EP3487447B1 - Method of detection and indication of the preparatory phase of birthing in an animal, corresponding device, recording medium and computer program product - Google Patents

Description

1. Field of the invention

The field of the invention is that of agriculture, and in particular the monitoring of cattle and horses.

More specifically, the invention relates to a method and a device for detecting and signaling the preparatory phase of parturition of an animal of the bovine or equine family.

The invention also makes it possible to discriminate whether farrowing is likely to take place without difficulty or not.

2. State of the art

It is well known that, without human intervention, some cows or mares lose their newborn during parturition.

It is therefore necessary for a breeder to exercise, day and night, a surveillance of animals that have reached term, in order to be able to intervene if necessary during parturition in an attempt to save the newborn or reduce post-partum complications. partum.

It is also common to carry out several matings simultaneously within the same herd, which complicates the breeder's task of monitoring insofar as calvings or foalings will occur over the same period.

In order to prevent a breeder from having to remain permanently mobilized both day and night, and to enable him to rest during the calving or foaling period, various techniques have been proposed.

There are known calving detection techniques consisting of introducing sensors measuring the temperature, pressure and/or acidity inside the genital tract of a cow in order to identify the imminence of calving. . One of these techniques consists for example of introducing a thermometer into the vaginal canal of the animal which will be expelled with the bag of waters.

A disadvantage of these intrusive techniques is that the introduction and the positioning of the sensors in the genital apparatus of the cow is delicate, and that it is a source of infection and/or irritation for the latter.

Another disadvantage of these intrusive techniques is that normal metabolic changes in the animal due to its activity can cause false alarms to be triggered.

Another disadvantage of these intrusive techniques is that they are expensive to implement.

It has also been proposed to attach a mercury or ball inclinometer to the tail of a cow or a mare, which will trigger an alarm when nearing parturition, when the animal's tail remains raised. .

A disadvantage of this technique is that it poses reliability problems. Indeed, it is found that alerts can be triggered inadvertently, for example when the animal changes position or lies down.

It has also been proposed to strap a cow's waist with a belt that measures the animal's uterine activity and sounds an alarm when the frequency of contractions increases.

This technique has several drawbacks: the belt presents a danger to the animal if it is too tight. However, it should not be too loose at the risk of not working properly, which requires tightening it every 3 to 4 hours. In addition, this technique is expensive and does not make it possible in certain cases to distinguish the deformations due to the work of the animal from other deformations resulting from a normal activity of the animal for eating or moving.

We also know from the document WO-A1-2013/186235 a technique for detecting the imminence of a calving putting an accelerometer fixed on the tail of a cow, in order to estimate the overall activity of the animal from the measurements of the acceleration.

A drawback of this known technique, which consists in comparing the overall activity alone to a threshold to detect whether the animal is about to farrowing, is that it is not precise enough to detect the imminence of farrowing in more than 50 to 60% of the animals.

We also know from the document US2015/230903 A1 a technique for detecting the imminence of a parturition of a bovine or an equine animal using a three-dimensional acceleration sensor placed on the tail of the animal, consisting in comparing a sequence of measurement with a reference sequence representative of a behavioral manifestation of the animal such as muscle contractions, colic resulting from uterine pain, tail lifting or overactivity.

1. Objects of the invention

The object of the invention is in particular to overcome the drawbacks of the state of the art mentioned above.

More precisely, the aim of the invention is to provide a technique for detecting and signaling the preparatory phase of giving birth to an animal which is reliable.

The object of the invention in particular, in at least one particular embodiment of the invention, is to propose such a detection and signaling technique which has a rate of detection of the preparatory phase of farrowing at least equal to 80 % and advantageously greater than 85%.

Another object of the invention is to provide such a detection and signaling technique which is simple to implement and has a reduced cost price.

The invention also aims to provide a technique for detecting and signaling the preparatory phase of parturition of an animal which makes it possible to detect whether the parturition takes place without difficulty or not.

An object of the invention is also to propose such a detection and signaling technique which is reusable.

Another object of the invention is to provide a technique for detecting and signaling the preparatory phase of parturition of a bovine or an equine which is reusable on several animals, which requires little maintenance and which is self-sufficient in energy over a period of several years.

2. Disclosure of Invention

These objectives, as well as others which will appear subsequently, are achieved with the aid of a method of detecting and signaling the preparatory phase of giving birth to an animal of the family of bovines or of the family of horses according to claim 1.

In the context of the invention, the expression “preparatory phase of farrowing” is understood to mean in its general sense. This is the phase preceding the engagement of the fetus in the pelvis, generally lasting between 2 and 24 hours, during which the first uterine contractions and the dilation of the cervix occur. .

