See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254049458

Some research challenges of acoustic camera

Article · November 2011

DOI: 10.1109/TELFOR.2011.6143725

CITATIONS READS

6 658

1 author:

Miljko Eric
University of Belgrade
65 PUBLICATIONS   217 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced techniques for efficient use of spectrum in wireless systems View project

All content following this page was uploaded by Miljko Eric on 10 December 2014.

The user has requested enhancement of the downloaded file.

 19th Telecommunications forum TELFOR 2011 Serbia, Belgrade, November 22-24, 2011.

Some Research Challenges of Acoustic Camera

0LOMNR0(ULü

 use of rectangular planar microphone array placed

Abstract —Application of array processing principle and relatively close to the acoustic source (some wavelengths of
techniques in acoustic camera is a challenging research the highest frequency in the sound spectrum).
problem. Discussions about some of those challenges are Microphones in microphone array are placed on the
given in this paper. Theoretical background of acoustical
near-field and far-field beamforming is highlighted in this distance equal to the half of the wavelength of the highest
paper. Generic versions of near-field and far-field delay and frequency in the spectrum of the signal (Niquist criteria for
sum of beamformers are discussed. Those versions are the the spatial domain). So, since limitation in choosing of
basis for further research steps. Some experimental results of microphone distances due to their physical dimensions,
near-field beamforming are presented using realized limits the spectral bandwidth of the signal on the upper
microphone array and multichannel acquisition system.
limit, NAH is mainly applied for the low frequency range.
Keywords — acoustic camera, near-field beamforming,
spherical beamforming, acoustic holography, microphone
Spacing of microphones and size of microphone array
array, array processing determines spatial resolution. 2D Fourier transformation is
applied for the signals acquired with microphone array.
I. INTRODUCTION From this spatial spectrum and knowledge of the law of

A coustic camera appeared on the market as a

commercial product more than ten years ago and it is
now commercially available from various sources
acoustic signal propagation, acoustic pressure of the signal
source can be back restored using known Green functions.
But that process can be done just for the propagation
[1,2,3,4,5]. The idea of acoustic camera to do noise/sound planes which are parallel to the measurement plane (plane
source identification and quantification and to perform a of microphone array). In order to localize a source in 3D,
picture of acoustic environment by array processing of measurement plan (microphone array) has to span the
multidimensional acoustic signals received by microphone complete object. NAH technique is very efficient with
array and to overlay that acoustic picture to the video scenario with one point sound source, but the story is more
picture. So, it is a device with integrated microphone array complicated in multisource signal scenario and some
and digital video camera, which provides visualization of
complex preprocessing techniques have to be applied in
acoustic environment.
that case. In NAH technique size of tested object has to be
With acoustic camera it is possible to visualize the
smaller than array size. Data acquired from the plane
results of acoustic source characterization so it became an
array in NAH can be also used for near-far beamforming.
attractive commercial test instrument used not just in
Beamforming is another technique applied in acoustic
research laboratories but in industry, as well. Acoustic
camera. Acoustic beamforming has a long history. There
camera is just one phase in long history of development of
are lot of scientific papers and books related to the
microphone array technology, [6]. Possible applications of
theoretical as well as practical aspects of acoustic
acoustic camera as test equipment are nondestructive
beamforming but there are some specific aspects of its
measurements for noise/sound identification in interior and
application in acoustic camera. Two basic acoustic
exterior of vehicles [7], trains and airplanes [8,9],
beamforming principles are applied in acoustic cameras:
measurements in wind tunnels, etc. Acoustic camera can
far-filed and near-field beamforming.
also be built in complex platforms such as underwater
Far-field beamforming is applied in the first
unmanned vehicles [10], robots and robotized platforms
commercial solutions of acoustic camera. Far-field
etc. It can be also used for passive acoustical sensing in
beamforming is beamforming of plane waves, so it is
battlefield [11].
assumed that acoustic source is far from the microphone
Acoustic camera is used both for near-field as well as
far-field signal scenario. Two basic well known principles array. Acoustic beamforming provides Direction Of
applied in solutions of acoustic cameras are: Near-Filed Arrivals (DOAs) of acoustic signals, so the measurement
Acoustic Holography (NAH) [12] and beamforming of source distances are additionally needed in order to
[13,14,15,16,17]. identify the sources. In that case line of sight between
Near-field Acoustic Holography (NAH) is based on the microphone array and sound sources are required. Since
the mathematical model of the plane waves on the
microphone array does not depend on sound source
0LOMNR 0 (ULü LV QRZ ZLWK WKH School of Electrical Engineering, locations, far-field beamforming provides estimation of
University of Belgrade, Bul. kralja Aleksandra 73, 11120 Belgrade, Serbia relative sound preasure on the microphone array but not on
(phone: 381-64-3128807; e-mail: [email protected]).
the locations of the sound sources.

