AENTATION AD-A277 671 ENAO PPAAGGEE I~hII I II OMB No°,o70 08 r cc, Vet ,eo.nse(cid:127),re'cg tn OP 0,ra't oeqr eeo sl,co,cOs sea,, "3r 4eIflQo 32*(cid:127) ,ounces 5411e4,oaqr c d.i.c.! (cid:127)e~ts,(cid:127) D I 'oecitc oea?. 000 04,f.l0o',f imSWS asnqOSpownt ooDnCs 2.-05a0o3 Aep oS -'5 ~e-e~sorDD irs "on., , ne 204 Au 9W VA AGENCY USE ONLY ILNIM OA-) REPORt DATE 3 REPOP' TYPFA ,, DATESL 23'LRFD March 1994 Professional Paper TITLEA ND SUBTITLE 5 FUND0NGN UMBERS TARGET SONAR DISCRIMINATION CUES PR: MMB2 "PE: 0602435N 6 AUTHOR(S) WU DN688674 W L. Au 7 PERFORMINGO RGANIZATIONN AMEISIA ND ADORESSIE$S 94 -09653 Naval Command, Control and Ocean Surveillance Center tNCCOSC) RDT&E Division San Diego. CA 92152-5001 9 SPONSORINGcIJOTORWN AGENCYN AME(S) ANDA DORESSIES) Naval Command, Control and Ocean Surveillance Center (NCCOSC) RDT&E Division San Diego, CA 92152-5001 I1 SUPPLEMENTARYN OTES 120 OtSTRIUTIOFA/AVAILABIUTSYT ATEMENT 12b DISTRIBUTI CODE S~Approved for public release; distribution is unlimited. 13 ABSTRACT PAS-4- 20 -*'US) The primary emphasis in this paper is on possible cues used by dolphins in performing different sonar discrimination taska. Published in Marine Mammal Sensory Systems, Edited by J. Thomas, et al., Plenum Press, New York. 1992. pp. 357-376 1yv3 DTIC QUAL:'y rNSPEC7. 3 'S %UMBERO F PAES marine animal marine biology - ,6 PRICEC ODE Smarine biosyatems 8 17 SECUrY CbiASSIFICATIONI SECURITYC LASSIFICATION 1B SECURITYC LASSIFICATION 20 LIMITATIONO F ABSTRACT OF REPORT OF THIS PAGE OF ABSTRACT UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED SAME AS REPORT NSN ?S40.'1-2W0-50 Stwnwo so-m M (FRoWn TARGET SONAR DISCRIMINATION CUES Whitlow W. L. Au Naval Ocean Systems Center Kailua, Hawaii 96734 One of the outstanding characteristics of the dolphin sonar system which distinguishes it from any man-made sonar is the ability to make fine distinctions in the features or properties of targets. Sonar experiments have shown that dolphins can discriminate between objects differing in size, structure, shape and material composition. This discrimina- tion capability has amazed and sparked the interest of many involved in the development and use of active sonar systems. The ability to perform fine target discrimination, recog- nition, or classification are often considered synonymous however, there are subtle differences in each function. Target discrimination means the ability to discern some fea- ture or features in the sonar returns that would allow a signal processing unit to decide that target A and B are different targets. Target recognition means the ability to recognize features or qualities 'f the sonar returns associat- ed with specific targets compared with returns from any other targets. Target recognition involves a discrimination capa- bility, an ability to recall from memory the features of sonar returns from specific targets and the abili'ty to compare present sonar returns with those stored in memory. Target classification means the ability to separate targets into different classes such as metal versus non-metal, organic versus inorganic, eatable versus non-eatable, smooth surface versus rough surface etc. Most of the experiments that will be discussed in this chapter will involve target discrimina- tion; a few will involve target recognition. Target classifi- cation experiments involving many different classes of targets are generally difficult to construct and train with animals. The primary emphasis in this paper will be on possible cues used by dolphins in performing different sonar discrimi- nation tasks. Most of the dolphin sonar discrimination re- search has been performed in the Soviet Union (Ayrapet'yants and Konstantinov, 1974; Bel'kovich and Dubrovskiy, 1976) and in the United States of America. An extensive review of these experiments has been performed by Nachtigall (1980). Marine Mammal Sensory Systems, Edited by J. Thomas et a., Plenum Press, New York, 1992 357 'L4 8 29 065 In sonar discrimination experiments dolphins are gener- ally required to choose