• mesure des composantes de l'accélération d'une portion de la queue dudit animal selon trois axes orthogonaux ;
• traitement desdites mesures destiné à détecter une occurrence de ladite phase préparatoire comprenant :
  • une étape de projection desdites mesures des composantes dans un repère orthonormé direct X, Y, Z où l'axe Z est vertical et dirigé vers le bas et l'axe X est horizontal et dans le prolongement du sacrum dudit animal, dans le cas où le repère formé par lesdits trois axes orthogonaux ne coïncident pas avec ledit repère orthonormé X, Y, Z ;
  • au moins deux étapes parmi les étapes a), b), c) d) et e) suivantes :
    1. a) détermination, pour chaque instant de mesure, d'une donnée Q représentative d'une position levée ou abaissée de ladite queue, comprenant une étape de calcul d'une valeur de l'angle de la queue par rapport à la verticale à partir des projections des mesures des composantes de l'accélération suivant les axes X et Z, ou des mesures des composantes de l'accélération, lorsque le repère formé par lesdits trois axes orthogonaux coïncide avec ledit repère orthonormé X, Y, Z, et une étape de comparaison dudit angle calculé avec une valeur d'angle seuil prédéterminée ;
    2. b) détermination d'une donnée A représentative de l'activité dudit animal, pour chaque instant de mesure, comprenant une étape de calcul des variations de chacune des composantes de l'accélération par rapport à sa valeur moyenne sur une période prédéterminée à partir des projections des mesures des composantes de l'accélération suivant les axes X, Y et Z ou des mesures des composantes de l'accélération, lorsque le repère formé par lesdits trois axes orthogonaux coïncide avec ledit repère orthonormé X, Y, Z ;
    3. c) détermination d'une donnée C représentative de l'occurrence d'une crise de colique chez ledit animal comprenant une étape d'application d'un filtre passe bande auxdites mesures des composantes de l'accélération sur au moins deux axes, et de préférence des composantes de l'accélération sur les axes X et Z ;
    4. d) détermination d'une donnée T représentative de l'occurrence de contractions chez ledit animal comprenant une étape d'estimation du coefficient de corrélation, pour chaque composante de l'accélération, entre une séquence des mesures de cette composante de l'accélération et une séquence de données de référence ;
    5. e) détermination d'une donnée TR représentative de l'occurrence d'un piétinement chez ledit animal comprenant une étape d'identification d'une fréquence prédominante dans le spectre de fréquence pour une séquence des mesures de la composante de l'accélération suivant l'axe Y afin d'établir si ladite fréquence est dans une plage de fréquences prédéterminée ;
  • classification automatique binaire de l'état dudit animal entre un état de préparation de la mise-bas et un état non-préparatoire, à partir desdites données Q, A, C, T et/ou TR déterminées précédemment, représentatives respectivement d'une position levée ou abaissée de la queue, de l'activité, de l'occurrence d'une crise de colique, de l'occurrence de contractions et de l'occurrence d'un piétinement ;
• génération d'un premier signal d'alerte lorsque le résultat de ladite classification automatique binaire est un état de préparation de la mise-bas.

According to the invention, such a method for detecting and signaling the preparatory parturition phase comprises the following steps in this order:

• measurement of the components of the acceleration of a portion of the tail of said animal along three orthogonal axes;
• processing of said measurements intended to detect an occurrence of said preparatory phase comprising:
  • a step of projecting said measurements of the components into a direct X, Y, Z orthonormal frame where the Z axis is vertical and directed downwards and the X axis is horizontal and in the extension of the sacrum of said animal, in the case where the frame formed by said three orthogonal axes does not coincide with said orthonormal frame X, Y, Z;
  • at least two steps from among the following steps a), b), c) d) and e):
    1. a) determination, for each instant of measurement, of a datum Q representative of a raised or lowered position of said tail, comprising a step of calculating a value of the angle of the tail relative to the vertical from projections of the measurements of the components of the acceleration along the axes X and Z, or of the measurements of the components of the acceleration, when the frame formed by said three orthogonal axes coincides with said orthonormal frame X, Y, Z, and a step comparing said calculated angle with a predetermined threshold angle value;
    2. b) determination of a datum A representative of the activity of said animal, for each instant of measurement, comprising a step of calculating the variations of each of the components of the acceleration with respect to its average value over a period predetermined from the projections of the measurements of the components of the acceleration along the axes X, Y and Z or of the measurements of the components of the acceleration, when the frame formed by the said three orthogonal axes coincides with the said orthonormal frame X, Y, Z ;
    3. c) determination of a datum C representative of the occurrence of a colic attack in said animal comprising a step of applying a band-pass filter to said measurements of the components of the acceleration on at least two axes, and of preference of the components of the acceleration on the X and Z axes;
    4. d) determination of a datum T representative of the occurrence of contractions in said animal comprising a step of estimating the correlation coefficient, for each component of the acceleration, between a sequence of measurements of this component of the acceleration and a reference data sequence;
    5. e) determination of TR datum representative of the occurrence of trampling in said animal comprising a step of identifying a predominant frequency in the frequency spectrum for a sequence of measurements of the component of the acceleration according to the Y axis to establish whether said frequency is within a predetermined frequency range;
  • binary automatic classification of the state of said animal between a state of preparation for whelping and a non-preparatory state, from said data Q, A, C, T and/or TR determined previously, respectively representative of a position raising or lowering of tail, activity, occurrence of colic attack, occurrence of twitching, and occurrence of trampling;
• generating a first alert signal when the result of said binary automatic classification is a farrowing readiness state.

Thus, the invention proposes to exploit the measurements of the acceleration of the tail of a cow, a mare or a donkey to classify the state of the animal between a state of preparation for low and a non-preparatory state from a reduced number of parameters Q, A, C, T and/or TR.

The inventors have in fact observed that in animals tail lifting, activity, colic attacks, contractions and trampling constitute the most characteristic manifestations that can be observed through the movements of the animal's tail. during the preparatory phase of farrowing and that each of these manifestations triggers a specific and reproducible sequence of movements at least on part of a bovine or equine population. The inventors have also observed that these specific sequences could vary between two breeds of cows or mares.

In nine particular embodiments of the invention, said step of processing said measurements respectively comprises steps a) and b), a) and c), a) and d), a) and e), b) and c) , b) and d), b) and e), c) and d), c) and e).

In a particular embodiment of the invention, said step of processing said measurements comprises at least three steps from among steps a), b), c), d) and e).

In a particular embodiment of the invention, said step of processing said measurements comprises steps a), b) and c) or d).

In a particular embodiment of the invention, said step of processing said measurements comprises at least four steps from among steps a), b), c), d) and e).

In a particular embodiment of the invention, said step of processing said measurements comprises steps a), b), c) and d).

In a particular embodiment of the invention, said step of processing said measurements comprises the five steps a), b), c), d) and e).