978-1-4577-1500-6/11/$26.00 ©2011 IEEE 1036

 Near-field beamforming is beamforming of sphere techniques. The first can be used for low and second for
waves so it is the case when the sound sources are on the high frequency range of acoustic signals.
small distances (some wavelengths) form the microphone Side-lobe levels in beamforming are a very important
array. Mathematical model of the signal on microphone fact in acoustic camera application since it directly limits
array in near filed scenario depends on the locations of the dynamic range in acoustic mapping. It is directly
sound sources, so the near-field beamforming can directly influenced by the geometry and aperture of the microphone
identify locations of acoustic sources and provide acoustic array. Bigger aperture and the bigger number of
microphones in the array means smaller side lobe levels.
picture in spatial sector of interests (DOAs in far field,
The level difference between the two sound sources that
became locations in near-filed beamforming). In literature
can be clearly identified and detected is called plot
near-field beamforming is also named focalization [18].
dynamic [7]. If the level difference between two real
Spherical beamforming [19] is a special version of sources is greater than the plot dynamic, louder sources
acoustic beamforming applied in acoustic camera. It is mask quieter sources. Side lobes can be further suppressed
based on the use of spherical microphone array. It can be by some additional processing but it is still one of the key
applied in reflective sound environment and it is used for problems in acoustic beamforming. Known leading
measurement in interior of vehicles, trains and airplanes. producers of acoustic cameras claim that dynamic range of
Advantage of spherical geometry is that spatial resolution its acoustic cameras is from 25 to 40 dB but it is not quite
is almost equal in all spatial directions. It is also possible clear how it is provided, since they reference to the their
to do some transformation of the signals acquired from non patented solutions.
spherical microphone array to provide virtual spherical Motivation for this research is related to the Serbian
array and then apply spherical beamforming. national project of Ministry of Science and Technological
In acoustic beamforming there are no limitations Development TR 32026 called "Integration and
regarding the geometry of microphone array. Microphone Harmonization of Sound Insulation in Buildings in the
array can be of any geometry but all possible microphones Context of Sustainable Housing". In this project we tend to
geometries with the same number of microphones are not apply technical solutions of acoustic camera for
development and verification of solution for sound
equally good. Ambiguity, spatial resolution and side lobes
insulation in buildings. To the author’s best knowledge
properties of acoustic beamforming are directly related to
acoustic camera is not so far used for this purpose so it
the geometry of microphone array. In known solutions of seems that it is our research challenge.
commercial acoustic cameras planar circular and volume
spherical geometry are mostly used. Planar 'star' geometry II. THEORETICAL BACKGROUND OF NEAR-FIELD AND FAR-
with non-uniform microphone distances and big FIELD BEAMFORMING
microphone size are also used for far-field measurements.
It is known that linear microphone arrays are fully A. Mathematical model of acoustic signal on
microphone array in near field scenario
ambiguous so this geometry is not used in acoustic camera.
All planar geometries of microphone array have ambiguity
of the front/back type. Volume microphone geometry is Suppose near-field signal scenario with M at the locations
generally un-ambiguous in volume, but if spacing between in 3D space denoted by set of vectors rs( m ) ; m 1, M . Time
microphones is random, then the level of the side-lobes in samples x p ( n ) of the signal on p-th microphone, in
beamforming can be significant. For ambiguity microphone array with P microphones on the locations
characterization of the antenna/microphone array an denoted by the set of vectors r p ; p 1, P , can be expressed
ambiguity function [20] can be used defined as a
in the next form:
normalized measure of co-linearity of two steering vectors M
in array manifold for different DOAs. If values of x p (n ) ¦b (m) (m)
s ( n  W (pm ) )  K p ( n ) (1)
p 0
m 1
ambiguity function is equal to one, perfect ambiguity
occurs. where b p( m ) and W (pm ) d p( m ) / v rs( m )  r p / v denote
Spatial resolution is defined as the shortest distance at attenuation and time delay of the m-th signal propagation
which two point sources can be separated. In far-field from its source location to the p-th microphone. So it is
acoustic beamforming it is equivalent to the smaller Time Of Arrival -TOA based signal model.
difference of DOAs which can be separated. From the Relative time delays can be expressed as:
theory of array processing it is well known that if the size ' W (pqm ) ' W (pm )  ' W q( m ) ( rs( m )  r p  rs( m )  rq ) / v (2)
of antenna/microphone array (aperture) is as big as
Time Difference Of Arrival TDOA model of the signal on
possible, the resolution properties are better. With the
microphone array can be formulated in spectral domain in
same number of microphones, array aperture can be
such a way:
increased with non-uniform distances between
X ( k ) A ( k )S 0 ( k )  Ș( k ) (3)
microphones, but in that case, side-lobes in beamforming
appear. It is also known that spatial resolution is frequency where X ( k ) and Ș ( k ) denote vectors of the k-th spectral
dependent, so the lower frequency means lower spatial sample of the signal and noise spectrum respectively on
resolution. Due to those properties, near field acoustic
holography NAH and beamforming are complementary