between at least two targets present- ed either simultaneously or successively. The targets will usually differ along a single physical dimension and the animal's discrimination threshold is determined by progres- sively making the difference smaller. Although differences in targets may exist in a single physical dimension (e.g. diameter, wall thickness, length, etc.), several acoustic features may Le affected as this single dimension is varied. Therefore, the important consideration in sonar discrimina- tion experiments is to determine what acoustic features are being tested and how these features change with changes to the physical characteristics of the targets. Unfortunately, this is easier said than done since the backscattering pro- cess is often quite complex even with simple geometrically shaped targets. SIZE DISCRIMINATION Cylinder Length and Diameter Ayrapet'yants et al. (1969) found that a Black Sea bottlenose dolphin (Tursiops t) could discriminate a 30 mm long cylinder from the 25 mm long standard at the 70% correct response level. Zaslavskiy et al. (1969) also found that a harbor porpoise (Phocoena phocoena) could discriminate cylinders that were 75 mm versus 95 mm in length at the 80% correct response level. Evans (1973) also studied cylinder size discrimination capability of an echolocating Atlantic Bottlenose dolphin (TursioRs trmtus) and an Amazon River dolphin (Inia aeofrensis). Solid chloroprene cylinders were presented simultaneously and the blindfolded dolphins were required to discriminate the standard from the non standard cylinder. The diameter of the non-standard cylinders was varied in 1 dB target strength increments. The results indi- cated that both species could discriminate target strength differences of 1 dB at performance levels above 70% correct. The experiments of Ayrapet'yants et al. (1969) and Zaslavskiy et al. (1969) were actually target strength dis- crimination experiments. Highlights or echo components were probably present in the echoes from the sonar signal pene- trating and propagating along different acoustic paths within the targets and from circumferential waves (Neubauer, 1986). However, for a signal that is incident normal to the longitu- dinal axis of a cylinder, the echo structure is affected by the diameter and material composition and not length. Since the diameter and material composition were fixed and only length varied, only the amplitude of the target echoes was affected by different lengths. The target strength of an acoustically rigid or soft cylinder of finite length can be expressed as (Urick, 1983) TS = 10 Log (aL2/2A) (1) where a is the radius, L is the length of the cylinder and X is the wavelength of the signal. The differences in target strength were approximately 1.6 and 2.1 dB for the targets used by Ayrapet'yants et al. (1969) and Zaslavskiy et al. (1969), respectively. These values compare well with the 358 1 dB difference observed in the experiment of Evans (1973). However, since the diameter of the cylinders was varied in Evans experiment, additional cues from circumferential waves (Barnard and McKinney, 1961; Diercks et al., 1963) may have been present. Sphere Diameter There has been an abundance of sphere size discrimina- tion experiments performed with metallic targets using Tursiois. A summation of these experiments are given in Table 1. The 4th column of Table 1 is the difference in the target strength calculated for a standard and comparison sphere at the animal's threshold. The target strength of a large, rigid or soft sphere can be expressed as (Urick, 1983) TS = 20 Log (a/2) (2) where a is the radius of the sphere. Two cues associated with sphere size discrimination are the differences in target strength and the highlight struc- ture of the echoes. The incident signal penetrating and propagating along different paths within a sphere will result in the presence of many highlights or echo components (Shirley and Diercks, 1970). Circumferential wave components will also contribute to the echo structure (Wille, 1965; Uberall