ES2275410B1 - PROCEDURE FOR DETECTION OF TERMS THROUGH ELECTRONIC AND COMPUTATION TECHNIQUES USING MULTIDIMENSIONAL SPECTERS. - Google Patents

Abstract

Termite detection procedure using electronic and computational techniques using multidimensional spectra. It consists in analyzing the frequency response of orders three and four, or Fourier transforms of the cumulants, in order to detect predominant amplitudes in the non-redundant zone of the bispectro and the triespectro. Eliminates noise coupled to signals of interest that have symmetrical probability distributions, pre-processing the original time series, in order to eliminate low frequency components, which are not observed in the acoustic and vibrational emissions of insects. It can be implemented using a wide range of vibrational sensors and laptops, in which its sound card data acquisition interface can be used. The use of the procedure is claimed in the field of early pest detection, and it is also possible to characterize the animal species according to its three and four order spectra.

Description

Termite detection procedure by electronic and computational techniques using spectra multidimensional

Industrial sector

Electronic Instrumentation Industry, Programming, Signal Processing and Insect Detection (Pest Control).

Previous General Concepts Acoustic and vibrational emission

The acoustic emission, defined by the standard E-610-82 of the American Society of Testing and Materials (ASTM), is the kind of phenomenon because of which generates transient elastic energy originated in the Interior of a structure. The acoustic emission test has for end the detection of "active defects" by methods liabilities Active defects are defined as those that cause acoustic emission when the material is tested. The acoustic emission detection is produced by sensors piezoelectric Figure 1 summarizes the previous definition and the generic arrangement of the measuring equipment in the environment in which The vibrations spread.

Acoustic emission standards published by the ASTM

The ASTM has published the following standards applicable to acoustic emission and its monitoring:

?

E650-97 Standard Guide for Mounting Piezoelectric Acoustic Emission Sensors

?

E569-97 Standard Practice for Acoustic Emission Monitoring of Structures During Controlled Stimulation.

?

E749-96 Standard Practice for Acoustic Emission Monitoring During Continuous Welding

?

E750-98 Standard Practice for Characterizing Acoustic Emission Instrumentation

?

E751-96 Standard Practice for Acoustic Emission Monitoring During Resistance Spot Welding

?

E976-99 Standard Guide for Determining the Reproducibility of Acoustic mission Response sensor

?

E1065-99 Standard Guide for Evaluating Characteristics of Ultrasonic Search Units

These rules are considered in the description of the invention object of the present specification.

Types of emissions, forms of signals. The technique of acoustic event account

There are 4 basic types of acoustic emissions, which are usually classified according to temporal patterns. Figure 2 shows an example of continuous acoustic emission. This type of emission is measured in terms of the effective value of the signal monitored (root mean square, RMS) of the monitored signal.

Figure 3 shows a burst type emission (burst) This emission is characterized by setting a threshold of tension, and counting the events that cross it. Everytime that the threshold is crossed, a pulse is produced, which constitutes a Account for electronic measurement circuits. The device of processing counts the total impulses for a range of given time. In addition, each pulse generated by the comparator electronic enters a delay circuit, which is the detector of hits. This detector remains active while detecting the pulses  from the comparator. After the time of wait (timeout), if it does not detect more pulses, then the detector will turns off (inhibited) and all detected pulses are recorded as an "event."

Figure 4 shows the concept of threshold and record of acoustic events, or vibrations of the environment register.

Figure 5 shows a continuous noise process but "weevil" or "pointed." This type of process can found in fluids with turbulence and is also characteristic of the Termite feeding and digging (activity). This activity is the one that is mostly wanted to detect.

Figure 6 shows a burst type emission with continuous background noise. These types of time series are frequent in gas cavitation, partial discharges in power transformers with background noise, loss signals in storage tanks, and in our case, in emissions from termite alarms Figure 8 shows an example of bursts of alarms with their characteristic spectrum.

Termites Brief entomological introduction

Termites are social animals that dig intricate networks of organized galleries in order to feed the "queen", egg laying, to guarantee the next generation and, consequently, the survival of the species. The "workers" feed the queen and dig galleries, the "soldiers" protect the colony from external aggressions, and the "winged" are responsible for colonizing new places of feeding and breeding. A colony can extend within a radius of 1 km.

