NL2015576B1 - Method and system for determining the condition of an animal. - Google Patents

Abstract

The invention relates to a method and system for determining the condition of an animal. Movements of the animal are measured over a certain period of time. The movements are converted into a movement signal that represents the measured movements. A frequency spectrum of the movement signal is determined. The frequency spectrum is subdivided into a plurality of frequency subareas. For a multiple of the frequency subareas, the amount of energy in the relevant frequency subarea is determined per frequency subarea. For each of a plurality of frequency subareas, the determined amount of energy in the relevant frequency subarea is compared to each of a plurality of expectation values for the amount of energy in the frequency subarea, each expectation value corresponding to one state of a plurality of states of the animal, for the determining the current condition of the animal.

Description

Title: Method and system for determining the condition of an animal

Background

Monitoring the health status of animals is important in every sector in which animals are kept for professional purposes. For example, within dairy farming it is important to know when cows are pregnant, in order, for example, to be able to dry the cow in time or to adjust the care of the cow. In addition, early detection of sick animals is also important in order to be able to treat them in time and, if necessary, to be able to isolate the animals from the other animals. Animal health monitoring is therefore not limited to dairy farming, but is also important in other sectors of animal husbandry.

Various state monitoring systems for animals are known, including also monitoring systems which use movement sensors for determining the current condition of the animal. However, a problem is that the accuracy for determining the current condition (for example: standing, lying down, eating, ruminating, walking) often leaves something to be desired. This is partly due to the fact that in many cases the recorded movements could fit in with various such situations.

International patent application WO2013 / 006056 describes for this purpose a system which evaluates the data of a motion sensor in combination with position determination with the aid of various filters for determining a current condition of the animal.

Summary of the invention

The present invention has for its object to provide a state determination method and associated system for animals, with which accurate state determination is possible on the basis of a motion sensor.

To this end, according to a first aspect thereof, the invention provides a method for determining the condition of an animal, the method comprising the steps of: a. Measuring movements of the animal during a certain period of time; b. converting the movements measured in step a. into a motion signal representing the measured movements; c. determining a frequency spectrum of the motion signal; d. subdividing the frequency spectrum into a plurality of frequency subareas; e. determining, for a multiple of the frequency subareas, per frequency subarea, the amount of energy in the relevant frequency subarea; f. determining the current condition of the animal in predetermined manner from the information obtained in step e.

The method according to the present invention is based on recognition of the most likely instantaneous state on the basis of the energies in the respective frequency subareas in the frequency spectrum of the measurement. As is known, specific behaviors of an animal can be recognized by the movements that the animal makes. However, those specific behaviors can be distinguished much more accurately because they are characterized by specific common frequencies and intensities thereof in the motion signal. By determining the frequency spectrum of the motion signal (i.e. the energy distribution in the frequency domain for the entire measurement signal over the measured time period) and dividing the frequency domain into frequency subareas, the amount of energy measured therein can be determined for each frequency subarea. According to the invention, the current condition of the animal can then be determined.

In particular, it applies that step f. comprises: comparing, for each of a plurality of frequency sub-regions, the ones in step e. determined amount of energy in the relevant frequency sub-region with each of a plurality of expectation values for the amount of energy in the frequency sub-region, each expectation value corresponding to one state of a plurality of states of the animal, for determining the current state of the animal. By comparing the measured amount of energy determined in this way with the expected values for the amounts of energy that could be expected for each of a plurality of possible states in which the animal might be located. This comparison step can be implemented in various ways.

According to an embodiment of the invention, for example, step f. are performed by it, for each in step e. determined amount of energy and for each state of a plurality of states, determining a probability value that the determined amount of energy belongs to the relevant state. This probability value can be determined based on probability distribution data for the expectation values for each state of a plurality of states and for each frequency sub-region. The probability distribution data indicate the probability that a measured amount of energy in a frequency sub-area fits in with the relevant state.

Thus, for example, if the measured amount of energy in a frequency subarea is equal to A, then the probability value PioPen, Afi (A) indicates the probability that in the frequency subarea Afi a measured energy value A matches an energy value that would be expected to be measured if the animal would be in the "walking" state. If that probability value is very small (for example Pi0pen, Afi (A) = 0.01) then it is not very likely that if the animal were walking, an energy value with size A would be found in this frequency sub-area. If, on the other hand, the probability is relatively large (for example, Pipen, Afi (A) = 0.1), the measured energy value in this frequency sub-area could fit well with the 'walking' state. When this is done in accordance with the invention for all possible states in which the animal may be, it follows from the most likely state in which the animal is at that moment.

