INTERPOLATING HEAD RELATED TRANSFER FUNCTIONS

1
INTERPOLATING HEAD RELATED TRANSFER FUNCTIONS

• Takanori NISHINO, Mikio IKEDA,
• Kazuya TAKEDA and Fumitada ITAKURA
• Meijo University
• Nagoya University
• Center for Integrated Acoustic Information
  Research
• 3 October 2000

2
INTRODUCTION

• Interaural Time Difference
• Interaural Level Difference
• Reverberations
• Doppler Effect

3
3D SOUND LOCALIZATION

• Head Related Transfer Functions (HRTFs)
• Depend on the source directions
• horizontal and vertical
• The 3D audio systems based on the HRTFs need the
  combination of a large number of different HRTFs

4
PREVIOUS WORK

• The use of linear interpolation
• (E. M. Wenzel and S. H. Foster, 1993)
• HRTFs were interpolated at the intervals of
  30azimuth and 36 elevation
• Up-down confusions were increased

5
PURPOSES

• To investigate
• the minimum number of measured HRTFs
  require for interpolation,
• the optimal arrangement in the horizontal and
  vertical directions

6
OVERVIEW

• HRTF measurement conditions and equipment
• Experiments
• Subjective tests
• Conclusions

7
MEASUREMENT

• Turn Table (Azimuth)
• Minimum Moving Angle 1
• Accuracy 0.3
• Arched Traverse (Elevation)
• Minimum Moving Angle 1
• Accuracy 0.3
• Range from 50 to 90

8
Measurement Conditions
Two Head-and-Torso Simulators Sampling
Frequency 44.1 kHzFilter Length 512 point
(11.6ms)Azimuth 72 (0, 5,,
355)Elevation 28 (-45, 40,, 90)
9
BACK
FRONT
10
INTERPOLATION
11
EXPERIMENTS

• Experimental Conditions
• Azimuth Intervals of 45
• Elevation (1) Equally-spaced (10,15,30,45,90
  )
• Elevation (2) Optimal (from 3 to 6 elevations)

12
(No Transcript)
13
OBJECTIVE MEASURE

• Spectral Distortion

14
RESULTS
15

• The SD is about 3 dB when 41 directions are used
• The linear method is superior to the spline
  method when the number of HRTFs is decreased
• There is no significant difference between the
  equally-spaced and the optimal cases

16
SUBJECTIVE TESTS

• HRTFs for evaluation
• Interpolation Method Linear Interpolation
• Number of Directions 41
• Azimuth 8 (0,45,,315)
• Elevation 6(-45,-30,0,,90)
• Phase Minimum phase with interpolated pure
  delay

17
SUBJECTIVE TESTS (Azimuth)
Stimulus White noise ( 200 ms )Azimuth 0,
30,, 330Elevation -15, 15Subject 4
Subjects
18
RESULTS
Measured HRTFs (63.6 Correct)
Interpolated HRTFs (57.8 Correct)

• No significant difference between the measured
  and interpolated HRTFs

19
SUBJECTIVE TESTS (Elevation)
Stimulus White noise ( 200 ms )Elevation -35,0
,35,70Azimuth 180Method The paired
comparison method (Referenced elevation is the
horizon)Subject 3 Subjects
20
RESULTS

• Interpolated HRTFs

Measured HRTFs
21
CONCLUSIONS

• Effective method of interpolating HRTF is
  investigated
• From objective tests,
• Linear interpolation is more effective
• The SD is about 3 dB when the HRTFs are
  interpolated from 41 directions
• No significant difference between the
  equally-spaced and the optimal cases

22
CONCLUSIONS

• From subjective tests,
• Good performance is obtained in azimuth
• All elevations are perceived correctly

23
HRTF DATA

• URL
• http//www.itakura.nuee.nagoya-u.ac.jp/HRTF/
• Contents
• HRTFs used in this experiment
• Over 50 subjects HRTF in the horizontal plane