These invertebrates feed primarily cellulose; usually invade places where there is wood and Their derivatives. During the feeding, the "workers" produce random vibrational signals (figure 9). If the termites "soldier" detect an intruder in the colony, emit trains from impulses Each pulse train or "burst" contains three or four equalized impulses. They are also equally spaced in the Bursts time. The separation time depends on the kind of termite. The analysis of these vibrations in the time domain and of the frequency it allows to decide whether or not there is invasion (plague). In addition, each species has an emission spectrum with one or several characteristic carrier frequencies. This spectrum can extend to ultrasound, so sensors have been designed that they enhance this area of the spectrum, in order to detect them.

The elimination of these pests has been based until recently on the use of "Temiticides." These products heavily pollute the environment, so other products have been developed (applied in "baits"), which prevent "workers" feed the termite "queen."

State of the art electronic and computational techniques of detection Signal processing

At present, the electronic and signal processing procedures used in the detection of insects, and in particular termites, are based on the use of high-cost equipment (acoustic and vibrational sensors). Signal processing techniques employ order 2 statistics, such as spectral analysis ( Fourier transform of the autocorrelation function), and the acoustic event count (as described in the previous section).

The spectral analysis technique (the domain of frequency) is based on diagnostic procedures Traditional In the frequency domain, the impulse response of the acoustic sensor or accelerometer (provided or not provided by the manufacturer), with the signal spectrum of interest under study, produced by the insect, which in itself it constitutes an impulse response of the system formed by the sensor and medium to termite activity. That is, it must be considered that the answer that is finally analyzed, in the presentation element of the electronic measuring equipment, is the system response formed by the medium and the sensor to a impulse. This impulse can be transmitted by direct contact or through the air, producing vibrational signals and acoustic The frequency response obtained has a zone of ultrafrequency to be considered.

Dr. R. Mankin's team ( http://cmave.usda.ufl.edu/\simrmankin/index.html , United States Department of Agriculture, Agricultural Research Service) is a benchmark in the digitization of sounds produced by different species of invertebrates, and in the spectral analysis of the sounds. These experiences have been carried out in laboratories and under controlled conditions.

The detection methods are also based on the use of electronic counters that count the number of impulses or disturbances produced by insects in the middle in which they develop. Specifically, the AED2000L device and its probe SP-1L (from AEC, Acoustic Emission Consulting Inc.) they operate as described in the section "the event account acoustics. "The system is the result of a collaboration between AEC and "Agricultural Research Service" of the States United. This equipment has been conceived in order to detect the following pests:

?

Termites (in wood, trees and subsoil),

?

beetles and devouring larvae of roots,

?

ants and bees "carpenters",

?

"horned" beetles of The citrus

?

Asian beetles,

?

palm weevils, Y

?

warehouse test grain.

The equipment is based on the detection technique passive acoustic emission, which means that it behaves like a simple microphone, which records acoustic events, and subsequently they are counted through circuits electronic

Equipment and sensor patents related to invention

There are several equipment and sensors designed ad hoc for pest detection, in addition to the one described above. This section describes related patents.

Equipment based on the use of microwaves has been developed. In the system devised by JD Robert et al (US6313643; PCT G01N 22/00, 33/46; equivalent WO9709611), an emitter generates a beam that strikes the affected surface or object. Part of the beam is reflected and is detected and processed by arrays of neural networks, whose mission is to decide from the reflected beam if there is a pest.

The sensor described in JP7115887, and devised by O. Hirotaka, allows to measure the vibrations produced by the activity of termites thanks to the electrical output of a microswitch or relay, which is activated as a result of the movement of insects when interacting with the end of sensor (nailed) in the middle. The end contains a bait food

Reference invention JP9168359 (US5877422), devised by O. Hirotaka, incorporates an "array" of relay type sensors that, when activated by insects, produce electrical signals that are recorded in a unit external to the medium where the sensors are connected.

The invention more similar to ours has been registered in patent number WO2004088454 (Inventor AH Andeson et. Al ; applicant HARRIS CORP). It is based on the reception of a signal and the calculation of its self-accumulators of order 4. These statistics of order 4 are compared with those stored in a database, which allow the system to classify, characterize and identify (fingerprint) the signal. This invention is very focused on the military field in Communications Electronics. The system aims to inform the user about the location and type of enemy emissions. The user can thus determine the type of troop movement and the amount, as well as the distance they are. It is based on the automatic recognition of communication signal patterns in a digital modulation format.

