WO2000003586A2 - Fish feeding control in aquaculture on the basis of sound emitted by fish - Google Patents

Abstract

The present invention, relates to an electronic processor-based system for fish farming plants. The system receives the overall sound that is generated by fish in the defined cage or pond at the feeding time interval (16 and 17). On the basis of this data, the system calculates the dynamics of fish activity during feeding and generates a prediction of when to add a feed (13). In addition, the biomass of fish can be calculated from the sound data. Fish sound is a function of fish biomass. In this way feeding can be accurately adjusted to the population growth of the fish. Accurate knowledge of the fish biomass provides a method of indication the presence of disease (decreased biomass), enabling quick preventative action. Furthermore, knowledge of fish biomass can indicate theft or the presence of fish predators.

Description

FISH FEEDING CONTROL IN AQUACULTURE ON THE BASIS OF SOUND EMITTED BY FISH

FIELD OF THE INVENTION A simple non-invasive method for controlled feeding of fish with sound emitted by fish during a feeding period.

BACKGROUND OF THE INVENTION Intensive aquaculture is based on the supply of expensive food of high protein concentration (Cowey, 1979, see attached appendix for full reference). Feeding strategy is generally chosen according to the data of healthy and unstressed fish, and is adjusted to fit the fish varying appetite in the long term. Nevertheless, considerable short-term and occasional fluctuations in food intake cause daily and hourly deviations from the preprogrammed ration level. Partial control of feeding monitoring is carried out by on- demand- feeding techniques, where fish trigger the feeding by biting or hitting an object. In this case the food ration depends on the varying number of individuals that learn the mechanism of the feeder, and does not necessarily show the appetite of all of the fish. Another approach is associated with feedback in the feeding process, which is based on the detection of uneaten food pellets by hydroacoustic or sonar technology (Juell et al., 1992; Baras & Lagarbere, 1995). The application of these methods is limited because the installation and equipment costs are high. Moreover these methods give feedback with delay only and for that reason can not prevent waste of food.

Disruption of normal feeding behaviour is among the first sign of trouble in fish farming (Rice, 1990; Beitinger, 1990). The feeding process is influenced by stressors such as changes in temperature, pH and oxygen and ammonia concentrations. In cages the main stressors are diseases and parasites. The experienced farmer can identify stress intuitively through variations in swimming patterns, colour variations, apathy or reduction in feed consumption. Invasive measurements of stress are based on analysis of blood and tissue samples. Invasive measurements of electrophysiological signals from individual fish by miniature transmitters provide information of heart and muscle activity, respiration rates and locomotion. However, these methods are limited to experimental use only.

Non-contact measurement of electrophysiological signals by electrodes located in a holding chamber away from the fish has a common application only in water quality management.

Fish in both freshwater and marine environments produce sounds. The specific relation between acoustical variables and the physiological status of the fish have rarely been specified and only in a few cases have the sounds been correlated with a characteristic behavioural function such as spawning and mating (Tavolga, 1980; Tavolga et al., 1981 ; Ladich, 1990; Myrberg, 1993; Lobel, 1992; Crawford et al., 1997; Brantley and Bass, 1994). Generally, the sounds emitted by fish are divided into 3 types, stridulatory, swimbladder vibration and hydrodynamic sounds (Tavolga, 1980). The behaviour of the fish before and during feeding is affected by the physiological condition of the fish (Heath, 1987). Usually, feeding behaviour is composed of the motivation to eat (appetite), searching, detection, capture and ingestion of prey (Keenleysid, 1979). The feeding behaviour of cultural fish was studied by video, computerised monitoring of self feeders and underwater telemetry (Ponxin & Ruwet, 1994).

Swimming activity is directly connected to feeding activity through the intensity and frequency of the fish tail beat needed for changing directions and thus to the frequency of visits to the feeding sites (Stasko, 1976). Similarly, fish well being is also associated with the response time to food attraction, the time required to eat, the time from capture until the prey is swallowed, and the rate of prey lost after capture due to "spitting" (Nyman, 1981).

Feeding sounds are attributed to the activity rhythms of the fish. The search, detection and capture of prey are characterised by sounds of hydrodynamic origin such as fish movement in the water, the breaking of the water surface to catch food and then the leaving of the surface with a strong tail flip. The actual eating sounds include the jaw movement and the scraping of the teeth, or the gill rakers (Maniva, 1976; Fillips, 1989). These sounds were studied in an attempt to stimulate feeding behaviour of fish (Tavolga, 1980).

U.S. Patent No. 4509151 discloses an apparatus for locating, classifying and identifying underwater marine animals by means of the sound they make. The apparatus is made up of subarrays of multiple hydrophones, which are arranged along the length of each subarray in groups. The output of the groups can be selectively analysed by Doppler effect for characteristic fish sounds. The received sounds are classified and used for identification of newly found fish. PCT Application No. WO 9719587 discloses a computerised system for controlling feeding in a cultural fish tank with optical sensors located in the tank and sensing feed rate from the supply according to the needs of the cultured species. A funnel collects a sample of the system and the sensor is able to detect data relating to the amount of feed passing through the system and convey this to a control unit. U.S. Patent Application No. 4922856 discloses a system that receives signals in the form of strikes from fish, which control the operation of a scatter type feeder to feed the fish in response to the fish strikers. The apparatus transforms vibrations of strikes by even the tiniest of fish into electrical signals, which may be amplified and analysed by the control circuitry to control the operation of the scatter type feeder. U.S. Application No. 5133287 discloses a continuous feeding process provided in a flow-through container where the input oxygen content of the water flow is measured and a signal is sent to a system processor. The system processor controls the output of a feeder, which supplies a continuous controlled amount of feed to the fish, with the amount of the feed being directly related to the oxygen content of the water. The output oxygen content is also measured and a signal is sent to the system.

It is evident that the background art do not provide solutions to the problems associated with control of fish appetite during feeding and the prediction of the moment to stop feeding. There is no disclosure of monitoring the immediate response of a feeding system to short term stress of the fish population during feeding. There is no method in the prior art related to reducing metabolic waste and associated food wastes, nor a method to optimise food conversion ratio or feeding to provide the best food assimilation. There is no effective method for preventing theft, assessing biomass or disease. In addition, acquisition of data by systems described in the prior art are affected by wind, waves and currents making their use limited.

There is therefore a need for a system and method, such as is disclosed in the present invention, to provide a solution to the aforementioned problems.

SUMMARY OF THE INVENTION The system of the present invention, is an electronic processor-based system for fish farming plants. The system receives the overall sound that is generated by fish in the defined cage or pond at the feeding time interval. On the basis of this data, the system calculates the dynamics of fish activity during feeding and generates a prediction of when to add a feed. The data gathering, processing and analysing is based on the Digital Signal Processing (DSP) technology.

The system of the present invention reduces production cost and increases fish growth. The system minimises feed wastage by managing the feeding on the bases of estimation of fish feeding activity. This results in a decrease of food expenses and environmental pollution by giving the fish the amount of feed needed for optimal feed assimilation.