1037
microphone array, A ( k ) is a matrix whose columns are where X ( k ) and Ș ( k ) denote vectors of the k-th spectral
steering vectors whose elements have a general form: sample of the signal and noise spectrum respectively on
 j 2 S k ' W (pm ) / N microphone array, A ( k ) is a matrix whose columns are
a (pm ) b p( m ) e (4)
steering vectors whose elements have a general form:
(m)
 j 2 S k ' W cp /N
a (pm ) e (8)
Elements of the vector
1 2 M T
S 0 ( k ) [ S 0 ( k ) S 0 ( k ) ... S 0 ( k )] are the k-th
spectral samples of all M superposed acoustic signals on
the reference point (microphone) in the space.
So it can be seen that key differences of mathematical
models of near and far-field signal scenario are related to
the modeling of the propagation delays. Also,
Fig.1 Near-field signal scenario mathematical model of far-field signal scenario depends on
the level of acoustic signal on the reference point in space
Elements of the vector so far-field beamforming provides relative level of acoustic
1 2 M
S 0 ( k ) [ S 0 ( k ) S 0 ( k ) ... S 0 ( k )]T
are the k-th power at the locations of the microphone array. However,
spectral samples of all M superposed acoustic signals on mathematical model of near-field signal scenario depends
the sound source locations. It can be noticed that signal on the level of acoustic signal on the sound location so
model given by (3) depends on the signal levels and source near-field beamforming can provide direct estimation of
signal locations, regardless of far-field signal model, where the signal level at the location of acoustic source.
there is a dependence of the signal model on Directions of III. GENERIC VERSIONS OF NEAR-FIELD AND FAR-FIELD
Arrivals and signal levels in reference point (microphone) DELAY-AND-SUM BEAMFORMERS
in space, so far-field signal model is basically TDOA base
model. It can be noticed that steering vectors are frequency From the array processing point of view acoustic signals
dependent so this model describe wideband nature of has to be modeled as wideband signals, so acoustic
acoustic signals in array processing context. beamforming is broadband beamforming and it can be
basically implemented in spectral domain or in time
B. Mathematical model of acoustic signal on domain with implicitly included some transformations in
microphone array in far-field scenario spectral domain. Generic versions of near-field and far
field Delay-and-Sum beamformers are presented in the
Time samples x p ( n ) of the signal in far-field scenario figures 3 and 4. Time delays in generic versions of
beamformers can be realized in frequency domain using
on p-th microphone can be expressed as:
M
the following matrix transformation:
x p (n ) ¦s (m)
( n  ' W cp( m ) )  K p ( n ) (5) 1 M
c
xp ¦ (S (0m ) P p( m ) W H  Ș: H ) (9)
m 1
N m 1
where ' W cp( m ) ( k ( m ) ) T r p / v denote relative propagation
where x p denotes vector of time samples of the signal on
time delay of the m-th signal on the p-th microphone
related to the referent point (microphone) in the space. the p-th beamformer finger, P p( m ) denotes steering matrix
Vector k ( m ) is unit vector represent direction of arrival with spectral elements of the steering vector of the m-th
(m)
(azimuth M and elevation T (m)
) of the m-th signal and source on the p-th microphone and W H is inverse FFT
matrix.
it can be expressed as:
k (m) [(cos(T ( m ) ) sin(M ( m ) ) i x  (cos(T ( m ) ) sin(M ( m ) )) i y  (cos(T ( m ) )) i z
(6)

Fig.3. Generic version of near-field Delay-and-Sum

beamformer

IV. SOME EXPERIMENTAL RESULTS OF ACOUSTIC NEAR-

Fig.2 Far-field signal scenario FIELD BEAMFORMING
Time Difference Of Arrival TDOA model of the signal on
Experiment was perfomed using measurement setup
microphone array can be formulated in spectral domain in
presented in the figure 5. which contains circular low-cost
such a way:
microphone array with diameter 0.75 m and 8
X ( k ) A ( k )S 0 ( k )  Ș( k ) (7)
microphones in the array, multicahnnel microphone
preamplifiers and signal acquisition and lap-top for the

1038
 control of merasurement setup, collection of signal samples performances of the near-field beamforming regarding
and off-line implementation of beamformers More detatils spatial resolution and dynamic range (side-lobe
about this microphone array development platform are suppression). Some theoretical background of near and far-
presented in [21]. field beamforming is highlighted and generic version of
near and far-field delay and sum beamformers are
formulated. Presented experimental results prove that
realized development platform with low cost microphone
array and multichannel acquisition system is a good basis
for future research.