This improves the rate of detection of the preparatory phase of parturition in a group of animals.

In a particular embodiment of the invention, the classifier implemented in said binary automatic classification step is a linear classifier.

According to a particular aspect of the invention, during said binary automatic classification step, the output of the classifier implemented in said automatic classification step is compared with a predetermined classification threshold.

Advantageously, said step a) of determining data representative of a raised or lowered position of the tail and/or said step b) of determining data representative of the activity and/or said step d) determining data representative of the occurrence of contractions and/or said step e) of determining data representative of the occurrence of trampling comprises a step of filtering said projections of the measurements or of said measurements.

According to a particular aspect of the invention, said step a) of determining data representative of a raised or lowered position of the tail comprises a step of assigning to said data Q a first value, preferably 1, when said angle value is greater than or equal to said threshold angle value and a second value, preferably 0 or -1, when said angle value is less than said threshold angle value.

According to a preferred aspect of the invention, said threshold angle value is between 32° and 37°, preferably is substantially equal to 35°.

The inventors have in fact observed that a threshold angle value close to 35° allows an effective classification of the state of the animal.

In a particular embodiment of the invention, said step a) of determining a datum Q representative of a raised or lowered position of said tail further comprises a step of reallocating to said value Q, for each instant of measurement t _measurement , of an average, over a predefined time interval Δt, of the values assigned to the datum Q for the measurement instants preceding the measurement instant t _measurement included in the interval [t _measurement -Δt; t _measure ].

In an advantageous embodiment of the invention, said step a) of determining a datum Q representative of a raised or lowered position of said tail comprises a step of weighting said value of the angle of the tail by the value of the measurement of the acceleration component, or its projection, on the Y axis.

According to a particular embodiment of the invention, said step b) of determining the activity of said animal comprises a step of calculating the quadratic mean of the variations of the components of the acceleration along the axes X, Y and Z with respect to their mean value.

Preferably, during said step c) of a datum representative of the occurrence of a colic attack, the cutoff frequencies of said band pass filter are substantially equal to 6 mHz and 12 mHz.

The inventors have indeed observed that the colic attacks are manifested by jolts repeating themselves over time, spaced from 2 min 45 s to 5 min 30 s.

Preferably, in said step d) of determining data representative of the occurrence of contractions, said reference sequence of data is a sequence of data representative of a peak of substantially triangular shape.

According to a particularly advantageous embodiment of the invention, in said step e) of determining the occurrence of trampling, said range of frequencies is limited by the frequencies 150 mHz and 300 mHz.

In a particular embodiment of the invention, a method for detecting and signaling the preparatory phase of parturition of an animal as described above further comprises a step of determining a datum P representative of the standing or lying state of said animal comprising a step of comparing the average over a predetermined period, preferably substantially equal to 1 minute, of the values of the measurements of the components of the acceleration, or of their projection, on the Y axis with respect to a predetermined state threshold value and/or a step of comparing the number of sign changes of the acceleration on the Y axis over said predetermined period with respect to a threshold number.

Indeed, the inventors have found that unlike the case where the animal is lying down, the average acceleration of the component along the Y axis of the acceleration is substantially zero and a minimum number of sign changes of the acceleration along Y over a predetermined period, when the animal is standing.

Preferably, said state threshold value is substantially equal to 0.2 g.

In a particularly advantageous embodiment of the invention, a method for detecting and signaling the preparatory phase of parturition of an animal as described above further comprises a step of binary automatic classification of the state of said animal between a difficult farrowing state and an easy farrowing state based on at least three data, preferably at least four data, even more preferably at least five data, among the data Q, A, C , T, TR and P.

• obtention à l'instant t tel que t=m.ΔT+T₀ avec m un entier, ΔT un intervalle de temps prédéterminé et T₀ est l'instant où ledit premier signal d'alerte est généré, d'un vecteur de cinq variables réelles continues (M¹,M²,M³,M⁴,M⁵) à partir d'une séquence de n valeurs successives desdites données Q, A, C, T et P sur une période prédéterminée précédant l'instant t, où M_i ¹=Q_i, M_i ³=T_i, M_i ⁴=C_i, M_i ⁵=A_i, avec i entier tel que 1≤i≤n, et M_i ² est la durée depuis le dernier changement d'une position couchée à debout ou vice-versa à l'index i estimée à partir de la séquence des valeurs successives de P ;
• détermination d'un vecteur de caractéristiques F_t=(F_t ¹,F_t ²,F_t ³,F_t ⁴,F_t ⁵) où F_t ^k, 1≤k≤5, est le maximum des valeurs moyennes de M^k sur des sous-intervalles de ladite période prédéterminée ;
• attribution à chaque caractéristique F^k, 1≤k≤5, dudit vecteur de caractéristiques d'un vote P_t ^k =log(P(F^k=F_t ^k | H₀)/ P(F^k=F_t ^k | H₁)) où P(F^k=F_t ^k | H₀) est la probabilité d'observer la caractéristique F^k sachant que la mise-bas est facile et P(F^k=F_t ^k |H₁) est la probabilité d'observer la caractéristique F^k sachant que la mise-bas est difficile, obtenues par comparaison avec une distribution statistique de référence construite à partir d'observations sur un échantillon d'animaux de référence ;
• calcul d'un critère de décision f_t s'exprimant sous la forme :
  
  où α_k et O _k sont des constantes prédéfinies ;
• génération d'un deuxième signal d'alerte dans le cas où ledit critère de décision f_t est supérieur ou égal à une valeur seuil prédéterminée S.

