WAVELETS TRANSFORMS APPLIED TO TERMITE DETECTION

Carlos Garc

ıa Puntonet

Univ. Granada. Dept. ATC, ESII. Periodista Daniel Saucedo. E-18071-Granada-Spain

Juan Jos

e Gonz

alez de la Rosa

Univ. C

adiz. Electronics Area, EPSA. Ram

on Puyol S/N. E-11202-Algeciras-Spain

Isidro Lloret Galiana

Univ. C

adiz. Dpt. Computer Science, EPSA. Ram

on Puyol S/N. E-11202-Algeciras-Spain

Juan Manuel G

orriz S

aez

Univ. Granada. Dept. ECT, Fuentenueva S/N. E-18071. Granada. Spain

Keywords:

Acoustic emission, termite detection, ultrasonics, vibratory signal, seismic accelerometer, transient detection,

wavelets.

Abstract:

In this paper we present an study which shows the possibility of using wavelets to detect transients produced

by termites. Identiﬁcation has been developed by means of analyzing the impulse response of three sensors

undergoing natural excitations. De-noising by wavelets exhibits good performance up to SNR=-30 dB, in the

presence of white gaussian noise. The test can be extended to similar vibratory or acoustic signals resulting

from impulse responses.

1 INTRODUCTION

In acoustic emission (AE) signal processing a custom-

ary problem is to extract some physical parameters of

interest in situations which involve join variations of

time and frequency. This situation can be found in

almost every nondestructive AE tests for character-

ization of defects in materials, or detection of spu-

rious transients which reveals machinery faults (Lou

and Loparo, 2004). The problem of termite detection

lies in this set of applications involving non-stationary

signals (de la Rosa et al., 2004b).

When wood ﬁbers are broken by termites they pro-

duce acoustic signals which can be monitored using

ad hoc resonant acoustic emission (AE) piezoelectric

sensors which include microphones and accelerome-

ters, targeting subterranean infestations by means of

spectral and temporal analysis. The drawback is the

relative high cost and their practical limitations (de la

Rosa et al., 2004b).

The usefulness of acoustic techniques for detec-

tion depends on several biophysical factors. The

main one is the amount of distortion and attenua-

tion as the sound travels through the soil (∼600 dB

−1

, compared with 0.008 dB m

−1

in the air). Fur-

thermore, soil and wood are far from being ideal

acoustic propagation media because of their high

anisotropy and frequency dependent attenuation char-

acteristics (Mankin and Fisher, 2002). This is the rea-

son whereby signal processing techniques emerged as

an alternative.

Second order methods (spectra) failure in low SNR

conditions even with ad hoc piezoelectric sensors.

Bispectrum have proven to be a useful tool for char-

acterization of termites in relative noisy environments

using low-cost sensors (de la Rosa et al., 2004a),(de la

Rosa et al., 2004c). The computational cost could be

pointed out as the main drawback of the technique.

This is the reason whereby diagonal bispectrum have

to be used.

Numerous wavelet-theory-based techniques have

evolved independently in different signal process-

ing applications, like wavelets series expansions,

multiresolution analysis, subband coding, etc. The

wavelet transform is a well-suited technique to de-

tect and analyze events occurring to different scales

(Mallat, 1999). The idea of decomposing a signal

into frequency bands conveys the possibility of ex-

tracting subband information which could character-

ize the physical phenomenon under study (Angrisani

et al., 1999).

In this paper we show an application of wavelets’

de-noising possibilities for the characterization and

detection of termite emissions in low SNR condi-

tions. Signals have been buried in gaussian white

noise. Working with three different sensors we ﬁnd

163

García Puntonet C., José González de la Rosa J., Lloret Galiana I. and Manuel Górriz Sáez J. (2005).

WAVELETS TRANSFORMS APPLIED TO TERMITE DETECTION.

In Proceedings of the Seventh International Conference on Enterprise Information Systems, pages 163-167

DOI: 10.5220/0002512101630167

 SciTePress

that the estimated signals’ spectra match the spectra

of the acoustic emission whereby termites are identi-

ﬁed.

The paper is structured as follows: Section 2 sum-

marizes the problem of acoustic detection of termites;

Section 3 remembers the theoretical background of

wavelets; Section 4 describes the experiments carried

out. Conclusions are drawn in Section 5.

2 ACOUSTIC DETECTION OF

TERMITES

2.1 Characteristics Of The AE

Signals

Acoustic Emission(AE) is deﬁned as the class of phe-

nomena whereby transient elastic waves are generated

by the rapid (and spontaneous) release of energy from

a localized source or sources within a material, or the

transient elastic wave(s) so generated (ASTM, F2174-

02, E750-04, F914-03

). This energy travels through

the material as a stress or strain wave and is typically

detected using a piezoelectric transducer, which con-

verts the surface displacement (vibrations) to an elec-

trical signal.