The probability distribution data of the states with which the measured amount of energy is to be compared can, for example, be stored in a data memory. For each state, the probability distribution data, for all frequency subareas per frequency subarea, comprise a probability distribution profile that indicates the development of the probability value depending on the amount of energy. Comparison of the determined amount of energy in step e. against this probability distribution profile for the frequency sub-region in question, therefore, provides the aforementioned probability value for the relevant state.

One in step e. a certain amount of energy in a single frequency sub-area already provides insight, but is often still insufficient to be able to determine the current condition of the animal with great accuracy. According to the invention, step f. therefore preferably performed for a plurality of frequency sub-regions, and in particular - according to some embodiments - preferably for all frequency sub-regions of the frequency spectrum. For each state, optionally, for example, an overall probability value - or a parameter that is proportional to the overall probability value - can be calculated by compiling the calculated probability values for different frequency subareas for the relevant state.

According to a further embodiment, the method according to the invention comprises a step for this purpose: g. calculating a total probability value for the respective state for each of the plurality of states, wherein the total probability value for the relevant state is calculated by the probability values of the frequency subareas for that state as determined in step f. multiply with each other.

Suppose for a number of frequency subareas Afi ..... Mi the respectively in step e. certain energy values are equal to Ai ..... Ai, then the corresponding probability values for the "sleeping" state are, for example:

Pslaps, Afl (A) = pi, Pslapen, Af2 (A) = P2, ......, Pslapen, Afi (A) ~ pi.

The total probability value Ptotaai, sleeping can then be determined in accordance with the present embodiments as follows: ^ total sleepmi ~ ^ slapsn, & fn (- ^ m) A An = 1 ..i

By comparing the total probability values for the various possible states of the animal with each other, the current state of the animal can be accurately determined on the basis of this information. A suitable parameter that is proportional to the overall probability value for accurately determining the condition can also be determined in a different manner.

If desired, but not essential, the total probability values found can be normalized because the probability that the animal is in one of these states (apart from which state that is exactly) must always be equal to 1, provided that the multitude of possible states in the comparison is complete (ie if there are no longer any states in which the animal could be, but which have not been included in the measurement). However, if not all states are included in the comparison, a closing post can be used in accordance with some embodiments.

To that end, according to some embodiments, the states of the plurality of states relate to: standing ruminating, lying ruminating, standing resting, lying resting, eating, sleeping, walking, and closing post that concerns all other possible states of the animal other than standing ruminating, lying ruminated , standing, resting, eating, sleeping and walking. The states that are included in the measurement can include one or more or all of the above states, with or without the stated closing point. The invention is not limited to the use of all of the above states, but can also be applied based on some of them. Moreover, it is possible that additional states are used in the method which is not specifically mentioned herein.

Furthermore, according to some embodiments, the method further comprises a step h. wherein at least one group of states is formed wherein the group comprises a subset of the plurality of predetermined states and wherein the chance is determined for the group that the animal is in one of the states within the group, e.g. step g. summing up certain odds for all possible states of the relevant group or by selecting the state with the highest probability within the group.

For example, standing ruminating and standing resting can be categorized under the group of 'standing' states. However, portrait ruminations and portrait ruminations can also be categorized under the group of 'ruminate' states. Other compositions of groups of states are also possible. According to a specific embodiment, step h. formed a plurality of different groups. For example, the "standing" and "ruminating" groups can also be used simultaneously, in which case an animal could even be in both state groups at the same time. As already explained above, according to an embodiment, the states in step h. of each group thus include a common activity of the animal such as standing or lying. A special grouping to be used that in step h. can be formed, according to yet a further embodiment, a group that relates to standing ruminating, standing resting, eating and the closing post or lying resting, lying ruminated and sleeping. By grouping these combinations in this way we can distinguish between the two states standing or lying down.

The time period in which the movements are measured with the motion sensor can be suitably selected for obtaining an accurate state determination. The measurement of the movements with the aid of the motion sensor can, for example, be carried out during the time period to be measured by taking a series of sequential samples.