et al., 1966). Examples of the echo structure and Table 1. Results of biosonar size discrimination experiments with spherical targets. Stand. Diam. is the dia- meter of the standard sphere. Increm. Diam. is the incremental diameter of the comparison sphere at the dolphin's discrimination threshold. T.S. Diff. is the difference in target strength between the standard and comparison target (calculated). Stand. Increm. T.S. Diam. Material Diam. Diff. Range References (cm) (cm) (dB) (i) 5.71 Ni-steel 0.64 0.9 >.5 Turner and Norris (1967) 10.40 steel 3.90 2.8 2 - 6 *Dubrovskiy et al. (1971) 57.10 steel 6.40 0.9 2 - 6 " - i 5.00 lead 0.50 0.8 8 +Dubrovskiy (1972) 1.02 lead 0.15 1.2 3 *Fadeyeva (1973) 1.40 lead 0.20 1.2 4.8 *Dubrovskiy and Krasnov (1971) 10.20 lead 1.50 1.2 2 - 6 Ayrapet'yant and Konstantinov (1974) + - cited in Ayrapet'yant and Konstantinov (1974) "- cited in Bel'kovich and Dubrovskiy (1976) -lodes /or 35 TI TSl Wi 4) iREiY 1WH) FREE Y WHO FNEOUENY WK.) 0. FRE(cid:127)NY WHO FREGNCY ilH) FRECUENCY I Fig. 1. Example of a simulated dolphin sonar signal and the echo from a 7.62-cm diameter water-filled stainless- steel sphere and 1 2.54-cm diameter solid steel sphere (from Au and Snyder, 1980) frequency spectrum of echoes from a solid 2.54 cm steel sphere and a 7.62-cm water-filled steel sphere are shown in Fig. 1. A simulated dolphin echolocation signal was used to produce the echoes. The highlight structure (e.g. position and amplitude of the highlights) are determined by the diame- ter and material composition of the sphere. Planar Targets Barta (1969) conducted a size discrimination experiment using circular aluminum disks covered with neoprene. The Tursios was trained to choose the smaller of two simulta- neously presented targets. A divider between the targets restricted the minimum range between the dolphin and targets to 0.7 m. The dolphin discriminated a 16.1 cm from a 15.2 cm diameter disk at the 75% correct threshold. Bel'kovich et al. (1919) used plastic foam square targets and trained a common porpoise (Delphinus delphis) to choose the larger of two simultaneously presented targets. The dolphin discrimi- nated between a 100 cm2 and 90.25 cm= target at a 77% correct response level. The main cue available in the planar target size dis- crimination was differences in target strength. The target strength of a rigid or soft planar target at normal incidence of the signal can be expressed as (Urick, 1983) TS = 20 Log (A/l) (3) where A is the area of the target and l is the wavelength of the signal in water. The target strength differences between the standard and comparison targets at threshold were 1 dB and 0.9 dB for the targets used by Barta (1969) and 360 Barta (1969), respectively. Backscatter measurements with Barta's targets indicated that threshold size discrimination was performed with a 1 dB difference in target strength. STRUCTURE DISCRIMINATION Wall Thickness Evans and Powell (1967) were first to demonstrate that a blindfolded, echolocating TzriQ could discriminate the thickness of metallic plates. The dolphin was trained to discriminate a 30 cm diameter circular copper disc of 0.22 cm thickness from a comparison target. Both targets were pre- sented simultaneously in the same trial. The dolphin did not discriminate the 0.16 and 0.27 thick comparison copper discs from the standard but did discriminate the 0.32 and 0.64 cm thick discs from the standard at a 75% and 90% level, respec- tively. The targets used by Evans and Powell (1967) were acous- tically examined by Au and Martin (1988) at both normal and 140 from normal incidence angles. An incident angle of 140 corresponded to the incidence angle used by instrumented human divers in the study of Fish et al. (1976) using the same metal plates of Evans and Powell (1967). Au and Martin (1988) found that at normal incidence the backscattered echoes resembled the incident signal and did not seem to contain much useful information for discrimination. However as the incident angle increased to 100, the echoes began to have