Termites use a sophisticated system of vibratory

long distance alarm. When disturbed in their nests

and in their extended gallery systems, soldiers pro-

duce vibratory signals by drumming their heads

against the substratum (R¨ohrig et al., 1999). The

drumming signals consist of pulse trains which propa-

gate through the substrate (substrate vibrations), with

pulse repetition rates (beats) in the range of 10-25 Hz,

with burst rates around 500-1000 ms, depending on

the species (Conn

etable et al., 1999). Soldiers pro-

duce such vibratory signals in response to disturbance

(1-2 nm by drumming themselves) by drumming their

head against the substratum. Workers can perceive

these vibrations, become alert and tend to escape.

Figure 1 shows one of the impulses in a burst and

its associated power spectrum is depicted in ﬁgure

2. Signiﬁcant drumming responses are produced over

the range 200 Hz-10 kHz. The carrier (main com-

ponent) frequency of the drumming signal is around

2600 Hz.

The spectrum is not ﬂat as a function of frequency

as one would expect for a pulse-like event. This is due

American Society for Testing and Materials. F2174-

02: Standard Practice for Verifying Acoustic Emission Sen-

sor Response. E750-04: Standard Practice for Characteriz-

ing Acoustic Emission Instrumentation. F914-03: Standard

Test Method for Acoustic Emission for Insulated and Non-

Insulated Aerial Personnel Devices Without Supplemental

Load Handling Attachments.

to the frequency response of the sensor (its selective

characteristics) and also to the frequency-dependent

attenuation coefﬁcient of the wood and the air.

0.005 0.01 0.015 0.02 0.025 0.03

-0.6

-0.4

-0.2

0.2

0.4

0.6

Time (s)

Normalized amplitude

A single impulse in a burst

Figure 1: A single pulse of a four-pulse burst.

0 5 10 15 20 25 30 35 40 45

-9

-8

-7

-6

-5

-4

-3

-2

-1

Frequency (kHz)

Normalized power

Spectrum of a pulse

Figure 2: Normalized power spectrum of a single pulse.

2.2 Devices, Ranges Of

Measurement And HOS

Techniques

Acoustic measurement devices have been used pri-

marily for detection of termites (feeding and exca-

vating) in wood, but there is also the need of de-

tecting termites in trees and soil surrounding build-

ing perimeters. Soil and wood have a much longer

coefﬁcient of sound attenuation than air and the co-

efﬁcient increases with frequency. This attenuation

reduces the detection range of acoustic emission to 2-

5 cm in soil and 2-3 m in wood, as long as the sensor

is in the same piece of material (Mankin et al., 2002).

The range of acoustic detection is much greater at fre-

quencies <10 kHz, and low frequency accelerometers

ICEIS 2005 - HUMAN-COMPUTER INTERACTION

164

have been used to detect insect larvae over 1-2 m in

grain and 10-30 cm in soil (Robbins et al., 1991).

It has been shown that ICA success in separating

termite emissions with small energy levels in com-

parison to the background noise. This is explained

away by statistical independence basis of ICA, re-

gardless of the energy associated to each frequency

component in the spectra (de la Rosa et al., 2004c).

The same authors have proven that the diagonal bis-

pectrum can be used as a tool for characterization

purposes (de la Rosa et al., 2004a). With the aim

of reducing computational complexity wavelets trans-

forms have been used in this paper to de-noise cor-

rupted impulse trains.

3 THE WAVELET TRANSFORM

3.1 Continuous Wavelet Transform

(CWT)

A mother wavelet is a function ψ with ﬁnite energy

and zero average:

+∞

−∞

ψ(t)dt = 0, (1)

This function is normalized

, kψk = 1, and is cen-

tered in the neighborhood of t=0.

ψ(t) can be expanded with a scale parameter a,

and translated by b, resulting the daughter functions

or wavelet atoms, which remain normalized:

a,b

(t) =

√



t − b



; (2)

The CWT can be considered as a correlation between

the signal under study s(t) and the wavelets (daugh-

ters). For a real signal s(t), the deﬁnition of CWT

is:

CW T s(a, b) =

√

+∞

−∞

s(t)ψ

∗



t − b



dt; (3)

where ψ

∗

(t) is the complex conjugate of the mother

wavelet ψ(t), s(t) is the signal under study, a and b

are the scale and the position respectively (a ∈ ℜ

−

0, b ∈ ℜ). The scale parameter is proportional to the

reciprocal of frequency.