According to a further embodiment, at least steps a. - f. and optionally the step g. and / or the step h. performed for different time periods. For example, the measurement can be repeated periodically (for example, every 30 seconds). The time periods may also be sequentially concatenated, thus continuously measuring, but with the results per time period being processed to frequency spectrum and further analyzed as described.

Determining the frequency spectrum in step c. can, in accordance with an embodiment, be implemented by performing a Fourier transformation (FT) of the motion signal. In addition, the frequency subregions of some embodiments in step d. be chosen to correspond to a plurality of bins of the frequency spectrum obtained via FT.

The information obtained with the method about states of the animal for the different time periods can be collected for further processing. For example, the total time span within which the animal is in a specific state can be determined by measurement, and sent to a central server where this data is stored and / or further analyzed.

According to a second aspect, the invention provides a system for determining the condition of an animal, provided with a motion sensor adapted to be attached to an animal for measuring movements of the animal, and a signal processing unit for processing the animal. information received, wherein the system is arranged to form a motion signal representing the measured movements of the animal during a certain period of time and wherein the signal processing unit is arranged to perform the following steps: c. determining a frequency spectrum of the motion signal; d. subdividing the frequency spectrum into a plurality of frequency subareas; e. determining, for a multiple of the frequency subareas, per frequency subarea, the amount of energy in the relevant frequency subarea; f. determining the current condition of the animal in predetermined manner from the information obtained in step e.

In particular, step f. : comparing, for one or more frequency subareas, the steps described in step e. determined amount of energy with an expected value associated with at least one state for determining the state of the animal.

The motion sensor may, for example, comprise an accelerometer or G-sensor mounted on a label carried by the animal. Thus, according to some embodiments, the label to be worn by the animal is formed by a collar which can be hung, for example, around the neck of a cow. The signal processing unit can also be on the label. In fact, both the measurement and the condition determination as a whole can take place on the label. This data could then be sent later or simultaneously or transferred to a central processing system. Depending on the specific implementation, data communication means may therefore be present on the label, such as, for example, a communication unit for wireless communication via WIFI or mobile data communication (GPRS, UMTS, LTE, etc.). The person skilled in the art is familiar with various possibilities for implementing data communication that are suitable for the present application.

Brief description of the figures

The invention will be discussed below on the basis of specific embodiments thereof not intended as limiting, with reference to the accompanying figures, in which:

Figure 1 is an illustration of a system according to the present invention;

Figure 2 shows a label for use in a system according to the present invention;

Figure 3 is a schematic representation of sampling in a method according to the invention;

Figure 4 shows a probability distribution graph of the expected values for energies in a frequency sub-area per state;

Figure 5 shows a state observation obtained with a method according to the present invention;

Figure 6 is a schematic representation of a method according to the present invention.

Detailed description

Figure 1 shows schematically a system 1 according to the present invention. In Figure 1, a cow 3 wears a collar 5 on which there is, inter alia, an accelerometer or g-sensor. The shown collar 5 hangs over the neck of the cow 3, however, the skilled person will understand that a lowered collar 5 can also be used. Figure 2 shows a schematic representation of the collar 5 that is worn by the cow 3 in Figure 1. The collar 5 comprises an accelerometer 22 with which movements of the cow can be measured. The motion sensor 22 transmits these measured values to a signal processing unit 23. With this data, the current condition of the cow 3 can be determined in the signal processing unit 23 (or another processing unit communicatively connected thereto). This condition indicates the activity of the animal, and thus may, for example, include standing ruminating, lying ruminating, standing resting, lying resting, eating, sleeping, walking, and closing post concerning all other possible states of the animal other than standing ruminating, lying ruminating, standing resting, lying resting, eating, sleeping and walking.

In the embodiment shown, the method according to the invention is mainly carried out in the central signal processing unit 23, although the invention may also be implemented differently, whereby signal processing could, for example, take place centrally. The collar 25 may further comprise a memory 25 or other data storage medium in which the probability distribution data per frequency sub-area per state is stored. Alternatively, the signal processing unit 23 can retrieve this data from another storage medium, obtained if necessary by means of wireless data communication, for example from a server or from a telecommunications network. The central processing unit 23 is further connected to a data communication unit comprising an antenna 26 for transmitting measurement data, such as the determined states. Depending on the implemented embodiment, of course, other data can also be sent, such as the measured motion signal, the frequency spectrum, or the energies in the frequency subareas. Moreover, the data communication unit and antenna 26 may, depending on the implementation, possibly receive data from external sources.