multiple highlights which could be used for discrimina- tion. Echoes from four plates at 140 incidence angle are show 4n Fig. 2. The differences in echo structure between the 22 cm thick copper standard and the 0.32 cm thick copper comparison target are obvious. Two scattering pro- cesses were suspected of producing the multiple highlight echoes: "leaky" Lamb waves and edge reflection of internally trapped waves. The two scattering processes are described schematically in Fig. 3. The trapped wave situation is for the longitudinal wave. Transverse waves of lower velocity will also be excited in the plates and converted to longitu- dinal waves at a boundary upon exiting the plate. The time of arrival of the secondary echo components is a function of the thickness and material composition (velocity of sound in the material) of the plates. The experiment of Titov (1972) in which a Tursiops wa3 trained to discriminate the wall thickness of steel cylinders was briefly described by both Ayrapet'yants and Konstantinov (1974) and Bel'kovich and Dubrovskiy (1976). Presumably, a two-alternative forced choice procedure with simultaneous target presentation was used. The outer diameter and length of the cylinders were 50 mm. The dolphin was trained to choose the thinner of two cylinders presented simultaneously at a range of 5 m. The dolphin was able to discriminate a wall thickness difference of 0.2 mm at the 75% correct re- sponse level. Hammer and Au (1980) performed three experiments (gener- al discrimination, wall-thickness and material composition discrimination) to investigate the target recognition and 361 14* CU-.22cm 33 STANDARD 84 CU - .32 cm 73 AL -.32 cm 71 BR -.32 cn SI I I I TIME (/ASEC) 7 (/.SEC) Fig. 2. The echoes and the envelope of the matched filter responses of four plates used in the experiment of Evans and Powell (1967). The relative arrival time of different highlights or echo components are shown in the matched filter responses. The incident angle was 140 from normal incident (from Au and Martin, 1988). aBEAINCIDE[NT I / z N ]A b EAINCIDENT BEAM REFLECTED BACKSCATTERED , EAM I'. 14 EDGE REFLECTION L"WEAAKVYE I I TRANSMITTED BEAM I TRANSMITTED BEAM Fig. 3. Schematics describing possible backscattering pro cesses involved with the plates used by Evans and Powell (1967), (a) depicts a leaky wave back- scattering mechanism and (b) depicts a trapped wave and edge reflection mechanism for longitudinal waves. Transverse waves will also be generated at each reflection point in the plate (from Au and Martin, 1988). 362 discrimination capability of an echolocating TursioDp. Two hollow aluminum cylinders, 3.81 cm and 7.62 cm in diameter and two coral rock cylinders of the same diameters, all 17.8 cm long were used as standard targets. The coral rock tar- gets were constructed of coral pebbles encapsulated in de- gassed epoxy. The targets were presented 6 m and 16 m from the animal's pen. The dolphin was required to echolocate the target and respond to paddle A if it was one of the aluminum standards or paddle B if it was one of the coral rock stan- dards. After baseline performance exceeded 95% correct with the standard targets, probe sessions were conducted to inves- tigate the dolphin's ability to discriminate novel targets varying in structure and composition from the standards. All the probe targets were cylinders, 17.8 cm in length. Two probe targets were used in each probe session and only 8 of 64 trials of the session were used for probe trials, 4 for each probe target. In the wall thickness experiment, Hammer and Au (1980) investigated the dolphin's ability to discriminate hollow aluminum probe targets with the same outer diameters but different wall thicknesses from the aluminum standards. The results showed that the dolphin could reliably discriminate wall thickness differences of 0.16 cm for the 3.81-cm O.D. cylinders and 0.32 cm for the 7.62-cm O.D. cylinder. A thickness difference threshold was not measured. The targets used by Hammer and Au (1980) were acousti- cally examined using