Furthermore, the system of the present invention responds in real time to all stimuli that excite or stress fish and affect the fish appetite. Hence, the system controls the fish getting the right amount of feed in all events and provides the best food conversion ratio.

In addition the system of the present invention can be used to recognise the physiological state of the fish, such as stress and fish diseases and assessment of fish biomass in each cage.

Moreover, the system of the present invention can be used for protection against theft and predators.

The system, is capable of being operated with most types of fish farms using automatic feeders with automatic and manual modes, either in a mobile form or in a fixed form. Although the invention described here is described for the use of fish, it is additionally contemplated that the present invention will be useful for other types of aquatic and semi-aquatic animal life.

In a first embodiment, the present invention provides a method of feeding fish, the method comprising the steps of: (a) providing a receiver; (b) placing the receiver in a fish farm; (c) receiving a fish sound from the fish with the receiver; and (d) determining an amount of food to be placed in the fish farm according to the fish sound.

In a preferred embodiment the method of feeding fish further comprises the step of converting the fish sound to an electrical signal. In a preferred embodiment the method of feeding fish further comprises the steps of: (a) interfacing the electrical signal to a data processor; (b) processing the electrical signal to obtain data; and (c) controlling amount of food given by a feeder in the fish farm.

In a preferred embodiment the method of feeding fish further comprises the step of calculating the dynamics of fish activity during feeding through analysis of data and generating a prediction of feeding time.

In a preferred embodiment the step of processing the electrical signal to obtain data, includes the step of statistical analysis treatment.

In a preferred embodiment the step of processing the electrical signal to obtain data further comprises the steps of: (a) filtering the electrical signal to produce a filtered signal; and (b) compressing the filtered signal to produce a compressed signal, wherein a matrix of the compressed signals is proportional to the input signal intensity and is a function of accumulated amount of food for each fish and elapsed time of one feed portion. In a preferred embodiment the receiver is an acoustic sensor.

In a preferred embodiment the acoustic sensor is a hydrophone.

In a preferred embodiment the receiver is directional.

In a preferred embodiment the receiver is omnidirectional.

In a preferred embodiment the receiver includes an array of receivers. In a preferred embodiment the fish farm is selected from the group consisting of ponds, in shore cages, off shore cages and aquariums. In a preferred embodiment the fish is selected from the group consisting of aquatic and semi-aquatic animal life.

In a preferred embodiment the fish is selected from the group consisting of carp, tilipia, salmon, sea bream, trout, miltfish, tuna, eels, crabs, catfish and shrimps. In a preferred embodiment the fish sound is a function of the physiological state of the fish.

In a preferred embodiment the step of controlling amount of food given by a feeder in the fish farm is performed by interfacing the data with the software of the feeder.

In a preferred embodiment the power source of the feeder is regulated according to the data.

In a preferred embodiment the step of controlling amount of food given by a feeder is performed manually.

In a preferred embodiment the feeder includes an automatic feeder in automatic mode. In a preferred embodiment the feeder includes an automatic feeder in manual mode.

In a preferred embodiment the feeder includes a manual feeder in manual mode.

In a preferred embodiment the feeder is in a mobile form.

In a preferred embodiment the feeder is in a fixed form. In a preferred embodiment the physiological state of the fish is selected from the group consisting of stress and fish disease.

In a second embodiment the present invention provides a feeding apparatus responsive to fish sound from fish comprising: (a) at least one receiver to receive and transform the fish sound to an electrical signal; (b) a data acquisition system to collect the electrical signal from the receiver; (c) a data processor for calculation of feeding time, according to the electrical signal from the data acquisition system; and (d) a feeder responsive to the calculation of the data processor.

In a preferred embodiment the feeding apparatus further comprises a preamplifier to amplify the signal. In a preferred embodiment the feeding apparatus further comprises a local system controller. In a preferred embodiment the feeding apparatus further comprises a base station connected to the local system controller.

In a preferred embodiment the fish is selected from the group consisting of aquatic and semi-aquatic animal life. In a preferred embodiment the fish is selected from the group consisting of carp, tilapia, salmon, sea bream, trout, miltfish, tuna, eels, crabs, catfish and shrimps.

In a preferred embodiment the at least one receiver is a calibration acoustic sensor.

In a third embodiment the present invention provides a method of calculating fish biomass in a fish farm, the method comprising the steps of: (a) providing a receiver; (b) placing the receiver in the fish farm; (c) receiving a fish sound from the fish with the receiver; and (d) determining the fish biomass according to the fish sound.

In a fourth embodiment the present invention provides a method of monitoring the presence of non-fish bodies in a body of water, the method comprising the steps of: (a) providing a receiver; (b) placing the receiver in a body of water; (c) receiving a fish sound from the fish with the receiver; (d) receiving a non-fish body sound from a non-fish body with the receiver; and (e) establishing the presence of the non-fish body according to the non-fish body sound.

In a preferred embodiment the non-fish body includes a thief.

In a preferred embodiment the non-fish body includes a predator. In a fifth embodiment the present invention provides a security apparatus responsive to sound from fish and sound from non-fish bodies comprising: (a) at least one receiver to receive and transform the sound from fish and the sound from non-fish bodies to an electrical signal; (b) a data acquisition system to collect the electrical signal from the receiver; and (c) a data processor for establishing the presence of the non-fish body according to the sound from the non-fish body.

In a preferred embodiment the receiver is a calibration acoustic sensor.

In a sixth embodiment the present invention provides a method of determining the physiological state of fish, the method comprising the steps of: (a) providing a receiver; (b) placing the receiver in a fish farm; (c) receiving a fish sound from the fish with the receiver; and (d) determining the physiological state of the fish in the fish farm according to the fish sound. In a preferred embodiment the physiological state is selected from the group consisting of stress and fish disease.

In a preferred embodiment the physiological state of the fish is a function of the fish sound. The term 'non-fish body' as used herein refers to any body in the water with the exception of the fish in the fish farm.

BRIEF DESCRIPTION OF THE DRAWINGS FIG la shows a general block diagram of a possible way of hydroacoustic data acquisition, a processing and transfer system to control fish feeding in open water cages with automatic feeders;

FIG lb shows a general block diagram of a possible way of hydroacoustic data acquisition, a processing and transfer system to control fish feeding in open water cages and land based ponds with mobile feeders; FIG 2 shows a simple model of sound production during feeding by a school of fish;

FIG 3 shows the experimental set up during fish feeding sound pickup; FIG 4 shows the schematics of fish feeding sounds processing during one feeding portion; FIG 5 (a,b,c,d) shows examples of acoustic images of fish feeding behaviour in 24 fish (Carassius auratus), which were measured in one compartment in three successive days with same measurement conditions. Relative power of fish feeding sound E(kj) is shown as a shade of grey. 5d shows the scale of E(kj) values;

FIG 6 (a,b,c,d) shows acoustic images related to FIG 5 (a,b,c,d) after exponential smoothing with damping coefficient 0.5. Relative power of fish feeding sound E(kj) is shown as a shade of grey. 6d shows the scale of E(kj) values;