In a particularly advantageous embodiment of the invention, said step of binary automatic classification of the state of said animal between a difficult parturition state and an easy parturition state comprising the following steps:

• obtaining at instant t such that t=m.ΔT+T ₀ with m an integer, ΔT a predetermined time interval and T ₀ is the instant when said first alert signal is generated, of a vector of five continuous real variables (M ¹ ,M ² ,M ³ ,M ⁴ ,M ⁵ ) from a sequence of n successive values of said data Q, A, C, T and P over a predetermined period preceding instant t, where M _i ¹ =Q _i , M _i ³ =T _i , M _i ⁴ =C _i , M _i ⁵ =A _i , with i integer such that 1≤i≤n, and M _i ² is the duration since the last change from lying to standing or vice versa at index i estimated from the sequence of successive values of P;
• determination of a vector of characteristics F _t =(F _t ¹ ,F _t ² ,F _t ³ ,F _t ⁴ ,F _t ⁵ ) where F _t ^k , 1≤k≤5, is the maximum of the mean values of M ^k over sub-intervals of said predetermined period;
• allocation to each characteristic F ^k , 1≤k≤5, of said vector of characteristics of a vote P _t ^k =log(P(F ^k =F _t ^k | H ₀ )/ P(F ^k =F _t ^k | H ₁ )) where P(F ^k =F _t ^k | H ₀ ) is the probability of observing the characteristic F ^k knowing that farrowing is easy and P(F ^k =F _t ^k |H ₁ ) is the probability to observe the characteristic F ^k knowing that farrowing is difficult, obtained by comparison with a reference statistical distribution constructed from observations on a sample of reference animals;
• calculation of a decision criterion f _t expressed in the form:
  
  where α _k and O _k are predefined constants;
• generation of a second warning signal in the event that said decision criterion f _t is greater than or equal to a predetermined threshold value S.

Thus, in a novel way, the invention proposes to discriminate the state of farrowing of the animal between a difficult state of farrowing and an easy state of farrowing from the parameters Q, A, C, T and P.

According to an advantageous aspect of the invention, said time interval ΔT is between 30 seconds and 2 minutes, preferably is substantially equal to 1 minute.

According to a particular aspect of the invention, said predetermined period is of a duration substantially equal to 1 hour.

où α_k, 1≤k≤5, est une constante prédéfinie, et $c_t^k$

est la confiance apportée par la caractéristique k, 1≤k≤5, qui s'exprime sous la forme :

avec

la probabilité d'observer la caractéristique F^k sachant que la mise-bas est facile et

la probabilité d'observer la caractéristique F^k sachant que la mise-bas est difficile.In a particular embodiment of the invention, a method for detecting and signaling the preparatory phase of parturition of an animal as described above further comprises a step of calculating a confidence index C _t being expressed in the form:

where α _k , 1≤k≤5, is a predefined constant, and ${\mathrm{vs}}_{\mathrm{you}}^k$

is the confidence brought by the characteristic k, 1≤k≤5, which is expressed in the form:

with

the probability of observing the characteristic F ^k knowing that farrowing is easy and

the probability of observing the characteristic F ^k knowing that farrowing is difficult.

In an advantageous embodiment of the invention, a method for detecting and signaling the preparatory phase of parturition of an animal as described above comprises a step of determining the _end time t corresponding to the end of farrowing, comprising a step of calculating at each instant t the sum St of the variations of the values of data A representative of the activity over a predefined time interval preceding said instant t and a step of detecting the maximum of values of St over a predetermined period, making it possible to determine the _end time t for which said value of S _t is maximum.

According to a particular aspect of the invention, said predefined time interval implemented during said step of determining the instant t _end corresponding to the end of farrowing is of a duration substantially equal to 20 minutes.

Preferably, said predetermined period implemented during said step of determining the _end time t corresponding to the end of farrowing is of a duration at least equal to 30 minutes and preferably at least equal to 50 minutes.

In an advantageous embodiment of the invention, a method for detecting and signaling the preparatory phase of parturition of an animal as described above comprises a step of generating an alert signal intended to signaling the occurrence of an abnormally long farrowing, when no second alert signal is generated 2 hours after the generation of said first alert signal.

In variants of this embodiment of the invention, the alert signal signaling the occurrence of an abnormally long farrowing can generate at any time chosen between 1h and 2h after the generation of said first alert signal, without departing from the scope of the invention.

The invention also relates to a computer program product comprising program code instructions for implementing the step of processing the measurements of the method for detecting and signaling the preparatory phase of parturition of an animal as described above and/or of the step of binary automatic classification of the state of said animal between a difficult parturition state and an easy parturition state of the method for detecting and signaling the preparatory phase farrowing an animal as described above, when executed by a processor.

The invention also relates to a recording medium readable by a processor on which is recorded a computer program for the implementation of the measurement processing step of the method for detecting and signaling the preparatory phase of bottom of an animal as described above and/or of the step of the step of binary automatic classification of the state of said animal between a difficult birth state and an easy birth state of the method detection and signaling of the preparatory phase of parturition of an animal as described above.

• un accéléromètre 3D destiné à mesurer les composantes de l'accélération de ladite portion de la queue selon 3 axes ;
• un support d'enregistrement tel que décrit précédemment et un processeur apte à exécuter le programme d'ordinateur tel que décrit précédemment, enregistré sur ledit support d'enregistrement pour traiter lesdites mesures ;
• des moyens de génération d'un premier signal de radiocommunication destiné à alerter de l'imminence de la mise-bas dudit animal, configurés de sorte à générer ledit premier signal dans le cas où lorsque le résultat de ladite classification automatique binaire est un état de préparation de la mise-bas ;
• des moyens d'émission dudit premier signal généré à destination d'un collecteur formant relai ou d'un terminal.