Disadvantages appreciated in the previous inventions

The methods described present two drawbacks on which the present invention is based. First, they are based on the use of electronic components (hardware configurations) relatively expensive, such as Acoustic sensors and radiation emitters and receivers. By On the other hand, any vibration not produced by insects is would also record in the inventions that use relays.

The high hardware cost does not guarantee the detection under conditions of high SNR (signal to noise ratio). The biophysical factors, such as the high attenuation coefficient of the wood, and the environmental ones, also affect the degradation of The efficiency of the algorithms. In addition, background noise is overlaps the signal of interest, contaminating the records of signals and causing decisive errors regarding their interpretation.

Second, signal processing performed is second order, or based on neural networks. In In any case, the elimination of noise by application of statistical independence. This question results decisive in a problem of location of the emitting source.

Cumulants, or statisticians of order superior, HOS (Higher Order Statisitics) have demonstrated their success in the blind separation of signals in different fields of Science and Technology, as in Biomedicine, Radar, Communications, etc. He Successful employment of higher order statisticians is that retain the information related to the phases of the signals treated and to which (largely) eliminate noise processes with symmetric probability distributions. In addition, this tool Advanced serial processing, provides information relative to the nonlinearities of the systems.

Taking advantage of this circumstance, it presents the procedure for detecting termite activity (considering its double facet, power-movement and alarm) in environments confluence of multiple noise sources.

Description of the invention

The detection procedure is carried out by the combined analysis of the spectra of order three and four ( Fourier transforms of the cumulants of order three and four called biespectro and triespectro, respectively). The most important part of the procedure performs the fast Fourier transform of the cumulants of order 3 and 4; respectively obtaining a plane and a "cube" in the frequency domain, instead of a single frequency dimension wmo that obtained in the classical spectrum of a signal.

Matrices and complex number cubes gather Lots of redundant information. The properties of the cumulants for the selection of priority addresses in the organizational structure of the data. Specifically, the diagonal representations in the frequency domain.

The following steps are established in the process.

one.

Spectral characterization of the frequency response of the sensor to be connected, with the fm of set the resonance peaks of the profit diagram or amplifications depending on the frequency. The sensor is subjected to non-saturating vibrational impulses and its spectrum is obtained in frequency. Examples in figures 8 and 9.

2.

Low pass filtering of the signals in order to eliminate low frequency components (frequencies below 2000 Hz), which introduce little information useful since it is not typical of insect species. Example in figure 10.

3.

Obtaining of the two-dimensional transform of the cumulant of order three by means of a non-biased statistical estimator, and prior window with "Hanning" or "Blackmann" windows. The maximum inter-sample jump (lag) will be chosen as the tenth of the total samples of the analyzed record, rounded down. There is no need for overlap between records (sets of samples of the same signal); that is to say, the fast Fourier transform (Fast Fourier Trasform, FFT) of the filtered signal is calculated in its entirety and sold (multiplied by a "window" function, which smooths it at the ends without altering its frequency components, thus avoiding that high frequency components due to jumps are introduced at the ends of the acquired signal register, thus avoiding the phenomenon of "leakage or spectral loss"). Examples in figures 11 and 12.

Four.

Selection of relevant addresses of the plane in a contour diagram, which allows to confirm the choice of the main cross diagonal of the number matrix complex. This address reduces complexity Computational algorithm analysis. Examples in figures 13 and 14 (correspond to the cumulants of order 4, being analogous to  order 3).

5.

Repeat the procedure with the cumulant of order four, obtaining the "triespectro" in the frequency domain. This "cubic" frequency analysis allows corroborating the results of triespectro. Examples in Figures 13, 14 and 15.

6.

He detection criterion is based on the appearance or not of a peak spectral in the non-redundant zone of the bispectro and the triespectro. This spectral peak coincides with one of the frequencies of accelerometer resonance or vibration sensor used. If he team only records random noise, the shape of the diagrams Spectral is not pointed, and remains flat. Example in figure 12.

The procedure can be implemented in a generic equipment that responds to the block diagram of figure 7. The signal conditioning block will be characteristic of the equipment Signal acquisition used. A "display" allows visually analyze the amount of amplitudes of the Higher order statistics in the frequency domain.

The procedure can be implemented by doing use of a wide range of vibration sensors and computers laptops, in which your sound card can be used as data acquisition interface.

