A brief on synthesis of 3D sound and externalization in headphones

1. Aditya Srinivas
ECE 491

2.  Natural vs Virtual Spatial Hearing
 Applications
 Measured HRTFs Approach
 Constructing filters based on Measured HRTF
 Synthetic HRTFs
 Problem of In-Head Localization
 Externalization
◦ Reverberation
◦ Decorrelation

4.  Process of widening stereo image of a sound and create an
illusion of sound in three dimensional space.
 Uses attributes that complement or replace the spatial
attributes that existed originally with a given sound source.

5.  Immersive environments
 Hearing Aids
 Non Speech Audio inputs
 Representational Sounds

6.  Captures the frequency dependent amplitude and time delay
differences that result from the head, torso and the complex
shape of the Pinna.
 Amplitude difference of signal between the two ears is called
the Inter-aural Level Difference(ILD) and the time difference is
called the Inter-aural Time Difference(ITD).

9.  HRTFs captured using microphones placed
inside the ear of a KEMAR manikin for
different spatial locations.
 y(nT) = x(nT) * h(nT)
◦ x(nT) = monophonic audio sample
◦ h(nT) = HRTF
◦ y(nT) = localized audio sample

10.  Non-individualized HRTFs cause localization errors
due to unique pinna features.
 Limitations in measurement.
 Limitations to simulate a moving sound source.
 Interpolating algorithms have varying degree of
success.

11.  Using bandpass filters and delays to localize sound in
space using measured HRTFs as reference.
 Easy to change the texture and timbre of the audio
sample by manually modifying the frequency
components.
 Easier to simulate moving sound sources.
 The process of localizing the sound at a particular point
is intricate and time consuming.
original Measured Synthetic
(e10 a250)

13.  An easier approach is to model the filtering
characteristics of the head and the pinna.

14. 

15. Elevation 0 Elevation 30 Elevation 60

17.  Reverberation
 Decorrelation

18.  The process of echoes producing
reverberation
 The Unit Impulse Response Function of each
echo.
 Magnitude of each Unit Impulse Response

19. Visualizing the Process

20. The Unit Impulse Response Function

21. Distance to each virtual source
Unit Impulse Response for each virtual source

22. Magnitudes of Unit Impulse Response
 Distance dependence
 Number of Reflections
 Total Magnitude of each echo

23. The Impulse Response
No Reverb
With
Reverb

24.  “refers to a process where the audio source
signal is transformed into multiple output
signals with waveforms that appear different
from each other but, which sound the same
as the source.”
 Effects of Decorrelation:
◦ Produces diffused sound fields
◦ Prevents image shift
◦ Prevents the Precedence Effect
◦ Reduces combing

25.  The cross Correlation function determines the
correlation measure
 Correlation measure
◦ (+1)  Identical Signals
◦ (-1)  Signals are out of phase
◦ (0)  Signals are dissimilar

26.  Building Decorrelation filters

27.  The convolution operation is equivalent to a
FIR filter and the exemplar signals as its
coefficients.
 Correlation measure determined by the
correlation of the filter coefficients.
 FIR filter is made all pass by keeping the
magnitude specification as unity.

28.  The phase is constructed by combination of
random number sequences.
 Correlation measure of the output signals will
be dependent on correlation measure of the
random number sequences.

29.  H = A exp^(j*phi)

30.  Samples
CM = -0.9 CM = -0.5 CM = 0.0 CM = +0.5
CM = +0.99
Original

31.  Reverb + Decorrelation samples
CM -0.9(3) CM - CM 0.0(1)
0.3(4)
CM +0.99(2) CM +0.5(5)

32.  Demonstration Sample

33.  Improve frontal localization
 Phase issues
 Improve distance perception
 Alternative methods of HRTF measurement
and Externalization

34. 1. Begault, D.R.: 3D Sound for Virtual Reality and Multimedia.
Academic Press, Cambridge (1994)
2. Allen, J. & Berkley, D.: Image method for efficiently simulating
small room acoustics, J. Acoust. Soc. Am., Vol 65, No. 4, April
1979
3. Kendell, G.S.: The Decorrelation of Audio Signals and Its Impact
on Spatial Imagery. Computer Music Journal, Vol. 19, No. 4.
(Winter, 1995), pp. 71-87
4. McGovern, S.G.: A Model for Room Acoustics
5. Hartmann W.M. & Wittenberg A.: On the Externalization of
Sound Images
6. Brown P. & Duda R.: An Efficient HRTF Model for 3D Sound

35. 7. Freeland, Diniz, Biscainho: Using Interpositional Transfer Functions in
3D Sound.
8. http://www.audiologyonline.com/news/news_detail.asp?news_id=6
9. Torrez, Petraglia: HRTF Interpolation in the Wavelet Transform Domain.
10. Gardner B. & Martin K.: HRTFs Measurement of a KEMAR Dummy Head
Microphone” MIT Media Lab personal Computing – Technical Report
#280 May, 1994
11. J. Blauert. Spatial Hearing. MIT Press, Cambridge, MA, 1983.
12. Johansson, P.: Sound Externalization, Luleå University of Technology
13. Wenzel, E.M., Arruda, M., Kistler, D.J., Wightman, F.L.: Localization
using nonindividualized head-related transfer functions. J. Acoust. Soc.
Am. 94(1), 111–123 (1993)