acoustic localization by interaural level difference
DESCRIPTION
Acoustic Localization by Interaural Level Difference. Rajitha Gangishetty. d. q. sound source. f. compact microphone array. Acoustic Localization. Acoustic Localization: Determining the location of a sound source by comparing the signals received by an array of microphones. - PowerPoint PPT PresentationTRANSCRIPT
Acoustic Localization by Interaural Level Difference
Rajitha Gangishetty
7/14/2005 Acoustic Localization by ILD 2
Acoustic Localization
compactmicrophone
array
sound source
d
Acoustic Localization: Determining the location of a sound source by comparing the signals received by an array of microphones.
Issues: reverberation noise
7/14/2005 Acoustic Localization by ILD 3
Overview
• What is Interaural Level Difference (ILD)?
• ILD Formulation• ILD Localization• Simulation Results• Conclusion and Future Work
7/14/2005 Acoustic Localization by ILD 4
Techniques
• Interaural time difference (ITD): relative time shift
ITD
ILD
sound source
microphones
• Interaural level difference (ILD): relative energy level
All previous methods (TDE, beamforming, etc.) use ITD alone.
7/14/2005 Acoustic Localization by ILD 5
Previous Work• Time Delay Estimation
[M. S. Brandstein, H. F. Silverman, ICASSP 1997; P. Svaizer, M. Matassoni, M. Omologo, ICASSP 1997]
• BeamformingJ. L. Flanagan, J.D. Johnston, R. Zahn, JASA 1985;R. Duraiswami, D. Zotkin, L.Davis, ICASSP 2001]
• Accumulated Correlation[Stanley T. Birchfield, EUSIPCO 2004]
• Microphone arrays [Michael S. Brandstein, Harvey F. Silverman, ICASSP 1995; P. Svaizer, M. Matassoni, M. Omologo, ICASSP 1997]
• Hilbert Envelope Approach[David R. Fischell, Cecil H. Coker, ICASSP 1984]
7/14/2005 Acoustic Localization by ILD 6
A sneak peek at the resultsLikelihood plots, Estimation error, Comparison of different
approacheslikelihood function computed by
horizontal and vertical microphone pairs
contour plots of likelihood functions (overlaid and combined)
microphones true location
7/14/2005 Acoustic Localization by ILD 7
ILD Formulation• N microphones and a source signal s(t)
• Signal received by the i th microphone
di = distance from source to the ith microphone = additive white Gaussian noise
• Energy received by i th microphone
7/14/2005 Acoustic Localization by ILD 8
ILD Formulation• For 2 mics the relation between energies and distances is
• Given E1 and E2 the sound source lies on a locus of points (a circle or line) described by
where,
7/14/2005 Acoustic Localization by ILD 9
ILD Formulation• For E1 ≠ E2 the equation becomes
which is a circle with center and radius
In 3D the circle becomes a sphere
• For E1= E2 the equation becomes
which becomes a plane in 3D
7/14/2005 Acoustic Localization by ILD 10
Isocontours for 10log(delta E)
7/14/2005 Acoustic Localization by ILD 11
ILD Localization
• With only two microphones source is constrained to lie on a curve
• The microphones cannot pinpoint the sound source location
• We use multiple microphone pairs• The intersection of the curves yield
the sound source location
Why multiple microphone pairs?
7/14/2005 Acoustic Localization by ILD 12
Combined Likelihood Approach
• Then the estimate for the energy ratio Then the estimate for the energy ratio at at candidate location candidate location isis
• Define the energy ratio asDefine the energy ratio as
where is the location of the where is the location of the ith microphoneith microphone
• is treated as a Gaussian is treated as a Gaussian random variablerandom variable
• Joint probability from multiple microphoneJoint probability from multiple microphone pairs is computed by combining the pairs is computed by combining the individual log likelihoodsindividual log likelihoods
Localize sound source by computing likelihood at a number Localize sound source by computing likelihood at a number of candidate locations:of candidate locations:
7/14/2005 Acoustic Localization by ILD 13
Hilbert Transform• The Hilbert transform returns a complex
sequence, from a real data sequence.
• The complex signal x = xr + i*xi has a real part, xr, which is the original data, and an imaginary part, xi, which contains the Hilbert transform.
• The imaginary part is a version of the original real sequence with a 90° phase shift.
• Sines are therefore transformed to cosines and vice versa.
7/14/2005 Acoustic Localization by ILD 14
Hilbert Transformer
xr[n]Complex Signal x[n]
Hilbert Transformer
h[n]
xr[n]
xi[n]
-j , 0<w<pi
j , -pi<w<0
H(ejw) =
where ‘w’ is the angular frequency
The Hilbert transformed series has the same The Hilbert transformed series has the same amplitude and frequency content as the original amplitude and frequency content as the original real data and includes phase information that real data and includes phase information that depends on the phase of the original data.depends on the phase of the original data.