ACKNOWLEDGMENT
Fig.4. Generic version of far-field Delay-and-Sum This work was supported by the Serbian Ministry of
beamformer Education and Science project TR32026

Acoustic signal (preamble of the IR UWB signal under REFERENCES

standard 802.15.4a) is sequentially emitted form loud- [1] www.acoustic-camera.com
speakers placed in near-field of the microphone array. [2] www.lmsintl.com
Generic version near-field beamformer presented in the [3] www.bksv.com
[4] www.norsonic.com
Fig.3. is applied for sound source localization. Results of [5] www.microflown.com
sound localizations for loadspeakers are presented in the [6] Ulf Michel, “History of acoustic beamforming”, Berlin
fig.6. Results of near-field localization of loudspeakers Beamforming Conference (BeBeC) 2006, 21-22. Nov.2006
[7] S.Guidati, "Advanced beamforming techniques in vehicle acoustic",
provided in MATLAB are overlayed with picture, realized Berlin Beamforming Conference (BeBeC) 2010.
with digital camera, from the position of the centre of [8] C.Cariou, O.Delvedier "Localizing aircraft noise sources with large
microphone array. scale acoutic antenna" 27th International congress of the aeronautical
sciences
[9] J.S.Pascal, J.F.Li "Use of double layer beamforming antenna to idetiy
and locate noise in cabins, EURONOISE 2006, Finland
[10] S.C.Yu,T.W.Kim,A.Asada,S.Weatherwax,B.Collins,J.Yuh
"Development of High-Resolution Acoustic Camera based Real-Time
Object Recognition System by using Autonomous Underwater
Vehicles", Proc.of the MTS/IEEE OCEANS 2006, Boston,
Massachusetts, September 15-21, 2006
[11] "Battlefield Acoustic", Microflown ebook, chapter 21
[12] "Acoustic holography Noise Source identification and Quantification",
LMS Application note.
[13] Joe C. Chen, Kung Yao, Ralph E. Hudson "Acoustic Source
Localization and Beamforming:Theory and Practice", EURASIP
Journal on Applied Signal Processing 2003:4, 359–370 Hindawi
Fig.5. Measurement setup using circular microphone array Publishing Corporation
[14] "Beamforming", Brüel&Kjar Technical Review, No.1. 204 available
on site www.bksv.com
[15] F.Dablauwe, K.Jansen, M.Robin "Extending the usability of near-field
acoustic holography and beamforming by using focalization, 14th
International congress on Sound and Vibration, July 2007, Cairns,
Australia
[16] "Near field Acoustic", Microflown ebook, chapter 8.
[17] "Far field Acoustic", Microflown ebook, chapter 9
[18] F.Dablauwe, K.Jansen, M.Robin," Combining acoustic imaging
techniques to localize and identify sources", Berlin Beamforming
Conference (BeBeC) 2008
[19] K.Haddad, J.Hald,"3D Localization of acoustic sources with spherical
array", Acoustics'08,Paris, June 29- July 4. 2008
[20] 0(ULü$=HMDN02EUDGRYLü ³$PELJXLW\ &KDUDFWHUL]DWLRQ RI
Arbitrary Antenna Array: Type I Ambiguity”, Fifth IEEE International
Symposium on Spread-Spectrum Techniques and Applications, IEEE
ISSSTA’98, Sun City, pp.399-402
[21] M.0LMLü '.0DãRYLü '.Šumarac PavORYLü DQG 0. $GQDÿHYLü A
Model of Planar Microphone Array Realized with Low-cost
Multimedia Microphones", this conference
[22] B. Zimmermann, C. Studer, “FPGA-based Real-Time Acoustic
Fig.6. Results of localization loudspeakers by near-field Camera Prototype” Integrated Systems Laboratory ETH Zurich, 8092
Zurich, Switzerland
beamforming [23] J.Lanslot,F Deblauwe, K.Janssens," Sound Source Localization",
INDUSTRIAL TECHNOLOGY • March 2011
V. CONCLUSION [24] C.H.Jones, G.A.Gimour "Acoustic Camera Apparatus" United States
Patent, Appl.No. 456.419, 5,August 1975.
Application of near-field beamforming in acoustic [25] Y.A.Gurovich,K.J.Plotkin,D.H.Robinson,W.K.Blake,P.R.Donavan,
cameras is the key research challenge since this kind of "Acoustic Beamforming: Mapping Sources of Truck Noise", NCHRP
application is of a relatively new date. So the future report 635, National Cooperative Highway Research Program,
Washington DC,2009.
research will be focused on the improvement of

1039

View publication stats