Eco localization by the analysis of the characteristics

of the reflected waves in audible frequencies

Josue Pareja-Contreras Manuel Sotomayor-Polar

School of Electronic and School of Electronic and
Telecomunication Engineering Telecomunication Engineering
Catholic University San Pablo Catholic University San Pablo
Arequipa, Peru Arequipa, Peru
Email: [email protected] Email: [email protected]

Abstract—This paper presents the design and implementation and 128 sensors in a circular and square arrays to find
of an echo-localization system based on an array of micro- sound sources using the Sum Delay Beamforming method
phones with a sound source that works at the audible range. (SDBF) in the frequency range of human voice. Lutsberg in
The received signals are processed by a spatial filter that was his work [6] developed a Beamforming system that works
designed using the Sum Delay Beam Forming method (SDBF). from 250 Hz to 7200 Hz and it is able to recognize the
After that, the echoes from the targets are analyzed in order to predominant of two sound sources with a precision of 95%.
calculate the relative distance from the source to an objective. Sun [7] designed a sound source localization based on an
array of 4 microphones. The objective of this work was
implement a low cost platform to study the performance of
adaptive beam- forming algorithms. The tests were carried
1. Introduction out with two sound sources and the system was able to
recognize the direction from both. The reduction of spatial
The spatial filtering methods are definitely one of the noise is also one of the most common applications, Byung-
most common techniques used by radar and sonar systems Jun [8] integrateed a robust generalized sidelobe canceller
that work with very high frequencies, in which common de- (RGSC) with a probabilistic estimation of the signal’s arrival
vices such as personal PCs would not be able to work. That direction to suppress the noise from a real automotive envi-
is why some studies and researches use high frequencies ronment. The studies mentioned above, as this paper, use a
for echo-localization and low frequencies to detect sound single processing node and a data acquisition system which
sources or reduce the spatial noise. is based on the Single Instruction Multiple Data (SIMD)
Echo-localization is very important, but the high cost architecture described in [9].
equipment could be a limitation. This is why, this paper
proposes a system capable of locating targets using low
frequencies at audible range.

2. State of the art 3. Methodology

Currently, the most popular method for echo-localization
is primarily based on spatial filtering tools which process the
information digitally. Beam forming is the best known tool
used so far. In [1] Josserand develops a 3D reconstruction Pulses of sine waves at frequency of 1 kHz are emitted,
system based on a miniature sonar by using the Frequency these pulses are separated from each other by a prudent
Steered Phased Array (FSPA) method and ultrasonic sen- time, called timeout, in order to avoid overlapping waves
sors. Matthew in [2] [3] proposes one dimension array which from the pulse source and the echoes signals that would
uses only one sound source working at 40KHz and MEMS be still arriving. The echoes are collected by a linear array
microphones as ultrasonic sensors. In lower frequencies, of six microphones, and a digital to analog conversion is
such as human-perceptive sounds, a sensor array was used performed so that the information can be processed using
to locate sound sources as demonstrated in [4] .In this MATLAB. The Sum Delay Beamforming method is carried
study, Dannis, in collaboration with his research group, out for spatial filtering and the timeouts are extracted to find
designed a low cost video conference system that worked the time when the echoes arrive. Finally the decision process
with frequencies between 300 Hz and 3KHz using electret is applied to obtain the results. The block diagram of the
microphones. Large scale projects, such as in [5], use 24 system is shown in figure (1).
 Figura 1. Block diagram of the system