In Frequency domain, Xi(ejw) = H(ejw)Xr(ejw)
7/14/2005 Acoustic Localization by ILD 15
Hilbert Envelope Approach
• All-pass filter circuit produces two signals with equal amplitude but 90 degrees out of phase.
• Square root of the sum of squares is taken.
Input Square
Square
SumSquare
R oot
H ilbert Speech Envelope
phase splitter
(0 o )
(90o)
(90o)2
(0o)2
7/14/2005 Acoustic Localization by ILD 16
Simulated Room
7/14/2005 Acoustic Localization by ILD 17
Simulation Results
The algorithm
• Accurately estimates the angle to the sound source in some scenarios
• Exhibits bias toward far locations (unable to reliably estimate the distance to the sound source)
• Is sensitive to noise and reverberation
7/14/2005 Acoustic Localization by ILD 18
Results of delta E Estimation
The estimation is highly dependent upon the
• sound source location
• amount of reverberation
• amount of noise
• size of the room
• relative positions of source and microphones
7/14/2005 Acoustic Localization by ILD 19
Likelihood plots5x5 m room, theta = 45 deg , no noise, no reverberation, d = 2m
7/14/2005 Acoustic Localization by ILD 20
Likelihood plots5x5 m room, theta = 90 deg , SNR = 0db, reflection coefficient = 9, d
= 2m
7/14/2005 Acoustic Localization by ILD 21
Likelihood plots5x5 m room, theta = 0 deg , SNR = 0db, reflection coefficient = 9, d =
1m
7/14/2005 Acoustic Localization by ILD 22
Likelihood plots10x10 m room, theta = 0 deg , SNR = 0db, reflection coefficient = 9, d
= 1m
angle error = 6.5 degrees
7/14/2005 Acoustic Localization by ILD 23
Likelihood plots5x5 m room, theta = 36 deg , SNR = 0db, reflection coefficient = 9, d =
2m
angle error = 9 degrees
7/14/2005 Acoustic Localization by ILD 24
Angle Errors in a 5x5 m room
d = 1m, only noise d = 2m, only noise
d = 1m, only reverberation
d = 2m, only reverberation
0.7 = solid line, blue
0.8 = dotted, red
0.9 = dashed, green
20 dB = solid line, blue
10 dB = dotted, red
0 dB = dashed, green
7/14/2005 Acoustic Localization by ILD 25
Angle error in degrees for the 10x10 m room when the source is at a distance of 1m
Angle error in degrees for the 5x5 m room when the source is at a distance of 1m
7/14/2005 Acoustic Localization by ILD 26
Angle error in degrees for the 10x10 m room when the source is at a distance of 2m
Angle error in degrees for the 5x5 m room when the source is at a distance of 2m
7/14/2005 Acoustic Localization by ILD 27
Comparison of errors with the Hilbert Envelope Approach in a 5x5 m room
0.7
0.8
0.9
20 dB 10 dB 0 dBWithout Hilbert = solid line, blue
Matlab Hilbert = dotted line, red
Kaiser Hilbert = dashed line, green
Reflection coefficient
7/14/2005 Acoustic Localization by ILD 28
Comparison of errors with the Hilbert Envelope Approach in a 10x10 m room
0.7
0.8
0.9
20 dB 10 dB 0 dBWithout Hilbert = solid line, blue
Matlab Hilbert = dotted line, red
Kaiser Hilbert = dashed line, green
Reflection coefficient
7/14/2005 Acoustic Localization by ILD 29
Likelihood plots without Hilbert Envelope
angle error = 27 degrees
5x5 m room, theta = 18 deg , SNR = 0db, reflection coefficient = 9, d = 2m
7/14/2005 Acoustic Localization by ILD 30
Frames approach
• Signal divided into 50 frames• Frame size = 92.8ms• 50% overlap in each frame
5x5 m room, theta = 18 deg , SNR = 0db, reflection coefficient = 9, d = 2m (left), d = 1m (right)
Mean error = 15 deg
Std Dev = 11 deg
Mean error = 7 deg
Std Dev = 6 deg
7/14/2005 Acoustic Localization by ILD 31
Conclusion and Future Work•ILD is an important cue for acoustic localization ILD is an important cue for acoustic localization
• Preliminary results indicate potential for ILD Preliminary results indicate potential for ILD (Algorithm yields accurate results for several (Algorithm yields accurate results for several configurations, even with noise and reverberation) configurations, even with noise and reverberation)
• Future work:Future work:• Investigate issues (e.g., bias toward distantInvestigate issues (e.g., bias toward distant locations, sensitivity to reverberation) locations, sensitivity to reverberation)• Experiment in real environmentsExperiment in real environments• Investigate ILDs in the case of occlusionInvestigate ILDs in the case of occlusion• Combine with ITD to yield more robust resultsCombine with ITD to yield more robust results
7/14/2005 Acoustic Localization by ILD 32