head and ear
play

Head and Ear Thurs. March 29, 2018 1 Impulse function at = 0. , , - PowerPoint PPT Presentation

COMP 546 Lecture 20 Head and Ear Thurs. March 29, 2018 1 Impulse function at = 0. , , , = ( 0 , 0 , 0 , ) To define an impulse function properly in a continuous space


  1. COMP 546 Lecture 20 Head and Ear Thurs. March 29, 2018 1

  2. Impulse function at 𝑒 = 0. 𝐽 π‘Œ, 𝑍, π‘Ž, 𝑒 = πœ€(π‘Œ βˆ’ π‘Œ 0 , 𝑍 βˆ’ 𝑍 0 , π‘Ž βˆ’ π‘Ž 0 , 𝑒) To define an impulse function properly in a continuous space requires more math. Let’s not spend our time doing that, since we just want qualitative behavior here. Sound obeys the wave equation. So, how is this function defined 𝑒 β‰  0 ? 2

  3. Impulse becomes expanding sphere One can show that this follows from the wave equation. 𝑒 = 4 Ξ” 𝑒 𝑠 = 𝑀 𝑒 𝑒 = 3 Ξ” 𝑒 𝑒 = 2 Ξ” 𝑒 𝑒 = Ξ” 𝑒 3

  4. Impulse sound energy is spread over a thin sphere of fixed thickness and of area 4𝜌 𝑠 2 where 𝑠 2 = (π‘Œ βˆ’ π‘Œ 0 ) 2 + (𝑍 βˆ’ 𝑍 0 ) 2 + (π‘Ž βˆ’ π‘Ž 0 ) 2 . 𝑠 = 𝑀 𝑒 1 1 𝐽 2 ~ 𝐽 ~ So, SPL 𝑠 2 𝑠 4

  5. 𝐽 π‘Œ, 𝑍, π‘Ž, 𝑒 𝐽 𝑑𝑠𝑑 πœ€(π‘Œ βˆ’ π‘Œ 0 , 𝑍 βˆ’ 𝑍 0 , π‘Ž βˆ’ π‘Ž 0 ) , when 𝑒 = 0 = 𝐽 𝑑𝑠𝑑 πœ€ 𝑠 βˆ’ 𝑀 𝑒 , when 𝑒 > 0 and 𝑠 𝑠 = (π‘Œ βˆ’ π‘Œ 0 ) 2 + (𝑍 βˆ’ 𝑍 0 ) 2 + (π‘Ž βˆ’ π‘Ž 0 ) 2 𝐽 𝑑𝑠𝑑 is constant (~energy in impulse) 5

  6. We can write a general sound source a sum of impulse functions: π‘ˆβˆ’1 πœ€ 𝑒 βˆ’ 𝑒 β€² 𝐽 𝑑𝑠𝑑 (𝑒 β€² ) 𝐽 𝑑𝑠𝑑 𝑒 = 𝑒 β€² =0 6

  7. Far from the source, where r is large, the wavefront is approximately locally planar. 7

  8. Binaural hearing (preview of next lecture) If the sound arrives from the left (assuming planar wavefronts), what is the interaural delay? 𝑒 = 𝑒 .17 𝑀 = 340 β‰ˆ .5 𝑛𝑑 𝑒 = 17 cm 8

  9. Naïve model: cone of confusion Model head, shoulders, ears as a sphere. All incoming directions on a cone define the same delay & shadow effect. Exercise: use time delay 𝜐 to estimate cone angle 𝜚 9

  10. Interaural differences How can the auditory system estimate the delay and shadowing ? Here is a simple model: 𝐽 π‘š ( 𝑒 ) = 𝛽 𝐽 𝑠 ( 𝑒 βˆ’ 𝜐 ) + π‘œ(𝑒) noise shadow delay (attenuation) 10

  11. Maximum likelihood: find the 𝛽 and 𝜐 that minimize π‘ˆ { 𝐽 π‘š ( 𝑒 ) βˆ’ 𝛽 𝐽 𝑠 ( 𝑒 βˆ’ 𝜐 ) } 2 𝑒=1 where 𝜐 < 0.5 𝑛𝑑 . 11

  12. To find the 𝛽 and 𝜐 that minimize π‘ˆ {𝐽 π‘š ( 𝑒 ) 2 βˆ’ 𝛽 𝐽 π‘š ( 𝑒 ) 𝐽 𝑠 ( 𝑒 βˆ’ 𝜐 ) + 𝐽 𝑠 ( 𝑒 βˆ’ 𝜐 ) 2 } 𝑒=1 we first find the 𝜐 that maximizes 𝐽 π‘š ( 𝑒 ) 𝐽 𝑠 ( 𝑒 βˆ’ 𝜐 ) . 𝑒 This ignores the small dependence of the 3 rd term above on 𝜐. 12

  13. Then estimate 𝛽 (shadowing): 𝐽 π‘š ( 𝑒 ) 2 π‘ˆ 𝑒=1 𝛽 2 = 𝐽 𝑠 ( 𝑒 βˆ’ 𝜐 ) 2 π‘ˆ 𝑒=1 Note that this gives two cues which we can combine. 13