3.1. Spatial filtering

Figura 2. Sum Delay Beamforming for Eco Localization

Sum Delay Beamforming (SDBF) was used to obtain the
echoes that are collected in a 360 degree range, based on
timing delays. This system is capable of steering a listening
lobe in a desired direction and can get information from the
surroundings. The delays τn are calculated by: 3.2. Detection
The output from each angle y(t) passes through a
adapted filter which has a reference signal of 1 KHz pulse,
giving an increment of signal to noise ratio (SNR). The
n.d.cos(φ0 ) result, now called A(t) is squared, to obtain the maximun
τn = (1) point of correlation between the received signal and the 1
c
KHz pulse. Subsequently the envelope E(t) is obtained us-
ing a finite impulse response (FIR) low pass filter to reduce
Where n represents the number of microphones or sen- the oscillatory effect of A(t)2 . The estimator determines if
sors, d represents the distance between each element, c the received signal is noise or a hit. For that, the probability
is the speed of sound and φ0 is the desired angle. They density functions must be obtained.
The likelihood of the echoes px|H (x|1) and the like-
are subsequently added to obtain the output signal y(f )
lihood of the noise px|H (x|0) are used with a maximun
represented in the frequency domain as:
aposteriori estimator to differentiate the echoes from the
noise.

PH (1).px|H (x|1) > PH (0).px|H (x|0), D = 1

N
−2πf (4)
1 X 0 PH (1).px|H (x|1) < PH (0).px|H (x|0), D = 0
y(f ) = xn (f )ej c (n − 1)dcos(φ ) (2)
N n=1 Where PH (1) and PH (0) represent the a priori probabi-
lity of the echoes and noise respectively, these are obtained
by estimating how many samples from the processed data
corresponds to each of the hypotheses.
And in time domain as:

N
1 X
y(t) = xn (t − tn ) (3)
N n=1

Each calculated output passes to the detection stage. Figura 3. Signal processing to the output of Delay Sum Beamforming
4. Platform Design

The figure (1) is used as a reference to understand

all the stages of the system and what was used for their
implementation

4.1. Sound emission

A generic loudspeaker was used as a sound source and it

was located in the center of the microphone array, connected
to a Huawey Y635 smartphone to generate several pulses.

4.2. Sound reception

Figura 4. Array of microphones and their amplifiers

The elements of this stage are the sensors array and

their respective amplifiers. The sensor array is constituted
by electret microphones which are very accessible and easy
to use, but they require power supply that can be provided by 4.3. Acquisition Equipment
the amplifier circuit. The LM386 operational amplifier was
used to amplify the signal because of their low complexity For this stage, The TonePort UX8 preamplifier, from
for the implementation in a PCB circuit. The microphones Line 6 Company, was used as data acquisition system, figure
and amplifiers are mounted on a wooden frame that is (5). This device information is collected by the software
supported by a steel pedestal and where the sound source is Pro Tools 10 and has been sampled at 96 KHz. This is
also placed, see image (4). The amplifiers are powered by subsequently exported to Matlab.
a 5 volt source.
The microphone array is 50 cm (L) with 10 cm of
separation between elements (d). It is necessary to take into
account the restriction of the equation (5) to avoid facing
problems related to spatial aliasing.

λmin
d< (5)
2

Where λmin represents the minimum wavelength to be

used. Figura 5. TunePort Ux8
 objective at different angles, see the third column of (1).
Although, the detected angles are in an acceptable range,
they could generate a shift enough to vary the calculated
5. Experimentation distance.
-The speed of the sound was considered a constant
This system was tested in open areas with little or no throughout the processing because of the short distances to
reverberation, so the echoes reflect back from the targets be measured. This also could add some deviations.
which have been positioned in different locations around -The sound source could not be perfectly located in the
the sensor array. The distance between the target and the center of the array.
microphones array cannot be greater than r that is defined
on the equation (6).
5.1.2. Location. The spatial filtering gives the signal T e(t)
per each explored angle. Plotting this signals together shows
2L2 a power sweep which represents the predominant echoes
|r| > (6)
λ signals from the objective, see figure (7). In this case, the
Where λ is the wavelength of the signal emitted by the objective was located in front in front of the array.
sound source.
The output signal of the envelope filter E(t), has several
pulses that where received with their respective timeouts, in
there, the echoes can be found. A time averaging was carried
out to integrate the information of several pulses, see figure
(6). The first part of the pulse belongs to the sound source’s
coupling and it should be removed in order to use only the
signal T e(t) .

Figura 7. Barrido de potencia (Objetivo al frente)

Figura 6. Pulse integration

The highest power value was found at 90° at 5.9167
meters. One nonparametric and three parametric methods
were used to estimate the probability density function of
5.1. Results the noise signal.