FIG 7 (a,b,c,d,e) shows basic sound power images of fish feeding sounds of control and stressed fish in each of four similar compartments. Relative power of fish feeding sound E(kj) is shown as a shade of grey. (a,b) is data obtained from control fish in compartments a and b. (c,d) is data from fish after heavy metal stress exposure three weeks before, in compartments c and d. (e) shows the scale of E(kj) values; FIG 8 (ab,cd) shows the relative power of fish feeding sounds E(k,j) as a function of number of time interval k in successive feeding portions j. (ab) data are obtained from compartments a and b with control fish ( — ) compartment a, ( — ) compartment b. (cd) data are obtained from compartments c and d with stressed fish ( — ) compartment c, ( — ) compartment d;

FIG 9 (ab,cd) shows the relative power of fish feeding sounds E(k,j) as a function of number of feeding portions j in successive time intervals k. (ab) data are obtained from compartments a and b with control fish ( — ) compartment a, ( — ) compartment b. (cd) data are obtained from compartments c and d with stressed fish ( — ) compartment c, ( — ) compartment d;

FIG. 10 shows the general experimental arrangement for examples 2 and 3;

FIG. 11 shows raw signals during fish feeding with noise from boats passing near the cage;

FIG. 12 shows typical signals during fish feeding without additional noise; FIG. 13 shows ambient noise without fish feeding;

FIG. 14 shows a diagram of the dispersion of amplitude-frequency parameters that were received by the sensor installed in the center of the sea cage;

FIG. 15 shows relative levels of sound inside fish sea cages obtained in two separate feeding times versus elapsed time. In the cage was approximately 94000 fish, species Denis, average mass of each fish was about 120g;

FIG. 16 shows relative levels of sound inside fish sea cages obtained at time of feeding versus elapsed time. In the cage was approximately 57000 fish, species Denis, average mass of each fish was about 460g; and

FIG. 17 shows relative levels of sound inside fish sea cages obtained at time of feeding versus elapsed time. In the cage was approximately 42000 fish, species Denis, average mass of each fish was about 620g.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention is directed to a novel, simple, noninvasive and inexpensive method of acquiring information by using sound emitted by fish aggregation during a feeding period. From the sound data, the system calculates the dynamics of fish activity during feeding and generates a prediction of when to add a feed.

In addition, the biomass of fish can be calculated from the sound data. Fish sound is a function of fish biomass. In this way feeding can be accurately adjusted to the population growth of the fish. Accurate knowledge of the fish biomass provides a method of indicating the presence of disease (decreased biomass), enabling quick preventative action. Furthermore, knowledge of fish biomass can indicate theft or the presence of fish predators. In this way the present invention can be used as a security apparatus.

Sounds produced by fish during feeding are a result of locomotory and actual eating activity, which depend on appetite. Fish appetite is measured by voluntary food intake (Beitinger, 1990) and is affected by stress. Therefore, sound data from the present invention can be used to monitor stress and the physiological state of the fish. When food is given in portions, the appetite of the fish will decrease after each portion. The same tendency should be observed in parameters of fish feeding activity and in sound level related feeding activity.

Fish appetite A(t) is expressed by a fish feeding activity characteristic that is a function of current sound intensity I(t) emitted by fish during feeding. The relationship between the sound intensity emitted by fish during feeding and fish appetite is defined as:

I(t)=KA(t) (1) where K - calibration coefficient of the measurement system.

There is a dependency between fish appetite A(t) and amount of food ingested by fish in the unit of time q. The greater the appetite of fish, the greater the amount of food ingested by fish in the unit of time. If one assumes that the amount of food ingested by fish in the unit of time is proportional to the fish appetite then: q=kA(t) (2) where k - coefficient of proportionality with dimension (1/sec). The coefficient k is usually variable with time and depends on many factors, such as the amount of ingested food, fish size and species, environmental parameters, fish physiological state etc. The appetite of the fish changes during the feeding. It is not constant, but varies as the fish ingest the food. During the initial moments of feeding the fish are hungry and the appetite of the fish is at a maximum and as the feeding activity continues the food is ingested while the appetite reduces quickly. At a later time the fish become satiated and their appetite further reduces. During the periods of feeding, the quantity of feed that will be apportioned is distributed over a uniform rate throughout the period. This food is completely ingested by fish. The relationship between the reduction of fish appetite and the amount of food ingested by fish per unit of time is defined as:

-dA/dt = q (3) The minus sign indicates that the ingestion of food by fish leads to decrease of fish appetite. Therefore

-dA/dt = kA (4) and

A(t) = A°e^'kT (5) where A - fish appetite at the initial moment of feeding.

In real conditions the coefficient A is variable and depends on the amount of ingested food, stress of fish, fish size, species, environmental parameters, fish physiological state etc.

Substituting (5) into (1):

10Lg(I) = C - Gt (6) where: C = 101g(KA_o) G = 10k.lg(e)

(6) shows that the level of fish feeding sound intensity is reduced linearly versus elapsed time of uniform feeding of fish. Coefficients C and G are defined experimentally in accordance with size and species of fish and real acoustic situations in the cages. Interpretation of these parameters could also be used as an indicator of fish stress and disease.

The present invention is an electronic processor-based system for fish farming plants. It is a tool for reducing production cost and increasing fish growth. It minimizes feed wastage by managing the feeding on the basis of estimation of fish feeding activity. This results in lower food costs and environmental pollution by giving the fish the amount of feed needed for optimal feed assimilation. The disclosed invention responds in real time to all stimuli that excite or stress fish and affect fish appetite. The system controls the fish obtaining the correct amount of feed under all conditions and provides the best food conversion ratio. The information obtained using this system with regards dynamics of fish feeding behavior could assist in recognition of fish diseases and assess fish biomass in each cage. To calculate fish biomass in a fish farm an acoustic receiver, such as a calibration acoustic sensor must be provided and placed in the fish farm. The receiver will receive a fish sound from the fish and the fish biomass can be determined according to the fish sound.

The system of the present invention can also be used to prevent theft of fish. Using the same system for hydroacoustic data acquisition, as shown in Figures 1 or 2, the receiver, preferably a calibration acoustic sensor, which can be permanently on-line can monitor in the same way as fish sound is monitored, non ambient noise of ships, divers or of other moving non-fish bodies. The method comprises providing a receiver, such as an acoustic sensor and placing it in the body of water. The sensor will receive sound from the fish and sound from a non fish body in the water. In this way the presence of non-fish bodies in the water can be established. Furthermore, the presence of thieves or predators can be detected. Therefore, the present invention also provides a method for protection against theft and against predators. The system can be used with most types of farms. The system may utilize user friendly software and operation of the software is extremely simple and easy to learn. The sensors and system controller unit can be conveniently installed in the most suitable place.

It is to be understood that the present invention is not limited in its application to the details of construction and the arrangement of the components set forth in the following description. The invention includes other embodiments and can be practiced or implemented in various ways. Also it is to be understood that the phraseology and terminology employed herein is for the purpose of description only and should not be regarded as limiting.