simulated dolphin echolocation signals (see also Au and Hammer, 1980). The results for the 3.81 cm cylinders are shown in Fig. 4. The echo structure is shown on the left, the frequency spectrum in the middle and the envelope of the matched filter response on the right. The matched filter results are useful to determine the time of arrival of the various highlights in the echo. The aluminum standard is shown in part a and the comparison or probe targets are shown in parts b-e. From a visual inspection of the echo structures we can see that all of the targets have different arrival times for the secondary echo components, and therefore, different echo structures. Differences in echo structure probably also provided the major cue in the experiment of Titov (1972). Arrival time differences in the highlights may be per- ceived as a time-separation pitch (TSP), especially if the echo components are highly correlated. Humans when presented with a correlated pair of sound pulses perceive a pitch that is equal to l/T, where T is the separation time between pulses (Small and McClellan, 1963; McClellan and Small, 1965). In Fig. 5, the frequency spectrum of the first and second echo components for one of the aluminum targets used by Hammer and Au (1980) are overlaid on the total echo spec- trum. Note how well the total spectrum is described by the rippled spectrum for the first two echo components. Such a rippled spectrum is perceived as TSP by humans (Bilsen, 1966). Au and Pawloski (1992) performed a wall thickness dif- ference study in the free field and in the presence of mask- ing noise, using aluminum cylinders. Their primary emphasis was to determine the cues used by a bottlenose dolphin in 363 ECHO - FRElUENCY SECTRUPI ENP4ELtC MAOT0EDF ILTERR ESNM *I A(cid:127) /04 ,mw aill FREMANCYC 042 I b lip, '03?cmaII Q I - -a | Na , IW .l I FREMENCY (102) c lip, /048cmwL i FREQUENCY I KHZ 1 d liP 079crow41I io -n i (cid:127) + ii 4 + (cid:127) e FREaUENCY CK HZ I I. a~ I .. '" TiME FREQUENCYI KHZ I ii Fig. 4. Results of backscatter measurements of the 3.82-cm O.D. aluminum cylinders used in the wall thickness experiment of Hammer and Au (1980). Target IA, was one of the two aluminum standards. a o ~sac, $SEI II 1 -0o 20i -40 L-.- I 0 too 20U FREQUENCY (KHZ) Fig. 5. Example of the ripple spectrum from an echo. The dashed curve is frequency spectrum of the first two echo components or highlights. The target was a 3.82-cm O.D. aluminum cylinder. 364 performing the discrimination task. A standard cylinder of 6.35 cm wall thickness was compared with cylinders having wall thicknesses that differed from the standard by ± 0.2, ± 0.3, ± 0.4, and + 0.8 mm. All cylinders had an O.D. of 37.85 mm, and a length of 12.7 cm. The dolphin was required to station in a hoop while the standard and comparison targets, separated by an angle of ± 11 from a center line were simul- taneously presented at a range of 8 m. They found that the dolphin 75% correct response threshold occurred at wall thickness differences of -0.23 mm and +0.27 mm. The echoes from the standard and the 0.3 mm thinner comparison target for a typical dolphin echolocation signal are shown in Fig. 6. The animal was able to perform above 75% correct re- sponse threshold for this discrimination. The echoes from the standard and the 0.2 mm thinner comparison target are shown in Fig. 7. The animal performed below threshold for this discrimination. Let %, be the time between the first and second highlight for the echo from the standard target, and let vlbe the time between the first and second highlight for the comparison target, then the difference between the two times is Ar = , - rc. Values for Ar are given above the STANDARD o ao jis 0.604 a -.2s 0 No Ps 0 I I I a- ao o iso -4z 3. 1.4 kHZ FREQUENCY(KHZ) Fig. 6. Echo waveform, envelope and frequency spectrum for the standard and the comparison target having a wall thickness difference of -0.3 mm. The dashed curves for the envelope and the frequency spectrum are for the comparison target (from Au and Pawloski, 1992). 365