In Figure 1, the forwarding of this data from the collar 5 to a central system 8 is schematically represented with wireless data signal 15. The central server 8 comprises a central processing unit 9 connected to a memory 10 on which the data can be stored and subsequently processed. Furthermore, the central processing unit 9 is connected to data communication means 12 comprising an antenna 13 for receiving the wireless signal 15 from the collar 5. The central processing unit 8 is adapted to receive wireless data signals from a plurality of collars such as collar 5 for state monitoring of a multitude of animals at the same time.

Data can be collected with the help of the collar 5 during a certain period of time. If possible, the data is collected periodically during a certain period of time. This is schematically shown in figure 3. In figure 3 it is shown that during a period of time, for example 5 seconds, indicated by arrow 33, the movement sensor of the collar 5 takes a plurality of samples in order to record the movement. Every 30 seconds, the collar 5 performs such a measurement as shown by the subsequent measurement series 34. Between the measurements 33 and 34 there is a specific period of time during which no measurements are made. In this intermediate period between the measurements 33 and 34, the collar 5 can be passive. However, other measurements can also be made with collar 5 at that time, as indicated in Figure 3 by taking samples such as sample 30 periodically. Taking such intermediate samples is entirely optional and is not part of the present invention.

In the alternative case (not shown here), the collar 5 can perform a continuous measurement, the motion sensor continuously taking samples of the movements made by the cow. The samples obtained can be divided into time periods of, for example, 5 seconds, and analyzed per time period in accordance with the present invention.

When the collar has made a measurement, such as measurement 33, the obtained motion data is converted into a frequency spectrum. This can for example be done with the help of Fast Fourier transformation (FFT). With Fast Fourier transformation, the measured energy is divided into frequency sub-areas (bins) and stored in the memory of the collar 5. In an alternative embodiment, the frequency sub-regions can also occupy more than one bin. The current condition of the animal is obtained from this information in a predetermined manner. According to a special embodiment of the invention, the state is determined by comparing the measured energy value per frequency sub-region, i.e. per bin in the present example, with a probability distribution that the measured energy in that particular frequency sub-region is associated with a specific state.

Figure 4 illustrates the probability distribution for eight different states for a specific frequency division. The probability distribution, such as probability distribution curve 38, indicates what the probability (P) is that a certain measured amount of energy (E) fits with a certain condition (in the case of curve 38 this is chewing lying down). For this purpose, the measured energy E is plotted on the horizontal axis, the vertical axis shows the probability P. If, for example, the measured energy is 2.0, the corresponding probability that the cow at the time of measurement ruminates in the condition lying down 'is based on the energy found in this frequency division equal to 0.045. The adjacent curve 39 indicates the probability distribution for the "standing ruminating" state.

The chance that on the basis of the measurement in the relevant frequency sub-area the cow is in the state of ruminating is 0.09, and this state is therefore more likely on the basis of the comparison of this one frequency sub-area.

According to a preferred embodiment of the present invention, for each of the frequency sub-areas, the measured energy E is compared with the probability distributions for each of the states. For each state, the obtained probability values P for all frequency subareas are multiplied with each other to obtain a quantity that indicates the total probability value for that condition. The state with the largest total probability value is the instantaneous state in which the cow 3 is probably at that moment. This data can be sent via a wireless data signal 15 to the central processing unit 8, where they are stored in memory 10 for further processing.

Figure 5 shows schematically the measured probability values for a cow for a certain period of time. Graph 40 shows the probability value for each of the eight states. It can be seen from graph 40 that the cow slept at the start of the measurement, as indicated by chance curve 42. Because there is some correspondence between movements in the "sleeping" state and movements in the "lying resting" state, it can be seen that in graph 40 the "lying resting" state, indicated by curve 46, still provides a relatively large probability value. It appears convincingly from the measured measurement signal that the cow was in the "sleeping" state (curve 42). The subsequent states of "eating" 43, "ruminating" 45, and again "sleeping" 44 indicate the cow's activities during the day.