FIG la shows a general block diagram of an exemplary system for hydroacoustic data acquisition, a processing and transfer system to control fish feeding in open water cages with automatic feeders. A feeding system 10 for an open water farm plant 22 includes at least one acoustic sensor 16 placed in water 21 outside at least one cage 18 containing fish 19 in water 20. The system 10 is able to gather data from a number of cages 18, much larger than the number of acoustic sensors 16. Pre-amplifiers 17 are used when the distance from the acoustic sensor 16 location is greater than about 10 m. In this case they are integrated into two acoustic sensors 16 and can be used to provide a cable length of about 150 m. The acoustic sensor 16 is connected to a cage mounted electronics supply 13 which includes a DAQ system 14 and a system controller 15.

The cable connected to the acoustic sensor can optionally extend out of the water to a data and collection analysis unit. Preferably, the collection and analysis unit include a data acquisition board, which collects data from the acoustic sensors 16, and all signal processing is preferably either done in digital, or in analogue or in any other suitable way. A system controller 15 uses compatible software for signal logging, processing, data presentation and communication to the host computer 11. The software can be written in any suitable programming language, either an object oriented language, including but not limited to Labview and Matlab, or a regular programming language, including but not limited to C/ C++ and Visual Basic languages. The selection of which could easily be made by one of ordinary skill in the art. The software should be compatible with the operating system of the computer or other data processor, which is running the software. In a module design the system controller 15 has a plug in for connection via standard communication modem to an external monitor and or a computer such as a base station 11 or host computer. Output data and users interface are represented in graphical and numerical forms. Control of the feeding regime by the feeder control unit 10 is realised through an interactive menu. Communication can be realised in serial, for example RS232 and or RS485 standard, or in parallel, for example Ethernet, by wire or radio links 12 and depends from the real situation. Unlimited distance from the local system controller 15 to the base station 11 can be used. The information can be relayed to the feeder in several ways. It can be interfaced with the software of the feeder, or it can be connected to an on/off switch of any type of feeder. Alternatively, the information can be relayed by monitor for manual feeding.

FIG lb shows a general block diagram of an exemplary system for hydroacoustic data acquisition, a processing and transfer system to control fish feeding in open water cages, in in land and land based ponds with mobile feeders. A mobile feeding unit 23 includes one acoustic sensor 16 placed inside a fish cage 18 containing water 20 and fish 19. In the open sea the acoustic sensor is preferentially not placed inside the cage. The system 23 is able to gather data from each cage 18.

The acoustic sensor 16 is connected to a mobile unit mounted electronics supply 24 which includes a DAQ system 14, a system controller 15 and a base station (local users interface) 11. Preferably, the collection analysis unit includes a data acquisition board which collects data from the acoustic sensors 16, and signal processing is preferably done in digital. Signal processing is also possible in analogue or using any other suitable method. A system controller 15 uses compatible software for signal logging, processing, data presentation and communication. In a module design the system controller 15 has a plug in for connection via standard communication modem to an external monitor and or a computer such as a base station 11 or host computer. Output data and users interface are represented in graphical and numerical forms. Control of the feeding regime by the feeder control unit 10 is realized through an interactive menu. Communication can be realized in serial, for example RS232 and or RS485 standard, or in parallel, for example ethernet, by wire or radio links 12 and depends from the real situation. Unlimited distance from the local system controller 15 to the base station 11 can be used. The information can be relayed to the feeder in several ways. It can be interfaced with the software of the feeder, or it can be connected to an on/off switch of any type of feeder. Alternatively, the information can be relayed by monitor for manual feeding.

FIG 2 shows a flow chart diagram of an exemplary model for sound production during feeding by a school of fish. The first step is the appearance of food in the water 36. There is a delay 38 before the fish 34 detects the food. After this time interval, the fish detect the food 25 and there is a second delay 26, before the food is captured by the fish 27. This produces a sound 28. After capturing the food 27 there is another delay 29 before ingestion of the food 30, which produces another sound 31. The sound from capturing the food 28 and ingestion 31 causes a response in a sound propagation channel 32 and this is detected by the acoustic sensor 16. Ambient noise 33 is also detected by the acoustic sensor 16. The steps are repeated for most fish 35, through fish N 37, including an individual fish designated 34, being a number 2-N. Therefore, the noise from all the fish as a measurement point, will be represented as a sum of random sounds emitted by the individual passing through the water to the acoustic sensor.

FIG 3 shows an exemplary model of the experimental set up during fish feeding sound pick-up. The experiment is carried out in a system composed of two tanks, 45 and 51 containing fish 49, an acoustic sensor 16, an acoustic sensor amplifier 54, DAS-1400 computer interface board 55 and a PC 56. Each tank is divided by a vertical wall 46 to give two compartments 48 and 57. The acoustic sensor 16 is placed in every compartment in the same position and receives and transforms sound to an electrical signal. The signals pass to a lowpass amplifier 54 with high input impedance. The signals are then interfaced by DAS 1400 55 to the PC 56.

FIG 4 shows an exemplary state diagram of fish feeding sounds processed during one feeding portion. The processing can be done using either digital or analogue processing or in any other suitable way. An acoustic signal shown asp_tJ (t,r) is emitted by fish / at a distance r from the acoustic sensor. The signal is received by the acoustic sensor at moment t in relation to the dispersion of a portion of foody, where y represents number of feeding portions, / =T ,2....J.

The signal is transformed by the acoustic sensor and preferably the preamplifier to an electrical signal shown as U_j(t), where y and t are as before.

The electrical signal is filtered to form a filtered signal w_y(t), where U_j (t) is a convolution of u t) and h(t), such as ^ (t) =u/t) ®h(t), with a filtered signal having a pulse response h(t). The pulse response of the filter is a Fourier transformation h(t)oH(f).

The filtered signal is compressed to form a compressed signal {E(kJ)} , according to the following equation: kτ+ τ

T kT where the matrix of values is proportional to the input signal intensity, in time interval T, during portion of food j. FIG 5 (a,b,c,d) shows exemplary examples of acoustic images of fish feeding behavior in 24 fish (Cαrαssius αurαtus), which were measured in one compartment in three successive days with the same measurement conditions. Basic sound power images { E(k,j) ) > { E(k,j)}_Th of fish feeding are shown in FIG 5(a,b,c). {E(kj)} are matrices of data that were obtained in one compartment with normal fish, {E(kj)}τ_h is threshold level and FIG 5 shows the relative amplitude of (E(kJ)} . Distribution of shading is determined by the ratio of maximum value to threshold value (dynamic range) and therefore in the image it is not possible to see the variation of data smaller than the color resolution of the drawing. A dynamic range of 20 dB was chosen in order to get an overview of the data. The images 'a,b,c' are very similar and the correlation coefficients for these three matrices {E(kj)} demonstrate that there is no reliable differentiation from this raw data. FIG 6 (a,b,c,d) shows exemplary acoustic images related to FIG 5 (a,b,c,d) after exponential smoothing with damping coefficient of 0.5. Exponential smoothing analysis predicts a value based on the forecast for the prior period, adjusted for the error in that prior forecast. The equation uses the smoothing constant a, the magnitude of which determines how strongly forecasts respond to errors in the prior forecast as: F₁₊rF_t + _Ω(A_rF_t)

Where: F_t= previous smoothing value;

F_t+ι=current smoothing value; A_t= previous value After exponential smoothing there is observed an obvious tendency in feeding sound behavior with high correlation. There is a decrease in sound level with decreasing fish appetite.

