes depends on size of loudspeaker array Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 4 / 15
  9. Experiment Method array lengths: 10 m, 4 m, 1.8 m,

    0.75 m, 0.3 m, single loudspeaker reference virtual array with dynamical binaural resynthesis using the SSR 6 subjects from our Lab rating the attribute pairs few artifacts vs. many artifacts, left vs. right castanets and speech as stimuli SSR and stimuli available at our Blog: http://audio.qu.tu-berlin.de Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 5 / 15
  10. Focused Sources Pre-Echoes -5 -4 -3 -2 -1 0 0◦

    30◦ 60◦ t (ms) Listener direction ϕ R = 1 m L = 4 m wrong perception of direction perception of more than one source Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 7 / 15
  11. Focused Sources Pre-Echoes -5 -4 -3 -2 -1 0 0◦

    30◦ 60◦ t (ms) Listener direction ϕ R = 1 m L (m) 1.8 0.75 0.3 artifacts hypothesis: the shorter the array the fewer artifacts are audible position hypothesis: the perceived position depends not on the array length Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 8 / 15
  12. Focused Sources Pre-Echoes -5 -4 -3 -2 -1 0 0◦

    30◦ 60◦ t (ms) Listener direction ϕ R = 4 m L (m) 1.8 0.75 0.3 artifacts hypothesis: the shorter the array the fewer artifacts are audible position hypothesis: the perceived position depends not on the array length Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 8 / 15
  13. Results Artifacts few many ref 0.3 0.75 1.8 4 10

    Artifacts L (m) Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 9 / 15
  14. Results Position left right ref 0.3 0.75 1.8 4 0.3

    0.75 1.8 10 Position L (m) R = 1 m R = 4 m Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 10 / 15
  15. Focused Sources Truncation Artifacts L = 100 m -20 -15

    -10 -5 0 -2 -1 0 1 2 Amplitude (dB) x (m) xref = 2 m 0 1 2 3 -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 11 / 15
  16. Focused Sources Truncation Artifacts L = 4 m -20 -15

    -10 -5 0 -2 -1 0 1 2 Amplitude (dB) x (m) xref = 2 m 0 1 2 3 -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 11 / 15
  17. Focused Sources Truncation Artifacts L = 1.8 m -20 -15

    -10 -5 0 -2 -1 0 1 2 Amplitude (dB) x (m) xref = 2 m 0 1 2 3 -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 11 / 15
  18. Focused Sources Truncation Artifacts L = 0.75 m -20 -15

    -10 -5 0 -2 -1 0 1 2 Amplitude (dB) x (m) xref = 2 m 0 1 2 3 -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 11 / 15
  19. Focused Sources Truncation Artifacts L = 0.3 m -20 -15

    -10 -5 0 -2 -1 0 1 2 Amplitude (dB) x (m) xref = 2 m 0 1 2 3 -2 -1 0 1 2 y (m) x (m) -1 -0.5 0 0.5 1 Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 11 / 15
  20. Modeling truncation leads to wrong ILD cues, pre-echoes to wrong

    ITD cues modeling of the binaural perception by the Lindemann model (Lindemann 1986a) as model parameter the ones from Lindemann were used model is available at http://amtoolbox.sourceforge.net/ Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 12 / 15
  21. Results Position left right ref 0.3 0.75 1.8 4 0.3

    0.75 1.8 10 Position L (m) R = 1 m R = 4 m Model Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 13 / 15
  22. Conclusion resynthesis of focused sources in WFS is a challenging

    task spatial aliasing leads to pre-echoes, which can be reduced by using only a sub-array of loudspeakers truncation artifacts lead to wrong binaural cues and a perception of the source from the wrong direction modeling of the perceived direction shows that the precedence effect may not be considered for t < 3 ms there exists the possibility to enhance the direction perception by using image source (Oldfield et al. 2010, Caulkins et al. 2003) Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 14 / 15
  23. Thank you very much for your attention! Questions? Get example

    audio files here: http://audio.qu.tu-berlin.de/ Wierstorf, Geier, Spors Reducing Artifacts of Focused Sources in Wave Field Synthesis November 06, 2010 15 / 15