Finally, the invention also relates to a device for detecting and signaling the preparatory phase of parturition of an animal of the bovine family or of the equine family comprising a waterproof box capable of being fixed on a portion of the tail said animal containing:

• a 3D accelerometer intended to measure the components of the acceleration of said portion of the tail along 3 axes;
• a recording medium as described above and a processor capable of executing the computer program as described above, recorded on said recording medium to process said measurements;
• means for generating a first radio communication signal intended to alert of the imminence of the birth of said animal, configured so as to generate said first signal in the event that when the result of said binary automatic classification is a state of farrowing preparation;
• means for transmitting said first generated signal intended for a collector forming a relay or a terminal.

The device according to the invention thus cleverly integrates a software application recorded on a recording medium, ensuring complete processing of the measurements of the accelerometer, which makes it possible to limit the energy consumption of the device due to the fact that the quantity of signals that it emits to the outside is very small and which thus gives it a particularly extended energy autonomy.

Advantageously, said first signal transmitted by said transmitting means has a range of between 150 and 500 meters.

Thus, the signal can be transmitted to a relay located in, or close to, the barn or stall where the animal is parked, regardless of the location of the latter in the building. The signal can also be transmitted directly to a communication terminal of the breeder, such as a mobile telephone or a personal assistant for example, if the latter is near the building.

According to a particular embodiment of the invention, a device for detecting and signaling the preparatory phase of parturition of an animal as described above comprises means for generating a second radiocommunication signal intended to alert of the occurrence of a difficult farrowing of said animal, configured so as to generate said second signal in the event that said processor calculates a decision criterion f _t greater than or equal to said predetermined threshold value S, said generating means said second signal being able to be said means for generating said first signal, and in that said transmitting means are capable of transmitting said second signal generated intended for a collector forming a relay or a terminal.

5. List of Figures

• la figure 1 est une vue schématique d'un dispositif selon l'invention fixé sur la queue d'une vache ;
• la figure 2 est une représentation schématique du contenu du boitier présenté en référence à la figure 1 ;
• la figure 3 est une représentation, sous forme de diagramme des étapes d'un exemple de procédé de détection et de signalement de la phase préparatoire de mise-bas d'un animal selon l'invention ;
• la figure 4 détaille les sous-étapes de l'étape de classification automatique binaire de l'état d'un animal entre un état de mise-bas difficile et un état de mise-bas facile, du procédé présenté en référence à la figure 3 ;
• les figures 5A à 5E illustrent un exemple d'évolution au cours du temps des données Q, A, C, T et P à l'approche d'une mise-bas pour un animal.

Other characteristics and advantages of the invention will appear more clearly on reading the following description of an embodiment of the invention, given by way of simple illustrative and non-limiting example, and the appended drawings, among which:

• the figure 1 is a schematic view of a device according to the invention attached to the tail of a cow;
• the picture 2 is a schematic representation of the contents of the box presented with reference to the figure 1 ;
• the picture 3 is a representation, in the form of a diagram of the steps of an example of a method for detecting and signaling the preparatory phase of whelping an animal according to the invention;
• the figure 4 details the sub-steps of the binary automatic classification step of the state of an animal between a difficult parturition state and an easy parturition state, of the method presented with reference to the picture 3 ;
• them figures 5A to 5E illustrate an example of the evolution over time of the data Q, A, C, T and P as parturition approaches for an animal.

6. Detailed description of an exemplary embodiment of the invention

A device 10 for detecting and signaling the preparatory phase of farrowing according to the invention is shown with reference to the figure 1 , attached to the tail of a cow 11.

The device 10 is in the form of a sealed box 12 whose face 13 in contact with the tail of the cow 11 is of substantially frustoconical shape so as to closely match the shape of the tail.

As can be seen on the picture 2 , which is a schematic representation of the box 12, the latter contains a 3D accelerometer 21 connected to an electronic card 22 integrating a microcontroller 210, which processes the measurements provided by the accelerometer 21, and an integrated circuit 24 providing optimization of the energy consumed by the box 12. The software application making it possible to process the measurements is recorded in the flash memory 211 of the microcontroller 210, which is read by the processor μ1.

Inside the box 12, there is also an antenna 23 for transmitting a digital signal in accordance with the GSM standard to a relay or a mobile phone and a battery L5, which supplies the accelerometer 21, the card 22 and the antenna 23 in energy.

The microcontroller 210 fitted to the card 22 processes the measurements delivered by the accelerometer 21 and makes it possible to generate a first warning signal, when the preparatory phase of whelping is detected, and a second warning signal when a difficult calving is detected. The first and the second alert signal are transmitted by the antenna 23, in this particular embodiment of the invention.

In this particular embodiment of the invention, the measurements of the components of the acceleration are recorded according to a sampling frequency of 4 Hz.

It will be noted that the measurements of the acceleration are not kept after processing. Indeed, the device 10 being provided to be able to operate without interruption for several years, and more precisely from 5 to 10 years in this particular embodiment of the invention, providing a storage memory of sufficient capacity is not possible.

We represented on the picture 3 synoptically, in the form of a diagram, the steps E1 to E10 of processing the measurements of the method for detecting and signaling the preparatory phase of giving birth to a cow according to the invention implemented in the software application stored in the flash memory 211.

The components of the acceleration along three orthogonal axes x, y and z linked to the box 12 are measured every 25 ms by the accelerometer 21, in a step E0.

In a step E1, the measurements of the components are projected in a direct orthonormal frame X, Y, Z, where the Z axis is a vertical axis and directed downwards, the X axis is horizontal and in the extension of the sacrum of said animal. For this purpose, the position of the x, y, z marker linked to the box has been calibrated beforehand on the animal with respect to the X, Y, Z marker.

Is then determined in E2 for each instant of measurement, a datum Q representative of a raised or lowered position of said tail. During this step E2, the projected components of the acceleration are initially filtered by applying a first order filter (step E21) and a value of the angle of the tail with respect to the vertical θ _tail is calculated at from the filtered projections ax and az of the measurements of the components of the acceleration along the axes X and Z, by applying the formula θ _tail =atan(az/ax) (step E22). The calculated angle is then compared with a threshold angle value of 35° (step E23). A value 256 is assigned to Q for the measurement instant t _measurement if the calculated angle is greater than or equal to 35°; otherwise a value 0 is assigned to Q.