  14. The Human Ear 14

  15. Outer Ear Next ten slides: How do head and outer ear transform the sound that arrives at the ear from various directions ? 15

  16. Head related impulse response (HRIR) Suppose sound is from direction ( 𝜚, πœ„ ). The wave is planar when it arrives at the head. If the source is an impulse then sound measured at the ear drum of ear 𝑗 is: 𝐽 𝑒 = β„Ž 𝑗 (𝑒; 𝜚, πœ„ ) βˆ— πœ€ 𝑠 βˆ’ 𝑀𝑒 left or right 16

  17. Sound source 𝐽 𝑑𝑠𝑑 𝑒; 𝜚, πœ„ transformed Suppose sound is from direction ( 𝜚, πœ„ ) and emits 𝐽 𝑑𝑠𝑑 𝑒; 𝜚, πœ„ . Then the sound measured at the ear drum of ear 𝑗 is: 𝐽 𝑒 = β„Ž 𝑗 (𝑒; 𝜚, πœ„ ) βˆ— 𝐽 𝑑𝑠𝑑 𝑒; 𝜚, πœ„ (Ignoring time delay from source to ear.) 17

  18. KEMAR mannequin In following slides, I will show HRIR measurements β„Ž 𝑗 (𝑒; 𝜚, πœ„ ) . azimuth πœ„ elevation 𝜚 18

  19. Azimuth πœ„ (Elevation 𝜚 = 0) Suppose sound is measured at right ear drum. 19

  20. 0.7 ms HRIR Source direction (azimuth) 20

  21. Arrival time differences are not as significant when azimuth = 0 and elevation is varied. HRIR Source direction (elevation) 21

  22. If head is symmetric about the medial plane (left/right), then : β„Ž π‘šπ‘“π‘”π‘’ (𝑒; 𝜚, πœ„ ) = β„Ž π‘ π‘—π‘•β„Žπ‘’ (𝑒; 𝜚, βˆ’πœ„ ) azimuth πœ„ elevation 𝜚 22

  23. 𝐽 π‘ π‘—π‘•β„Žπ‘’ 𝑒; 𝜚, πœ„ = β„Ž π‘ π‘—π‘•β„Žπ‘’ (𝑒; 𝜚, πœ„ ) βˆ— 𝐽 𝑑𝑠𝑑 𝑒; 𝜚, πœ„ HRIR For each incoming sound direction ( 𝜚, πœ„ ), what is the Fourier transform with respect to variable t ? 23

  24. 𝐽 π‘ π‘—π‘•β„Žπ‘’ 𝑒; 𝜚, πœ„ = β„Ž π‘ π‘—π‘•β„Žπ‘’ (𝑒; 𝜚, πœ„ ) βˆ— 𝐽 𝑑𝑠𝑑 𝑒; 𝜚, πœ„ HRIR For each incoming sound direction ( 𝜚, πœ„ ), what is the Fourier transform with respect to t ? = 𝐽 π‘ π‘—π‘•β„Žπ‘’ πœ•; 𝜚, πœ„ β„Ž π‘ π‘—π‘•β„Žπ‘’ (πœ•; 𝜚, πœ„ ) 𝐽 𝑑𝑠𝑑 πœ•; 𝜚, πœ„ Head Related β€œTransfer Function” (HRTF) 24

  25. HRTF β„Ž π‘ π‘—π‘•β„Žπ‘’ (πœ•; πœ„, 𝜚 = 0) (plot for fixed elevation 𝜚 = 0) Shadowing effect dominates: HRTF for each frequency πœ• has a max at 90 degrees (right ear) and min at 270 degrees (left ear). πœ• 25 πœ„

  26. HRTF β„Ž π‘ π‘—π‘•β„Žπ‘’ (πœ•; πœ„ = 0, 𝜚) (plot for fixed azimuth πœ„ = 0. ) (medial plane) Curves shifted for visualization Valley is β€œ pinnal notch” (it distinguishes elevations) 26

  27. Middle Ear Ossicles (bones) β€œEar drum” 27

  28. Ossicles act as a lever, transferring vibrations from ear drum to fluid in cochlea pinna cochlea auditory canal outer middle inner 28

  29. Inner ear Vestibular apparatus Cochlea 29

  30. Cochlea (unrolled) TOP VIEW SIDE VIEW 30

  31. Cochlea (unrolled) TOP VIEW SIDE VIEW 31

  32. 𝑑 Recall vibrating string πœ• = 𝑀 Both 𝑀 and 𝑑 vary on fibres on basilar membrane. long (large L) short (small L) & loose (small c) & tense (large c) 20,000 Hz 20 Hz 32

  33. Basilar Membrane (BM) http://auditoryneuroscience.com/topics/basilar-membrane-motion-0-frequency-modulated-tone 33 http://auditoryneuroscience.com/ear/bm_motion_2

Recommend


More recommend