The use of the procedure described, in combination of hormonal techniques for pest eradication, allow the next generation not to feed the termite "queen", who dies of starvation.

Description of the figures

Fig. 1. Production and emission detection acoustic

Fig. 2. Continuous acoustic emission.

Fig. 3. Burst type acoustic emission.

Fig. 4. Acoustic event concept and threshold.

Fig. 5. Continuous acoustic event and "weevil."

Fig. 6. Burst emission with noise.

Fig. 7. Generic block diagram of the team detection. CAD.- Converter Analog-Digital

Fig. 8. a. Termite Alarm Issuance Reticulitermes lucifugus.

Fig. 8. b. Spectrum of acoustic emission. Taken with the seismic accelerometer KB12V.

Fig. 9. Acoustic emission signals of various quality. Q-D and Q-E are not audible and  can be identified with the specific accelerometer for the SP1-L insect detection. the background may mislead if gold sensor is used.

Fig. 10. Emission filtering procedure acoustics associated with food. The units are dispensed with For simplicity. Ordered measures in quantization states.

Fig. 11. Comparison of the second spectrum order with the third order spectrum for the filtered emission of Figure 10. A characteristic spectrum enhancement is observed.

Fig. 12. Third order diagonal spectra for signals of different quality, according to their contamination of noise and hearing. Q-D qualities and Q-E are not audible (in the time domain the Termite power signals are fully covered of noise). Background noise has a spectrum of third flat order while the masked emissions exhibit a peak in the resonance frequency of the sensor (6 kHz).

Fig. 13. Triespectro "slice" emission of termite feeding, used to obtain 2D diagrams of Figure 15. The peaks in the frequencies of sensor resonance.

Fig. 14. Top view by way of diagram of contours of the triespectro of Figure 13. Horizontal units and Vertical in kHz.

Fig. 15. Diagonal representations of various triespectros. Q-A, Q-B and Q-D correspond to different types of quality in acoustic emissions using the accelerometer SP1-L; the characteristic peak is observed in the presence of termites. The background is a measure of the background sound (most unfavorable case found) when there is no termite activity, with the SP1-L sensor, it observe a peak but not in the sensor resonance. Graphic KB12V presents a diagonal characterization of the triespectro but now using a seismic accelerometer, with another frequency of resonance; This sensor is not specific to the detection of insects

Claims (7)

1. Termite detection procedure using electronic and computational techniques using multidimensional spectra characterized in that it is based on the calculation of the cumulants of order three and four, and their respective frequency spectra (bispectro and triespectro), from the original signals in the time domain (time series). 2. Termite detection method by electronic and computational techniques using multidimensional spectra, according to claim 1, characterized in that it comprises the following steps:

-

Spectral characterization of the frequency response of the sensor to be connected, with the fm of set the resonance peaks of the profit diagram or amplifications depending on the frequency.

-

Low pass filtering of the signals, in order to eliminate low frequency components, that introduce little useful information, since it is not typical of insect species

-

Obtaining the transform module two-dimensional of the cumulant of order three (biespectro), by means of a statistical estimator not biased, and prior smoothing with windows from "Hanning" or "Blackmann".

-

Extraction of the diagonal elements of the data matrix of the bispectral plane of modules, which correspond to the same frequency in both dimensions (bisector of the first frequency quadrant). Representation of the elements in a contour diagram in order to confirm the choice.

-

Repeat the procedure with the order four cumulant: obtaining the module "triespectro" in the frequency domain, and selection of The main diagonal.

3. Termite detection procedure by electronic and computational techniques using multidimensional spectra, according to claims 1 and 2, characterized in that it is independent of the type of accelerometer or vibration sensor used. 4. Termite detection method by electronic and computational techniques using multidimensional spectra, according to claims 1, 2 and 3, characterized in that it can comprise the use of a sound card of a laptop computer, as data acquisition interface. 5. Use of the detection procedure of termites using electronic and computational techniques using multidimensional spectra, described in claims 1, 2, 3 and 4, for the characterization of the type of termite, or of the species of detected insect, from its order spectra higher. 6. Use of the detection procedure of termites using electronic and computational techniques using multidimensional spectra, described in claims 1, 2, 3 and 4, as a non-destructive and preventive test in the detection of insects that produce vibrational signals. 7. Use of the detection procedure of termites using electronic and computational techniques using multidimensional spectra, in combination of hormonal techniques for the eradication of pests.