FIG 7 (a,b,c,d,e) shows exemplary basic sound power images {E(kj)}> {E(kj)}_Tι₁ of fish feeding sounds of control and stressed fish. {E(kj)} are matrices of data that were obtained in compartments 'a,b,c,d' accordingly. 7e shows the relative amplitude of {E(kj)} . Images 'a' and 'b' of the control groups of fish are very similar as are 'c' and 'd' the two groups under stress.

FIG 8 (ab,cd) shows an exemplary example of the relative power of fish feeding sounds E(k,j), as a function of number of time intervals k, in successive feeding portions j. For control fish the values of E(kj) are reduced gradually with successive feeding portions (ab). With stressed fish the same order is not observed and higher peaks appear after certain delays (cd). This is explained by the ingestion activity of stressed fish being low and they do not have enough time to ingest food during one time period.

FIG 9 (ab,cd) shows an exemplary example of the relative power of fish feeding sounds E(k,j), as a function of number of feeding portions j, in successive time intervals k. Sound power associated with actual eating (k>3) from stressed fish (cd) is more than twice as (3 dB) small than from control fish (ab).

The present invention may be better understood with reference to the examples and the accompanying description.

Example 1

A simple model is suggested for understanding the origin of the feeding sounds in fish aggregation (Fig. 2). The delays between the stages of feeding for every individual fish and among fish do not follow a general pattern. Therefore, noise from all fish in a measurement point will be represented as a sum of sounds emitted by the individual fish passing through water to the acoustic sensor. For reliable detection of fish sounds, it is necessary to take into account some complex problems involved in passive acoustic systems, such as separation of fish eating sounds from ambient noise and conditions of sound propagation in a limited volume. In a limited volume, fish sounds can produce additional noises, as a result of reverberation and reflection sounds from walls and obstructions. Clearly, all these sounds are added in an omnidirectional acoustic sensor with a random amplitude and phase and therefore, it is very complex to select useful information from their patterns. If the fish appetite is to be considered as the main factor that influences the temporary, locomotive and actual eating response of the fish, then the main information will be in the patterns of space distribution of the intensity of feeding sounds, relative to the moment of giving the food.

If the food is given in equal time intervals in the same quantity, a new variable is obtained that is linked to fish appetite and accumulated amount of given food. Information can be obtained from the time-amplitude-frequency domain of a series of sound intensity emitted by fish, related to the number of uniform portions of food "f and elapsed time interval "k" of each portion. The effects of reverberation and reflection on the field of fish sounds are decreased in a limited fish volume. For example, for a distance of 100 m, the time of propagation of sound is about 0.15 s smaller than the stage of fish feeding (> Is) and the condition of sound propagation is practically not influenced by the place of sound on the time axis. When the distance is >1000 m the aforementioned values of acoustic phenomena's are small. It is simpler to cut off the ambient noise, as it is possible to receive the signals in a fixed time 'window'.

When the food is given in equal amounts in every feed portion, they is proportional to the accumulated amount of food W, that has been eaten by N fish. To exclude the influence of the scale from N, data was recalculated using mass of food vv, for each fish as w=W/N. This results in the matrix {E(k, j)} , where E is proportional to sound intensity and is a function of two variables, such as accumulated amount of food for each fish and elapsed time of one feed portion. MATERIALS AND METHOD The experiment was carried out in a system composed of two tanks containing fish, hydrophone, amplifier, computer interface board and PC (Fig. 3). Each tank (70 x 90 x 90 cm) was divided by a vertical wall to give two compartments, each with 24 fish. An ornamental carp specie Carassius auratus (goldfish), 8 to 12 cm long of both sexes, was used. The fish were 6 months old and were from the same batch. Before the start of the experiment, the fish were acclimatised for 6 weeks in each of the four compartments. During the acclimatisation time, oxygen levels were kept at 8+0.5 ppm, water temperature at 23 to 24°C and the ammonia level was undetectable. Dry pellets were used for feeding the fish once a day. Three weeks before the experiment, fish from one compartment per tank were exposed to heavy metals Cadmium (Cd, 0.5ppm)+ Lead (Pb,0.3ppm). These metals and concentrations were used to initiate a stress response.

During measurement, the water filters and aeration system were turned off to remove additional noise. The food was given in equal time intervals (-130 ms) in equal portions (weight 0.01 g per fish, -30 pellets) in the same area of the compartment. The number of pellets had to be chosen, so the fish could capture all portions concomitantly. The measurement system consisted of an acoustic sensor 8104 B&K, amplifier,

Keithley MetraByte/Asyst interface board DAS- 1401, IBM PC Pentium 100 with the following connections. The hydrophone was placed in every compartment in the same position and received and transformed the sound to an electrical signal. From the hydrophone, the signals passed to a lowpass amplifier with an high input impedance. The amplifier had a gain of 50 dB and frequency range 0-25 kHz. Level of feeding sounds in the frequency range >1 kHz was at least 10 dB higher than the ambient noise and therefore special noise avoid action was not done. After amplifying, these signals were interfaced by a DAS 1401 to a computer. The initial preprocessing and acquisition data of {E(k, j)} were carried out by VIEWD AC software and finally by MATLAB software. Digital filtering of signals was performed in a specific frequency range determined by size and form of the fish tank.

The recording of sounds (time of recording signals was about 125 s) was begun immediately after a food portion was given to a group of fish. The power E of the signal in every time interval was equal to 1/10 of the time record and was calculated in real time and was written to file. In order to exclude the distortion of the signal during preprocessing, the overall time of signal processing was minimised to compare with the feeding interval. This procedure was repeated to produce a matrix of values of E(j,k) for the J portion of food and the K time intervals. An extrapolation function was then applied to produce a more appropriate smooth image for visual analyses. The general scheme of the sound process is shown in Fig.4 where: p_tJ (t, r^_>) is the acoustic signal emitted by /-fish with a distance r^_> from the hydrophone and received by the hydrophone in a moment t in relation toy portion of food; j =1,2 .... J represents number of feeding portions;

N is number of fish emitted sounds in every moment in relation toy portion; u/t) is electrical signal after hydrophone and preamplifier; U_j (t) =U_j(t) 0 'h(t) is signal after filter having pulse response h(t); h (t) <=>H (f) is a Fourier transformation of amplitude frequency response of filter;

kτ+ r

T kr where the matrix of values is proportional to the input signal intensity, in time interval T, during portion of food j. k = 0,1,....K number of time intervals during one portion of food. The measurement was begun from the moment of placing food in the water body. In order to increase low frequency components of sound power related to fish appetite, the affect of the sound data was decreased, such as capture of pellets by one or two fish from the surface in the random moments of the feeding interval, by exponential first order smoothing.