5.1.1. Distance Measuring. The presented system was

compared with a laser distance meter GLM 80 BOSCH
because of the great accuracy in their measurements. The
results are shown in the table (1) The objective was located
in front of the microphone array and four tests were made
moving it one meter further from its previous position each
time.
GLM 80(meters) SDBF (meters) Angle of the objetive
2.910 m 2.8637 m 93°
3.9154 m 3.8623 m 91°
4.9475 m 4.8806 m 92°
5.9062 m 5.8685 m 91°
Cuadro 1. C OMPARISON

The explanations of the difference between the results Figura 8. Probability Density (Noise)
are as follows:
-Despite the target was located just in front of the sensor
array at 90 degrees, the sum delay beamforming detects the
 Gaussian Kernel estimation got the least mean squared
error compared to the histogram estimation. That is why it
was used to get the probability density function (PDF) of
noise and echoes signals.
The target was moved four more times to test the per-
formance in different locations. The results are shown in
the figures(9),(10),(11),(12) .

Figura 12. Estimator Output (Objective at 51°, Distance = 6.1240 m)

6. Conclusions

-An echo-localization system was designed and built

using low-cost, common and accessible devices. This could
Figura 9. Estimator Output (Objective at 71°, Distance = 6.1329 m) be useful and help in the development of new algorithms and
techniques that may be implemented in higher frequencies.
-The Near field and spatial aliasing restrictions should
be considered in order to reduce the minimum detectable
distance and to improve the radiation pattern of the sensor
array.
-A spatial filtering was implemented without expensive
elements such as hypercardioid microphones, giving good
results.
-Spatial resolution, in degrees, is directly proportional to
the sampling frequency of the acquisition device.
-The intensity of an echo signal depends on the pro-
ximity of the objective to the sensors and the area that is
exposed to the microphone array. A bigger objective would
give a more intense echo.

Figura 10. Estimator Output (Objective at 140°, Distance = 6.1240 m) References

[1] T. Josserand and J. Wolley, “A miniature high resolution 3-d imaging

sonar,” Ultrasonics, 2011.
[2] M. A. Clapp and J. Hopkins, “Sonar echo-location in 2-d using mini-
microphone array and spatiotemporal frequency filtering,” Circuits
and Systems, 2005. ISCAS 2005. IEEE International Symposium on,
vol. 5, pp. 4775 – 4778, mAY 2005.
[3] M. A. Clapp and J. Hopkins, “Single ping multiple measurements:
Sonar bearing angle estimation using spatiotemporal frequency fil-
ters,” IEEE Transactions on Circuits and Systems I: Regular Papers,
vol. Volume:53, pp. 769 – 783, April 2006.
[4] J. A. Danis, “The acoustic beamformer,” 2014.
[5] S. Kagami ,H. Mizoguchi , Y. Tamai and T. Kanade, “Microphone ar-
ray for 2d sound localization and capture,” Robotics and Automation,
2004. Proceedings. ICRA ’04. 2004 IEEE International Conference
on, vol. 1, pp. 703 – 708, May 2004.
Figura 11. Estimator Output (Objective at 93°, Distance = 5.9668 m) [6] R. J. Lustberg, “Acoustic beamforming using microphone arrays,”
Master’s thesis, Massachusetts Institute of Technology, 1993.
[7] Y. Sun , J. Zhang and X. Yang, “Design of experimental adaptive
beamforming system utilizing microphone array,” Radar Conference
2013, IET International, pp. 1 – 5, April 2013.
[8] B.-J. Yoon, I. Tashev , A. Acero, “Robust adaptive beamforming
algorithm using instantaneous direction of arrival with enhanced
noise suppression capability,” 2007 IEEE International Conference
on Acoustics, Speech and Signal Processing - ICASSP ’07, vol. 1,
pp. I–133 – I–136, 2007.
[9] A. D. George, J. Markwell, R. Fogartye, “Real-time sonar beamform-
ing on high-performance distributed computers,” Parallel Computing,
pp. 1231–1252, 2000.
[10] I. McCowan, “Microphone arrays : A tutorial,” 2001.