The expression for the modulus of CW T is:

|CW T s(a, b)| = k(a)

; (4)

f ∈ L

(ℜ), the space of the ﬁnite energy functions,

verifying

+∞

−∞

|f(t)|

dt < +∞.

kfk =

+∞

−∞

|f(t)|

1/2

= 1.

where α is the so-called Lipschitz exponent and k is

a constant. Looking at equation 4 one can discrimi-

nate the signal from the noise by analyzing the local

maxima of |CW T s(a, b)| across the scales.

Any ﬁnite energy signal s(t) can be decomposed

over a wavelet orthogonal basis

of L

(ℜ) according

to:

s(t) =

+∞

j=−∞

+∞

n=−∞

hs, ψ

j,n

iψ

j,n

(5)

Each partial sum can be interpreted as the details vari-

ations at the scale a = 2

(t) =

+∞

n=−∞

hs, ψ

j,n

iψ

j,n

s(t) =

+∞

j=−∞

(t)

(6)

The approximation of the signal s(t) can be pro-

gressively improved by obtaining more layers or lev-

els, with the aim of recovering the signal selectively.

For example, if s(t) varies smoothly we can obtain

an acceptable approximation by means of removing

ﬁne scale details, which contain information regard-

ing higher frequencies or rapid variations of the sig-

nal. This is done by truncating the sum in 5 at the

scale a = 2

(t) =

+∞

j=J

(t) (7)

3.2 Discrete Time Wavelet

Transform (DTWT)

In the DTWT the original signal passes through two

complementary ﬁlters and two signals are obtained

as a result of a downsampling process, correspond-

ing to the approximation and detail coefﬁcients. The

lengths of the detail and approximation coefﬁcient

vectors are slightly more than half the length of the

original signal, s(t). This is the result of the dig-

ital ﬁltering process (convolution) (Angrisani et al.,

1999). The approximations are the high-scale, low-

frequency components of the signal. The details are

the low-scale, high-frequency components.

A tree-structure arrangement of ﬁlters allows the

subband decomposition of the signal. In each stage

of the ﬁltering process the same two digital ﬁlters are

used: a highpass h

(·), the discrete mother, and its

mirror (lowpass) g

(·). All these ﬁlters have the same

relative bandwidth (ratio between frequency band-

width and center frequency). The results of the de-

composition can be expressed as:

DT W T s(j, n) =

N−1

k=0

j+1

n − k)s(k), (8)

j,n

(t) =

√

t−2

(j,n)∈Z

WAVELETS TRANSFORMS APPLIED TO TERMITE DETECTION

165

where N is the number of samples in the signal, j is

the decomposition level, n is the time shifting. The

same arguments are valid for the process of recon-

struction.

3.3 Wavelet Packets (WP)

The WP method is a generalization of wavelet decom-

position that offers more possibilities of reconstruct-

ing the signal from the decomposition tree. If n is the

number of levels in the tree, WP methods yields more

than 2

n−1

ways to encode the signal. The wavelet

decomposition tree is a part of the complete binary

tree.

When performing a split we have to look at each

node of the decomposition tree and quantify the infor-

mation to be gained as a result of a split. An entropy

based criterion is used herein to select the optimal de-

composition of a given signal. We use an adaptative

ﬁltering algorithm, based on the work by Coifman and

Wickerhauser (Coifman and Wickerhauser, 1992).

Functions that verify additivity-type property are

suitable for efﬁcient searching of the tree structures

and node splitting. The criteria based on the entropy

match these conditions, providing a degree of ran-

domness in an information-theory frame. In this work

we used the entropy criteria based on the p-norm:

E(s) =

; (9)

with p≤1, and where s(t) = [s

(t), s

(t), . . . , s

(t)]

in the signal of length N.

The results are accompanied by entropy calcula-

tions based on Shannon’s criterion:

E(s) = −

log(s

); (10)

with the convention 0 × log(0) = 0.

4 EXPERIMENTS AND

CONCLUSIONS

Two accelerometers (KB12V, seismic accelerometer;

KD42V, industrial accelerometer, MMF) and a stan-

dard microphone have been used to collect data from

termites in different places (basements, subterranean

wood structures and roots) using the sound card of a

portable computer and a sampling frequency of 96000

(Hz). These sensors have different sensibilities and

impulse response. This is the reason whereby we nor-

malize spectra.