One test performed was carried out with one group of control fish in one compartment during three successive days, to check the regularity of the feeding sound behaviour and the correlation with the amount of eaten food. A second experiment performed with all four fish groups, was to detect the influence of stress on the feeding sound pattern. The measurement was carried out for every compartment individually. Concomitantly, the record of signals produced a visual detection of the fish behaviour. The time when the fish stopped carrying away the pellets from the surface of the water was considered as cessation of feeding.

RESULTS AND DISCUSSION The collected data of fish feeding sound power were processed into a suitable form for analysis and conclusions. Basic sound power images (E(k,j)} >{E(k, j)}_Th of fish feeding are shown in Fig. 5 (a,b,c). {E(k,j)j are matrices of data that were obtained in one compartment with control fish. {E(k,j)}_τh, is threshold level and Fig. 5 shows the relative value of fish feeding {E(k )}. Distribution of shading is determined as the ratio of maximum value to threshold level (dynamic range), and therefore it is not possible to see from the image the variation of data smaller than the colour resolution of the drawing. In Fig. 5 a dynamic range of 20 dB was chosen to get an overview of the data. The images 'a,b,c' are very similar with similar correlation coefficients (Table 1) for these three matrices {E(k,j)}, which demonstrates that there is no reliable differentiation from this data. After exponential smoothing, an obvious tendency in feeding sound behaviour with high correlation is observed (Fig. 6) (Table 1). This result demonstrates regularity of feeding sound behaviour, with a tendency of decreasing sound level with decreasing fish appetite. This means that it is possible to use this data for prediction of fish feeding activity and hence for operative control of the feeding process.

Table 1 : Correlation coefficients of fish feeding sound power E(k, j) of data obtained in one compartment during three successive days under uniform measurement conditions and related to Fig. 5 and Fig. 6.

From Fig. 6 it can be seen that on smoothing the matrix (E(kj)) there is a tendency towards reduction of magnitude with increasing 'y' and '&' values. The tendency relating to amount of given food (j) can be explained by the following reasons. Firstly, there is a reduction of fish feeding activity (locomotory and actual eating activity) with a lowering of appetite, as a result of increasing the amount of eaten food. Secondly, all fish do not have the time to ingest food during the time interval between feeding portions. Thirdly, capture of food by fish is not uniform for a specific distribution of food, due to competition among a member of a fish population when the amount of food is restricted (Brett, 1979). In this case, the more active fish capture and eat food first and when their activity is lowered, the less active fish begin eating.

There is a tendency for fish feeding activity to decrease with elapsed time after capture of the food. The greatest magnitudes of sound power (Fig. 6) are in the second- third time interval in every portion of food i.e. with a delay relative to the moment of putting food in the compartment. According to the visual observation of fish feeding behaviour, these sounds have hydrodynamic origin and are related to fish which capture food near the surface. Sound power in the next time interval is considered due to actual eating, because all fish in general are in that depth without active movement. Basic sound power images {E(k,j)} >{E(k,j)}_Th of fish feeding sounds of control and stressed fish are shown in Fig. 7 (a,b,c,d). {E(kj)} -are matrices of data that were obtained in compartments 'a, b c, d' accordingly, Fig.7(e) shows the relative amplitude of {E(k,j)}. Images 'a' and 'b' of the control group of fish are very similar, as are 'c' and 'd' from the two groups under stress. It is very simple to recognise task visual symptoms of essential differences between values of sound intensity (level of darkness in Fig. 7e) and square of signal image area.

The data values of sound power matrices related to Fig. 7 were transformed to a series and were represented as a function of food portion number 'k' and time interval 'j' (Fig. 8 and Fig. 9). This effect to a certain extent explains some difference in the feeding sound power series of control groups and stressed fish.

For control fish the values oϊE(k,j) are reduced gradually in successive feed portions (Fig.8(ab)). For stressed fish the same order is not observed and greater peaks appear after a certain delay. This means that ingestion activity of stressed fish is low and they do not have enough time to ingest food during one time period. It is also possible to see this in Fig.9. Sound power associated with actual eating (k>3) from stressed fish (Fig. 9(cd)) is more than twice as (3 dB) small as control fish (Fig. 9(ab)) and does not have a tendency to reducing. The next peak of capture activity is shifted compared with control fish and it seems likely that distance between peaks depends on the level of stress. It is important to know the optimum amount of fish in order to get rid of this phenomena.

It is assumed that increasing sound intensity for control fish is directly attributed to temporary patterns of feeding behaviour. Usually capture and ingestion sounds emitted by each fish during feeding are a random series of burst pulses with a specific burst rate, carrier frequency and pulse amplitude. Therefore, for the same conditions of number of fish, amount, quality and type of pellets, the signal received by the acoustic sensor will be dependant on the aforementioned patterns of individual feeding sounds. It is known that stress increases delays of fish response to food attraction, capture and required time to eat (Nyman, 1981) and therefore there is a decrease in the amplitude of the signal. CONCLUSION

The application of passive acoustic imaging of fish feeding enables one to extract adequate information about fish conditions that can be used both for control of feeding and for warning of stress. This study demonstrates that passive acoustic imaging creates a new potential for accurately monitoring feed uptake in fish that can lead to improved management of fish stock farms. The passive acoustic method provides the information directly from the fish and reflects its physiological status without any artificial impact due to the measurement system.

There is a real need for a reliable, sensitive and realtime device based on acoustic imaging, that obtains quantitative data about the relationship of image patterns and fish voluntary food consumption, assimilation efficiency levels and fish stress.

Additional information can be obtained by using directional acoustic or a distributive fibre acoustic system, that picks up sound from a defined part of the aquatic medium. Such spatial filtration will enhance recognition of capture and ingestion sounds, to increase the ratio of fish feeding sound to ambient noise and therefore increase the reliability of fish feeding control.

Example 2: Measurement of fish feeding sounds and ambient noise in real sea cage conditions

The main objectives of the experiment were as follows: (1) To test the efficiency of underwater transducers, underwater connectors and workstations in real sea cage conditions; (2) Acquisition and saving of raw data of ambient underwater noise that exists in sea cages; (3) Acquisition and saving of raw data from sound produced by different species and sizes of fish during feeding for future analyses; (4) Using this data to choose algorithms for signal processing and for parameters of signal filtering in time-frequency domain.

The experiment was performed using the experimental arrangement of Figure 10 in a cage with diameter 10m and depth 15m containing 82,000 Sea Bream (Denis) with average weight of 448g. The water acoustic sensor model 8226 in series was located outside the cage at a distance of about 2 m from the cage and in the centre of the cage at fixed depth of lm, 3m, and 5m. During measurement, recording of all instances of fish jumping and passing boats were marked on the history of the recorder device. Ambient noise and / or fish feeding sound was picked up by the sensor, which was coupled through connection box to plug and play DSP board that was installed to a rugged workstation (WS) Pentium II 300 MHz. The predetermined number of raw signal data points received by one or more acoustic sensors were fed to an intermediate memory buffer of WS that provided continuous acquisition of data. Raw data was written at different times with and without feeding, during feeding and during times when boats passed near the cage. The recorded data was analysed in the amplitude-frequency time domain. These analyses were performed on specially developed software. The software for recording and analysing data was written by object oriented language - Labview 5.