Figure 6 is a schematic representation of the method according to the present invention. The method starts with step 100 (step a) in which the movements of the animal are measured with the aid of the motion sensor during a certain period of time (for example time period 33). In step 102 (step b), the movements measured in step 100 are converted into a motion signal representing the measured movements. Next, in step 104 (step c), the frequency spectrum of the motion signal as determined in step 102 is determined. In step 106 (step d), this frequency spectrum is divided into frequency sub-regions. Optionally, a frequency sub-region may coincide with a single bin, but a frequency sub-region may also contain multiple bins of the frequency spectrum, as those skilled in the art will appreciate.

In step 108 (step e), for a plurality of the frequency subregions, per frequency subregion, the amount of energy in that particular subregion is determined. This is the amount of energy that will subsequently be compared in step 110 (step f) with the expected value, in particular the probability distribution of the expected value of the energy, for the various states. In step 110 (step f) this comparison takes place by comparing for each of the states and for each frequency sub-region the measured amount of energy with the probability distribution curves (as schematically shown in Fig. 4) to calculate a probability value therefrom for each condition.

The obtained probability values can be multiplied for each state in step 112 (step g). Thus in step 112 a multiplication takes place of all calculated probabilities of all frequency sub-regions per state. This gives the overall probability or the total probability value per state, and this for all possible states. Subsequently, in step 114 (step h), certain states can optionally be brought together to form a group. For example, the chances for "standing ruminating" and "standing resting" can be grouped into a "standing" group. The state information thus ultimately obtained is sent via the wireless data communication signal 15 to the central server 8 for further processing. This further processing may involve summing the probability values of states within the group or selecting a highest probability value of a state within the group. Other possible groups of states are for example a group comprising standing ruminating, standing resting, eating, and a closing post. Yet another group can include the following states: lying down, lying down, and sleeping.

The data can also be processed by server 8, for example, to compile a graph as shown in Figure 5. The activities of several cows can also be compared with each other, or with a data history for the cow in question, or with reference values for different cows. states of condition (for example: no or very much appetite, above expectations much or just little sleep and / or rest activity, cow is very active or restless).

The specific embodiments of the invention described above are intended to illustrate the invention principle. The invention is only limited by the following claims.

Claims (26)