The de-noising procedure was developed using a

sym8, belonging to the family Symlets (order 8),

which are compactly supported wavelets with least

asymmetry and highest number of vanishing moments

for a given support width. We also choose a soft

heuristic thresholding.

We used 15 registers (from reticulitermes grassei),

each of them comprises a 4-impulse burst buried in

white gaussian noise. De-noising performs success-

fully up to an SNR=-30 dB. Figure 3 shows a de-

noising result in one of the registers. Figure 4 shows

2000 4000 6000 8000 10000 12000 14000

-2

2000 4000 6000 8000 10000 12000 14000

-1

2000 4000 6000 8000 10000 12000 14000

-1

buried burst

de-noised signal at level 4

de-noised signal at level 5

Figure 3: Limit situation of the de-noising procedure

(SNR=-30 dB). From top to bottom: a buried 4-impulse

burst, estimated signal at level 4, estimated signal at level

5. Time resolution: 1/96000 (s)

a comparison between the spectrum of the estimated

signal at level 4 and the spectrum of the signal to be

de-noised, taking a register as an example. Signiﬁcant

0 10 20 30 40

-10

-8

-6

-4

-2

Frequency (kHz)

Normalized power

Spectrum of the de-noised burst

0 10 20 30 40

-4

-3

-2

-1

Frequency (kHz)

Normalized power

Spectrum of the noisy burst

Figure 4: Spectra of the estimated signal and the buried

burst.

components in the spectrum of the recovered signal

are found to be proper of termite emissions.

ICEIS 2005 - HUMAN-COMPUTER INTERACTION

166

ACKNOWLEDGEMENT

The authors would like to thank the Spanish Gov-

ernment for funding the project DPI2003-00878, and

the Andalusian Autonomous Government Division for

funding the research with Contraplagas Ambiental

S.L.

REFERENCES

Angrisani, L., Daponte, P., and D’Apuzzo, M. (1999). A

method for the automatic detection and measurement

of transients. part i: the measurement method. Mea-

surement, 25(1):19–30.

Coifman, R. R. and Wickerhauser, M. (1992). Entropy-

based algorithms for best basis selection. IEEE Trans.

on Inf. Theory, 38(2):713–718.

Conn

etable, S., Robert, A., Bouffault, F., and Bordereau,

C. (1999). Vibratory alarm signals in two sympatric

higher termite species: Pseudacantotermes spiniger

and p. militaris (termitidae, macrotermitinae). Jour-

nal of Insect Behaviour, 12(3):90–101.

de la Rosa, J. J. G., Lloret, I., Puntonet, C. G., and G

orriz,

J. M. (2004a). Higher-order statistics to detect and

characterise termite emissions. Electronics Letters,

40(20):1316–1317. Ultrasosics.

de la Rosa, J. J. G., Puntonet, C. G., G

orriz, J. M., and

Lloret, I. (2004b). An application of ica to identify vi-

bratory low-level signals generated by termites. Lec-

ture Notes in Computer Science (LNCS), 3195:1126–

1133. Proceedings of the Fith International Confer-

ence, ICA 2004, Granada, Spain.

de la Rosa, J. J. G., Puntonet, C. G., and Lloret, I. (2004c).

An application of the independent component analy-

sis to monitor acoustic emission signals generated by

termite activity in wood. Measurement, In Press, Cor-

rected Proof, Available online 12 October 2004.

Lou, X. and Loparo, K. A. (2004). Bearing fault diagnosis

based on wavelet transform and fuzzy inference. Me-

chanical Systems and Signal Processing, 18(5):1077–

1095.

Mallat, S. (1999). A wavelet tour of signal processing. Aca-

demic Press, 2 edition.

Mankin, R. W. and Fisher, J. R. (2002). Current and po-

tential uses of acoustic systems for detection of soil

insects infestations. In Proceedings of the Fourth Sym-

posium on Agroacoustic, pages 152–158.

Mankin, R. W., Osbrink, W. L., Oi, F. M., and Anderson,

J. B. (2002). Acoustic detection of termite infesta-

tions in urban trees. Journal of Economic Entomology,

95(5):981–988.

R¨ohrig, A., Kirchner, W. H., and Leuthold, R. H.

(1999). Vibrational alarm communication in the

african fungus-growing termite genus macrotermes

(isoptera, termitidae). Insectes Sociaux, 46:71–77.

Robbins, W. P., Mueller, R. K., Schaal, T., and Ebeling, T.

(1991). Characteristics of acoustic emission signals

generated by termite activity in wood. In Proceedings

of the IEEE Ultrasonic Symposium, pages 1047–1051.

WAVELETS TRANSFORMS APPLIED TO TERMITE DETECTION

167