Results

(a) Location of acoustic sensor

Sonar equation for passive sonar is a relationship between certain quantities termed sonar parameters: Equation 1 : SL - TL = NL + DT

Where: SL = source level of the target; TL = 20 lg ( R ) transmission loss; R = distance from object to sensor; NL = noise level;

DT = detection threshold.

The overall ambient noise level (NL) was estimated at a level of 115 dB//lμPa in the water both inside and outside of the cage. The overall feeding sound level (SL) of the Sea Bream was detected in the range of the level of noise until 125 dB//l μPa. Usually for reliable identification of targets the detection threshold is set for (DT) >

NL + 3dB. According to equation 1 and TL definition the fish feeding sound could be detected when the fish were not more than 3-4 m from the sensor. This result was confirmed experimentally. These results showed that omnidirectional sensors do not receive feeding sound from neighbouring cages. Results from the experiment showed that the maximal level of intensity of the fish feeding sound was received through the water when the acoustic sensor model was in the centre of the cage at a depth of about 1 m. (b) Measurement of amplitude-time- frequency parameters of underwater sound Figures 11-13 show some typical results of analyses of signals recorded on the CD in the amplitude-time-frequency domain. They display spectograms of continuous time signals, together with their waveforms. Raw signals during fish feeding with noise from boats passing near the cage is shown in figure 11. In figure 12 typical signals during fish feeding without additional noise can be seen. Figure 13 shows ambient noise without fish feeding.

Figure 14 is a diagram of the dispersion of amplitude-frequency parameters of the maximum amplitude signal that was received by the hydrophone installed in the centre of the sea cage. After analyses of this graph, it can be seen that the signals from fish during feeding are in a separate region from signals of boats passing near the cage and ambient noise without feeding. The graph shows the obtained data in the form of parameters for signal filtering in the time-frequency domain and for the algorithm of signal processing.

(c) Additional interference During the experiments there sometimes appeared on the input of the DAQ, radio interference with a large amplitude. Radio interference depends upon the length and location of the cable (in air or in water). Grounding partially solved this problem.

Example 3: Testing the system of the present invention in real sea conditions The objective of the experiment was to test the algorithm of fish feeding sound detection, filtration and processing. Additionally, the principle of fish appetite measurement and properties of the system of the present invention were tested in real sea conditions.

The experiment was performed using the experimental arrangement of Figure 10. Series experiments were performed to verify the algorithm of the fish feeding sound detection, filtration and processing in different cages with different sizes and amount of fish. Experiments were performed in cage numbers 36, 34 and 32 with diameter 10m and depth 15m. In cage #36 there were 82,000 Sea Bream (Denis) with average weight of 448g. In cage #34 there were 94,000 Sea Bream (Denis) with average weight of 120g. In cage #32 there were 57,000 Sea Bream (Denis) with average weight of 466g. Two water acoustic sensors model 8226s in series were located in the different cages, at the centre of the cage, at a fixed depth of about lm.

The measurement system was started simultaneously with the start of the feeding in the cage. The operation was performed manually. Fish feeding sound and ambient noise were received by water acoustic sensor model 8226s coupled through connection box to plug and play DSP board that was installed to workstation (WS) Pentium II 300 MHz. The fish feeding sounds were filtered from ambient noise, using a special algorithm of signal processing and compression of the data. The fish appetite with elapsed time can be represented on the graph in real time during feeding. The stopping of the measurement system was simultaneous with the stopping of the feeding in the cage and was performed manually. The recorded data, considered processed, went through a second processing step for separation of main trends in non-real time. The software for recording and analysing data was written by object oriented language Lab View 5. Results

Relative levels of current fish feeding sound intensity inside of sea cages were obtained at a time of fish feeding versus elapsed time (figure 15). Trend lines were calculated not in real-time, but on the basis of field data. A good linear relationship was shown between fish feeding sound intensity (in dB) and elapsed time (amount of food) (Figures 15, 16 and 17). Their expression with high correlation coefficient R was in accordance with the following equation: 10Lg(I) = C - Gt where C = 101g(KA₀) G = 10k lg(e)

K = calibration coefficient of the measurement system A₀ = fish appetite at the initial moment of feeding k = coefficient of proportionality with dimension (1/sec)

This confirms the suggested algorithm of measurement of fish appetite by fish feeding sound with a high degree of validity, which can be used for estimation and prediction of fish satiation in real time.

Trend lines were reduced to a straight-line depiction via the linear law that was taken from the oscillations of sound levels that exist during the fish feeding. This result is measured by the real fish-feeding situation. All fish not ingesting all the food at the same time may cause oscillations. One part of the school captures and ingests food while the other section waits. This process is not continuous during the feeding. It is flowing and wavy. This phenomenon can also be seen visually during fish feeding. It will be appreciated that the above examples and descriptions are intended only to serve as examples, and that many other embodiments are possible within the spirit and the scope of the present invention.

APPENDIX

Baras, E. & Lagardere, J., 1995, Fish Telemetry in Aquacultural : Review and Perspectives. Aquacultural International 3,77-102.

Beitinger, T. L., 1990. Behavioral Reactions for the Assessment of stress in Fishes. J Great Lakes Res.. Internal. Assoc, Great Lakes Res., 16(4), pp. 495-528.

Brett, J.R., 1979. Environmental factors and growth. In: Hoar, W.S., Randall, D.Y. and Brett, J.R. Fish Physiology. Vol. VIII, Bioenergetic and Growth. Academic Press, New York. NY. pp. 279-352.

Brantley, R.K. and Bass, A.H. 1994. Ethology. 96:213-232.

Cowey, C. B., 1979. Protein and Amino Acid Requirements of Finfish. In Finfish Nutrition and Fishfed Technology, Ed. By J. E, Halver & K. Tiews. Schi, Bundesforscungsanst. Fisch. Hmb., 14115. Vol. 1 P,2-16.

Crawford, J.D., Cook, A.P. and Heberlein, A.S. 1997. J. Acoustical Soc. Am. 101 :1200- 1212.

Fillips, M.J. 1989. J. Fish Biol. Vol.35 No. 4, 589-592.

Heath, A.G. 1987. Water Pollution and Fish Physiology. Boca Raton, Florida: CRC Press.

Juell J., Furevik, D.M. & Bjordal, A. 1992. Demand Feeding in Salmon Farming by Hydroacoustic Food Detection. Aquacultural Engineering, 12, pp, 155-167. Keenleysid, M.H.A. 1979. Diversity and Adaptation in Fish Behaviour. New York: Springer- Verlag.

Ladich, F. 1990. Ethololgy 84, 192-201.

Lobel, P.S. 1992. Environmental Biology of Fishes. 33: 351-358.

Maniva, Y. 1976. Attraction of bony Fish, Squid and Crab by Sound. A. Schuijf & A D. Hawidns (Eds,). Sounds Reception in Fish, Elsevier, Amsterdam.