Subsequently (step E24) a mean of values of Q over the time interval [t _{measurement −Δt} ; t _measurement ] preceding the measurement instant t _measurement , with Δt=1 hour in this particular embodiment of the invention, so as to smooth out the variations of the data item Q.

In E3, a datum A representative of the activity of the animal is determined by calculating (step E31) a quadratic mean of the variations of the projected and filtered components of the acceleration along the axes X, Y and Z with respect to their value average over one minute.

In E4, a datum C representative of the occurrence of a colic attack in the animal is determined by applying (step 41) a band pass filter of the Chebyshev type of order 2 to the projections of the measurements of the components of the acceleration on the X and Z axes whose cutoff frequencies are 6 mHz and 12 mHz and the sum of the squares of the projections of the measurements filtered by the band pass filter is assigned (step 42) to the data C.

In E5, a datum T representative of the occurrence of contractions in the animal is determined by calculating an estimate (step 51) of the correlation coefficient, for each component of the acceleration projected on the X, Y or Z axis, between a sequence of 3.5s of the projection of the measurements of this component of the acceleration with a reference template formed of a data sequence of a 3.5s wide roughly triangular shaped spike. The correlation coefficient obtained on each axis X, Y and Z is then integrated over a period of 15 minutes (step 52) and the value T is assigned to the sum of the squares of the values of the correlation coefficient estimated on each axis X, Y and Z (step 53).

In E6, a datum TR representative of the occurrence of trampling in the animal is determined by identifying (step E61) a predominant frequency in the frequency spectrum for a sequence of the projection of the component of the acceleration according to the axis Y by one minute and a value of 1 is assigned (step E62) if the identified frequency is in the frequency range [150 mHz; 300mHz], -1 otherwise.

Examples of the evolution over time of data Q, A, C and T as a farrowing approaches are illustrated respectively on the figures 5A to 5D .

The processing of the measurements is completed by applying to the data Q, A, C, T and TR, in a step E7, a linear classifier of a binary automatic classification algorithm known per se, and by comparing the output of this classifier with a threshold S _class equal to 5.10 ⁵ in this particular embodiment of the invention (step E71). This classifier makes it possible to discriminate the state of the animal between a state of preparation for parturition and a non-preparatory state according to the value of the output of the classifier with respect to the threshold S _class .

It should be noted that this embodiment of the invention allows detection of the farrowing preparation phase in 88% of cases, without false alarms.

A first alert signal is generated by the microcontroller 210, when the result of the binary automatic classification is a state of preparation for farrowing, in step E8. This second alert signal is transmitted by antenna 23 in step E80.

In this particular embodiment of the invention, in a step E9, a datum P representative of the standing or lying state of said animal is also determined. In this step E9, the average over a period of 1 minute preceding the measurement instant t _measurement is compared (step E91) of the values of the projection on the Y axis of the measurements of the acceleration with respect to a state threshold value equal to 0.2 g and the value 1 is assigned (step E92), corresponding to a standing position, when the average is less than the state threshold value and the value 0 otherwise, corresponding to a lying position.

In a variant of this particular embodiment of the invention, it may be envisaged to determine the datum P from the number of sign changes of the acceleration on the Y axis over a predetermined period, for example of 1 minute. If this number of sign changes is greater than or equal to a threshold, for example equal to 10 over a period of 1 minute, the value 1, corresponding to a standing position, can then be assigned to the datum P, and 0 otherwise .

An example of evolution over time of the data P at the approach of a farrowing is illustrated on the figure 5E .

A step E10 of automatic binary classification of the state of the animal is then carried out between a difficult state of farrowing and an easy state of farrowing from the data Q, A, C, T and P.

A second warning signal is generated by the microprocessor 210 in a step E20, in the event that the output of the classifier implemented during step E10 is less than or equal to a predetermined threshold value S, equal to -0, 8741 in this particular embodiment of the invention. This second alert signal is transmitted by antenna 23 in step E30.

In relation to the figure 4 , step E10 is detailed.

This step E10 begins with the training (sub-step E101) from time T ₀ the time when said first alert signal is generated in step E8, at and each time t such t=m .ΔT+T ₀ with m an integer and ΔT= 1mn, of a vector of five continuous real variables (M ¹ ,M ² ,M ³ ,M ⁴ ,M ⁵ ) from the sequence of 240 successive data values Q, A, C, T and P over a period of one minute preceding time t. This vector of five continuous real variables (M ¹ ,M ² ,M ³ ,M ⁴ ,M ⁵ ) is obtained directly as follows: M _i ¹ =Q _i , M _i ³ =T _i , M _i ⁴ =C _i , M _i ⁵ =A _i , with i integer such that 1≤i≤240, and M _i ² is the duration since the last change from lying to standing or vice versa at index i estimated from the sequence of successive values of P.

Next, in a sub-step 102, a vector of characteristics F _t =(F _t ¹ ,F _t ² ,F _t ³ ,F _t ⁴ ,F _t ⁵ ) where F _t ^k ,1≤k≤5, is the maximum of the average values of M ^k during the hour preceding the instant t.

It should be noted that for the times preceding the time T ₀ , the value 0 is assigned to the discrete variables M ^k , 1≤k≤5.

In E103, each characteristic F ^k , 1≤k≤5, of the vector of characteristics is assigned a vote P _t ^k constructed from the ratio, in logarithmic scale, of the probability P(F ^k =F _t ^k |H ₀ ) of observing characteristic F ^k knowing that farrowing is easy and the probability P(F ^k =F _t ^k |H ₁ ) of observing characteristic F ^k knowing that farrowing is difficult, that is P _t ^k =log(P(F ^k =F _t ^k | H ₀ )/ P(F ^k =F _t ^k | H ₁ )).