A method for determining the condition of an animal, the method comprising the steps of: a. Measuring movements of the animal during a certain period of time; b. converting the movements measured in step a. into a motion signal representing the measured movements; c. determining a frequency spectrum of the motion signal; d. subdividing the frequency spectrum into a plurality of frequency subareas; e. determining, for a multiple of the frequency subareas, per frequency subarea, the amount of energy in the relevant frequency subarea; f. determining the current condition of the animal in predetermined manner from the information obtained in step e. The method of claim 1, wherein step f. includes: comparing, for each of a plurality of frequency subareas, the steps described in step e. determined amount of energy in the relevant frequency sub-region with each of a plurality of expectation values for the amount of energy in the frequency sub-region, each expectation value corresponding to one state of a plurality of states of the animal, for determining the current state of the animal. The method of claim 2, wherein the step f. includes: determining, for each in step e. determined amount of energy and for each state of a plurality of states, of a probability value that the determined amount of energy belongs to the relevant state, the probability value being determined based on probability distribution data for the expected values for each state of a plurality of states and for each frequency sub-region , wherein the probability distribution data indicates the probability that a measured amount of energy in a frequency sub-area matches the relevant state. The method of claim 3, further comprising the step of: g. calculating a total probability value for the respective state for each of the plurality of states, wherein the total probability value for the relevant state is calculated by the probability values of the frequency subareas for that state as determined in step f. multiply with each other. Method according to one of the preceding claims, characterized in that in step c. a Fourier transformation of the motion signal is determined to obtain the frequency spectrum or another type of transformation of the motion signal is determined to obtain the frequency spectrum. The method according to claim 5, characterized in that the plurality of frequency ranges corresponds to a plurality of bins of the frequency spectrum. Method according to one of the preceding claims, characterized in that in a step h. at least one group of states is formed wherein the group comprises a subset of the plurality of predetermined states and wherein the chance is determined for the group that the animal is in one of the states within the group, in particular by the in step g. process certain probabilities for all possible states of the relevant group in combination, in particular by summing all probability values or by selecting a highest probability value of a condition within the group. Method according to claim 7, characterized in that in step h. a plurality of different groups is formed. The method according to claim 8, characterized in that the states in step h. of each group include a common activity of the animal such as standing or lying. A method according to any one of the preceding claims, characterized in that the states of the plurality of states relate to standing ruminating, lying ruminating, standing resting, lying resting, eating, sleeping, walking, and closing post that all other possible states of the animal concerns other than standing ruminating, lying ruminating, standing resting, lying resting, eating, sleeping and walking. 11. Method according to claim 10 and according to at least one of claims 7-9, characterized in that in step h. at least one group is formed that relates to standing ruminating, standing resting, eating and the closing post or lying resting, lying ruminating and sleeping. A method according to any one of the preceding claims, characterized in that at least the steps a. - f. and optionally the step g. of claim 4 and / or the step h. of at least one of claims 7-9 is performed for mutually different time periods. A method according to claim 12, characterized in that the information obtained with the method about states of the animal for the different time periods is collected for further processing. 14. A system for determining the condition of an animal, provided with a motion sensor adapted to be attached to an animal for measuring movements of the animal, and a signal processing unit for processing the received information, the system is arranged to form a motion signal representing the measured movements of the animal during a certain period of time and wherein the signal processing unit is arranged to perform the following steps: c. determining a frequency spectrum of the motion signal; d. subdividing the frequency spectrum into a plurality of frequency subareas; e. determining, for a multiple of the frequency subareas, per frequency subarea, the amount of energy in the relevant frequency subarea; f. determining the current condition of the animal in predetermined manner from the information obtained in step e. The system of claim 14, wherein step f. comprises: comparing, for one or more frequency sub-areas, the ones in step e. determined amount of energy with an expected value associated with at least one state for determining the state of the animal. A system according to claim 14 or 15, wherein for performing step f. the signal processing unit is communicatively connected to a data storage medium for obtaining probability distribution data for the expectation values for each state of a plurality of states and for each frequency sub-region, the probability distribution data indicating the probability that a measured amount of energy in a frequency sub-region matches the relevant state. The system of claim 16, wherein the signal processing unit is further adapted to perform the step: g. calculating a total probability value for the respective state for each of the plurality of states, wherein the total probability value for the relevant state is calculated by the probability values of the frequency subareas for that state as determined in step f. multiply with each other. A system according to any one of the preceding claims 15-17, characterized in that the processing unit is arranged such that, in use, in step c. a Fourier transformation of the motion signal is determined to obtain the frequency spectrum. A system according to claim 18, characterized in that the plurality of frequency ranges corresponds to a plurality of bins of the frequency spectrum. A system according to any one of the preceding claims 15-19, characterized in that the processing unit is arranged such that, in use, at least in a step h. a group of states is formed in which the group comprises a subset of the plurality of predetermined states and wherein for the group it is determined what the probability is that the animal is in one of the states of the animal, in particular by the step g. process certain probabilities for all possible states of the relevant group in combination, in particular by summing all probability values or by selecting a highest probability value of a condition within the group. A system according to claim 20, characterized in that the processing unit is arranged such that, in use, in the step h. a plurality of different groups is formed. A system according to claim 21, characterized in that the processing unit is arranged such that, in use, in the step h. the states of each group include a common activity of the animal such as standing or lying. A system according to any one of the preceding claims 14-22, characterized in that the processing unit is arranged such that, in use, the plurality of predetermined states relates to standing ruminating, lying ruminating, standing resting, lying resting, eating, sleeping , walking and closing post that concerns all other possible states of the animal than standing ruminating, lying ruminating, standing resting, lying resting, eating and sleeping. System according to claim 23 and according to at least one of claims 20 to 22, characterized in that the processing unit is arranged such that, in use, in the step h. at least one group is formed that relates to standing ruminating, standing resting, eating and the closing post or lying resting, lying ruminating and sleeping. 25. System as claimed in any of the foregoing claims 14-24, characterized in that the system is arranged such that, in use, at least the steps c.-f. and optionally the step g. of claim 17 and / or the step h. of at least one of claims 20-22 is performed for mutually different time periods. A system according to claim 25, characterized in that the system is arranged such that, in use, the information obtained about the animal's states for the different time periods is collected for further processing.