Myrberg, A. A. Jr, 1993. Fish Communication, pp. 1260-1265. In: Encyclopedia of Language and Linguistics. Vol. 3. (R.E. Asher, ed.) Pergamon Press, New York.

Nyman, H. G. 1981, Sublethal Effects oflead (Pb) on Size Selective Predation by fish- Applications on the Ecosystem Level, Verh. Interat. Vtein. Linmo). 21. pp. 1126-1130.

Ponxin, P. & Ruwet, J. C. 1994. Application to Freshwwer Aquaculture of the Methods Used to Measure the Behavior of Fish: A Brief Review. In: Measures for Success, Proceedingng of The international Conference Bourdaux Aquaculture 94 (Ed's P. Kestemont, J. Muir, F. Sevilla And P. Williot) Cemegraf Editions: Bordeaux, France, pp. 271-275.

Rice A. J., 1990. Bioenergeric Modeling Approaches to Evaluation of Stress in Fishes. American Fisheries Society Symposium 8, pp. 80-92.

Stasko A. B, & Horrall R, M. 1976. Method of Counting Tailbeats of free Swimming Fish by Ultrasonic Telemetry Techniques. J. Fish Res, Board Can.. 33, pp. 2956-2598.

Tavolga W. N. 1980. Hearing and Sound Production in Fishes in Relation of fisheries Management. Fish Behavior and it's Use in Capture and Culture of fishes. ICLARM CONFERENCE PROCEEDING 5,512 p. International Center for Living Aquatic Resources Mana Gement, Manila, Philippines, pp, 102-123.

Tavolga W. N., Popper, A.N. and Fay R.R. 1981. eds. Hearing and Sound Communication in Fishes. Springer- Verlag, New York.

Claims

What is claimed is:

1. A method of feeding fish, the method comprising the steps of:

(a) providing a receiver;

(b) placing said receiver in a fish farm; (c) receiving a fish sound from the fish with said receiver; and

(d) determining an amount of food to be placed in said fish farm according to said fish sound.

2. The method of claim 1 further comprising the step of converting said fish sound to an electrical signal.

3. The method of claim 2, further comprising the steps of:

(a) interfacing said electrical signal to a data processor;

(b) processing said electrical signal to obtain data; and (c) controlling amount of food given by a feeder in said fish farm.

4. The method of claim 1 further comprising the step of calculating the dynamics of fish activity during feeding through analysis of data and generating a prediction of feeding time.

5. The method of claim 3, where said step of processing said electrical signal to obtain data, includes the step of statistical analysis treatment.

6. The method of claim 3, wherein said step of processing said electrical signal to obtain data further comprises the steps of:

(a) filtering said electrical signal to produce a filtered signal; and

(b) compressing said filtered signal to produce a compressed signal, wherein a matrix of said compressed signals is proportional to the input signal intensity and is a function of accumulated amount of food for each fish and elapsed time of one feed portion.

7. The method of claim 1, wherein said receiver is an acoustic sensor.

8. The method of claim 7, wherein said acoustic sensor is a hydrophone.

9. The method of claim 1, wherein said receiver is directional.

10. The method of claim 1, wherein said receiver is omnidirectional.

11. The method of claim 1, wherein said receiver includes an array of receivers.

12. The method of claim 1, wherein said fish farm is selected from the group consisting of ponds, in shore cages, off shore cages and aquariums.

13. The method of claim 1 , wherein said fish is selected from the group consisting of aquatic and semi-aquatic animal life.

14. The method of claim 13, wherein said fish is selected from the group consisting of carp, tilipia, salmon, sea bream, trout, miltfish, tuna, eels, crabs, catfish and shrimps.

15. The method of claim 1, wherein said fish sound is a function of the physiological state of the fish.

16. The method of claim 3, wherein said step of controlling amount of food given by a feeder in said fish farm is performed by interfacing said data with the software of the feeder.

17. The method of claim 3, wherein the power source of said feeder is regulated according to said data.

18. The method of claim 3, wherein said step of controlling amount of food given by a feeder is performed manually.

19. The method of claim 3, wherein said feeder includes an automatic feeder in automatic mode.

20. The method of claim 3, wherein said feeder includes an automatic feeder in manual mode.

21. The method of claim 3 , wherein said feeder includes a manual feeder in manual mode.

22. The method of claim 3, wherein said feeder is in a mobile form.

23. The method of claim 3, wherein said feeder is in a fixed form.

24. The method of claim 15, wherein said physiological state of the fish is selected from group consisting of stress and fish disease.

25. A feeding apparatus responsive to fish sound from fish comprising:

(a) at least one receiver to receive and transform the fish sound to an electrical signal;

(b) a data acquisition system to collect said electrical signal from said receiver; (c) a data processor for calculation of feeding time, according to said electrical signal from said data acquisition system; and (d) a feeder responsive to said calculation of said data processor.

26. The feeding apparatus of claim 25, further comprising a preamplifier to amplify said signal.

27. The feeding apparatus of claim 25, further comprising a local system controller.

28. The feeding apparatus of claim 27, further comprising a base station connected to said local system controller.

29. The feeding apparatus of claim 25, wherein said fish is selected from the group consisting of aquatic and semi-aquatic animal life.

30. The feeding apparatus of claim 29, wherein said fish is selected from the group consisting of carp, tilapia, salmon, sea bream, trout, miltfish, tuna, eels, crabs, catfish and shrimps.

31. The feeding apparatus of claim 25, wherein said at least one receiver is a calibration acoustic sensor.

32. A method of calculating fish biomass in a fish farm, the method comprising the steps of:

(a) providing a receiver; (b) placing said receiver in the fish farm;

(c) receiving a fish sound from the fish with said receiver; and

(d) determining the fish biomass according to said fish sound.

33. A method of monitoring the presence of non-fish bodies in a body of water, the method comprising the steps of:

(a) providing a receiver;

(b) placing said receiver in a body of water;

(c) receiving a fish sound from the fish with said receiver;

(d) receiving a non-fish body sound from a non-fish body with said receiver; and (e) establishing the presence of said non-fish body according to said non- fish body sound.

34. The method of claim 33, wherein said non-fish body includes a thief.

35. The method of claim 33, wherein said non-fish body includes a predator.

36. A security apparatus responsive to sound from fish and sound from non-fish bodies comprising: (a) at least one receiver to receive and transform said sound from fish and said sound from non-fish bodies to an electrical signal;

(b a data acquisition system to collect said electrical signal from said receiver; and

(c) a data processor for establishing the presence of said non-fish body according to said sound from non-fish body.

37. The security apparatus of claim 36, wherein said receiver is a calibration acoustic sensor.

38. A method of determining the physiological state of fish, the method comprising the steps of:

(a) providing a receiver;

(b) placing said receiver in a fish farm;

(c) receiving a fish sound from the fish with said receiver; and (d) determining said physiological state of the fish in said fish farm according to said fish sound.

39. The method of claim 38, wherein said physiological state is selected from the group consisting of stress and fish disease.

40. The method of claim 38, wherein said physiological state of the fish is a function of said fish sound.