P(F ^k =F _t ^k |H ₀ ) and P(F ^k =F _t ^k |H ₁ ) are, in sub-step E103, determined by comparison with a reference statistical distribution constructed from observations on a sample of reference animals.

où α₁ = 0,19, α₂ = 0,59, α₃ = 0,27; α₄ = 0,49 et α₅ = 0,27 et O _1=- 0,187, O _2=-0,107; O _3=- 0,206; O _4=- 0,003; O _5=- 0,138 dans ce mode de réalisation particulier de l'invention.A decision criterion f _t , or classifier, is then calculated in a sub-step E104, expressed in the form:

where α ₁ = 0.19, α ₂ = 0.59, α ₃ = 0.27; α ₄ = 0.49 and α ₅ = 0.27 and O ₁₌ -0.187, O ₂₌ -0.107; O _{3 =} -0.206; O _{4 =} -0.003; O _{5 =} -0.138 in this particular embodiment of the invention.

The value of this decision criterion f _t constitutes the output of the classifier implemented in state E10.

où α_k, 1≤k≤5, sont des constantes , et $c_t^k$

est la confiance apportée par la caractéristique k, 1≤k≤5, qui s'exprime sous la forme : $$c_t^k=\{\begin{array}{l}0.5\mathrm{si}{P}_t^{\left(0\right)k}\mathrm{ou}{P}_t^{\left(1\right)k}=0.\\ \frac{P_t^{\left(1\right)k}}{P_t^{\left(0\right)k}+P_t^{\left(1\right)k}}\mathrm{sinon}.\end{array}\mathrm{avec}{p}_t^{\left(0\right)k}=P\left(F^k=F_t^k|H_0\right)$$

la probabilité d'observer la caractéristique F^k sachant que la mise-bas est facile et $p_t^{\left(1\right)k}=P\left(F^k=F_t^k|H_1\right)$

la probabilité d'observer la caractéristique F^k sachant que la mise-bas est difficile.During step E10, a confidence index C _t is also calculated (sub-step E105) expressed in the form:

where α _k , 1≤k≤5, are constants, and ${\mathrm{vs}}_{\mathrm{you}}^k$

is the confidence brought by the characteristic k, 1≤k≤5, which is expressed in the form: $${\mathrm{vs}}_{\mathrm{you}}^k=\{\begin{array}{l}0.5\mathrm{if}{P}_{\mathrm{you}}^{\left(0\right)k}\mathrm{Where}{P}_{\mathrm{you}}^{\left(1\right)k}=0.\\ \frac{P_{\mathrm{you}}^{\left(1\right)k}}{P_{\mathrm{you}}^{\left(0\right)k}+P_{\mathrm{you}}^{\left(1\right)k}}\mathrm{otherwise}.\end{array}\mathrm{with}{p}_{\mathrm{you}}^{\left(0\right)k}=P\left(F^k=F_{\mathrm{you}}^k|{\mathrm{<figure-callout\; id="0"\; label="H"\; filenames="imgf0004.png"\; state="\{\{state\}\}">H</figure-callout>}}_0\right)$$

the probability of observing the characteristic F ^k knowing that farrowing is easy and $p_{\mathrm{you}}^{\left(1\right)k}=P\left(F^k=F_{\mathrm{you}}^k|{\mathrm{<figure-callout\; id="1"\; label="H"\; filenames="imgf0004.png"\; state="\{\{state\}\}">H</figure-callout>}}_1\right)$

the probability of observing the characteristic F ^k knowing that farrowing is difficult.

This confidence coefficient is transmitted to the breeder when the second alert signal is emitted.

The example embodiment detailed above makes it possible to detect a difficult calving in 82% of cases, with a confidence index greater than 0.78.

Furthermore, in step E40, the _end time t corresponding to the end of farrowing is determined. The purpose of this step E40 is to prevent a second alert signal from being emitted, untimely and inappropriately, after the end of calving.

This step E40 comprises the calculation (step 401) at each instant t, such as t=m.ΔT+T ₀ with m an integer and ΔT=1mn, of the sum St of the variations of the values of the data A representative of the activity on a time interval of 20 minutes preceding time t and detection of the maximum of the values of S _t over a period of 50 minutes, from which the _end time t is deduced (step 402).

In this particular embodiment of the invention, as soon as the _end time t is exceeded, the microcontroller is advantageously configured to no longer generate any second alert signal.

It will also be noted that an alert signal intended to signal the occurrence of an abnormally long farrowing is generated by the microcontroller 210 when no second alert signal is generated 2 hours after the generation of the first warning signal without the end of farrowing time t _end being exceeded.

Claims (15)

1. A method for detecting and signalling the preparatory phase of parturition of an animal of the bovine family or the equidae family, comprising a step of measuring the components of the acceleration of a portion of the tail of the said animal along three orthogonal axes, characterised in that it comprises the following steps in this order:

   - processing said measurements to detect an occurrence of said preparatory phase comprising:

   - a step of projecting said component measurements in a direct orthonormal frame X, Y, Z where the Z axis is vertical and directed downwards and the X axis is horizontal and in the extension of the sacrum of said animal, in case the frame formed by said three orthogonal axes does not coincide with said orthonormal frame X, Y, Z;

   - at least two of the following steps a), b), c), d) and e):

   a) determination, for each measurement instant, of a data item Q representative of a raised or lowered position of said tail, comprising a step of calculating a value of the angle of the tail with respect to the vertical on the basis of the projections of the measurements of the components of the acceleration along the X and Z axes, or of the measurements of the components of the acceleration, when the frame formed by said three orthogonal axes coincides with said orthogonal frame X, Y, Z, and a step of comparing said calculated angle with a predetermined threshold angle value;

   b) determination of a data item A representative of the activity of the said animal, for each instant of measurement, comprising a step of calculating the variations of each of the components of the acceleration with respect to its average value over a predetermined period on the basis of the projections of the measurements of the components of the acceleration along the X, Y and Z axes or of the measurements of the components of the acceleration, when the frame formed by the said three orthogonal axes coincides with the said orthogonal frame X, Y, Z;

   c) determining a data C representative of the occurrence of a colic attack in said animal comprising a step of applying a bandpass filter to said measurements of the acceleration components on at least two axes, and preferably the acceleration components on the X and Z axes;

   d) determining data T representative of the occurrence of contractions in said animal comprising a step of estimating the correlation coefficient, for each component of acceleration, between a sequence of measurements of that component of acceleration and a sequence of reference data;

   e) determining data TR representative of the occurrence of trampling in said animal comprising a step of identifying a predominant frequency in the frequency spectrum for a sequence of Y-axis acceleration component measurements to establish whether said frequency is within a predetermined frequency range;

   - automatic binary classification of the state of said animal between a state of preparation for parturition and a non-preparatory state, on the basis of said previously determined data Q, A, C, T and/or TR, representative respectively of a raised or lowered position of the tail, of activity, of the occurrence of a colic attack, of the occurrence of contractions and of the occurrence of trampling;

   - generation of a first alert signal when the result of said automatic binary classification is a parturition readiness.
2. The method for detecting and signalling the preparatory phase of parturition of an animal according to claim 1, characterised in that said step a) of determining a data item representative of a raised or lowered position of the tail and/or said step b) of determining a data item of activity and/or said step d) of determining a data item representative of the occurrence of contractions and/or said step e) of determining a data item representative of the occurrence of trampling comprises a step of filtering said projections of the measurements or said measurements.
3. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 2, characterised in that said step a) of determining a data item representative of a raised or lowered position of the tail comprises a step of assigning to said data item Q a first value, preferably 1, when said angle value is greater than or equal to said threshold angle value and a second value, preferably 0 or -1, when said angle value is less than said threshold angle value.
4. The method for detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 3, characterised in that said threshold angle value is between 32° and 37°, preferably substantially equal to 35°.
5. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 4, characterised in that said step a) of determining a data item Q representative of a raised or lowered position of said tail further comprises a step of reassigning to said value Q for each measurement instant t_measurement, of an average, over a predefined time interval Δt, of the values assigned to the data item Q for the measurement instants preceding the measurement instant t_measurement included in the interval [t_measurement-Δt; tmeasurement].
6. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 5, characterised in that said step a) of determining a data item Q representative of a raised or lowered position of said tail comprises a step of weighting said value of the angle of the tail by the value of the measurement of the component of the acceleration, or of its projection, on the Y axis
7. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 6, characterised in that said step b) of determining the activity of said animal comprises a step of calculating the root mean square of the variations of the components of the acceleration along the X, Y and Z axes with respect to their mean value.
8. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 7, characterised in that, during said step c) of determining a data item representative of the occurrence of a colic crisis, the cut-off frequencies of said bandpass filter are substantially equal to 6 mHz and 12 mHz.
9. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 8, characterised in that, in said step d) of determining a data item representative of the occurrence of contractions, said reference sequence of data is a sequence of data items representative of a substantially triangular shaped peak.
10. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 9, characterised in that, in said step e) of determining the occurrence of trampling, said frequency range is bounded by the frequencies 150mHz and 300mHz.
11. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 10, characterised in that it further comprises a step of determining a data item P representative of the upright or recumbent state of said animal comprising a step of comparing the average over a predetermined period, preferably substantially equal to 1 minute, of the values of the measurements of the components of the acceleration, or of their projection, on the Y axis in relation to a predetermined state threshold value and/or a step of comparing the number of changes in sign of the acceleration on the Y axis over the said predetermined period in relation to a threshold number.
12. The method of detecting and signalling the preparatory phase of parturition of an animal according to any one of claims 1 to 11, characterised in that it further comprises a step of automatic binary classification of the state of said animal between a difficult parturition state and an easy parturition state from at least three data, preferably at least four data, even more preferably at least five data, from among the data Q, A, C, T, TR and P defined in claims 1 or 11.
13. A computer program product comprising program code instructions for implementing the step of processing measurements of the method of detecting and signalling the preparatory phase of parturition an animal according to any one of claims 1 to 12 and/or the step of automatically binary classifying the state of said animal between a difficult parturition state and an easy parturition state of the method of detecting and signalling the preparatory phase of parturition of an animal according to claim 12, when executed by a processor.
14. A processor-readable recording medium on which is recorded a computer program for implementing the step of processing the measurements of the method of detecting and signalling the preparatory phase of parturition of an animal according to one of claims 1 to 12 and/or the step of automatic binary classification of the state of said animal between a difficult parturition state and an easy parturition state of the method of detecting and signalling the preparatory phase of parturition of an animal according to claim 12.
15. A device for detecting and signalling the preparatory phase of parturition of an animal of the bovine family or the equine family comprising a watertight case adapted to be fixed to a portion of the tail of said animal containing:

    - a 3D accelerometer for measuring the components of the acceleration of the said portion of the tail along 3 axes;

    - a recording medium according to claim 14 and a processor adapted to execute the computer program according to claim 13 recorded on said recording medium to process said measurements;

    - means for generating a first radio communication signal for alerting of the imminence of parturition of said animal, configured to generate said first signal in case the result of the automatic binary classification is a state of readiness for parturition;

    - means for transmitting said first generated signal to a relay collector or terminal.

Images