Electronic device for monitoring the neurophysiological state of an operator in an aircraft control station, monitoring method and associated computer program

Abstract

Title: Electronic device for monitoring the neurophysiological state of an operator in an aircraft control station, associated monitoring method and computer program. The invention relates to an electronic device for monitoring (22) the neurophysiological state of an operator (14) in an aircraft control station (16) comprising: a receiving module configured to receive data from a sensor (24); a categorization module configured to associate a category with the operator (14) from the received data; a processing module configured to extract from each data at least one parameter representative of the neurophysiological state of the operator (14);a detection module configured to apply a model from a machine learning method to determine, based on representative parameters, whether the operator (14) is in a nominal neurophysiological state or in an altered neurophysiological state, the model being chosen from a list of predetermined models based on the category associated with the operator. Figure for the abstract: Figure 2;

Description

Title of the invention: Electronic device for monitoring the neurophysiological state of an operator in an aircraft control station, monitoring method and associated computer program

[0001] The present invention relates to an electronic device for monitoring the neurophysiological state of an operator in an aircraft.

[0002] The invention also relates to a method for controlling and monitoring the neurophysiological state of an operator in an aircraft.

[0003] The invention also relates to a computer program comprising software instructions which, when executed by a computer, implement such a process.

[0004] The aircraft is typically an airplane, a helicopter, or a drone. The operator is, for example, the aircraft pilot, the co-pilot, or a radar operator. The control station is located either inside the aircraft or, in the case of a drone, remotely.

[0005] Monitoring the neurophysiological state of such an operator is essential for aircraft safety in order to detect any adverse condition: fatigue, mental workload, incapacitation, stress, etc., which could lead to an incapacity of the operator or a deterioration of his ability to perform the tasks planned under operational flight conditions.

[0006] Conventionally, the neurophysiological state of an operator is monitored by the other aircraft operators. For example, the co-pilot monitors the pilot's neurophysiological state, and vice versa.

[0007] To complement this monitoring, it has been proposed to use one or more sensors placed in the aircraft for monitoring. The sensors measure certain data, and the neurophysiological state of the operator is deduced from this.

[0008] However, such a method is not entirely satisfactory. Indeed, this method is not robust enough. In particular, the monitoring method must be robust to the variability of the situations encountered. In fact, the detection of an altered neurophysiological state must be robust to the variability of the physiological characteristics of the different operators (age, gender, etc.), but also to the variability of the environment in which the operator is located (aircraft cockpit in turbulence, cockpit where ambient light varies, etc.).

[0009] Furthermore, the method must be robust despite the limited operational data available corresponding to the desired behaviors. In the military field, This problem is all the more complex because the data collected and useful for the application of interest are not necessarily available since they are often classified, and few in number due to the difficulty of observing critical situations.

[0010] To meet this requirement, methods have been proposed to monitor only a specific mental state of the operator, such as fatigue or mental workload, etc. The method is then trained on a limited amount of data. Although monitoring may be satisfactory for training the algorithm, generalizing the model, that is, adapting it to a new, unknown subject, often leads to a sharp drop in performance. Such a method is therefore very specific to the data on which the training is performed and is thus, again, not robust enough to the variabilities encountered subsequently.

[0011] Conversely, other methods have been proposed where algorithms are trained on large amounts of data, covering a large number of variabilities, but they then offer lower performance, insufficient for the aeronautical application and the strong safety constraints that apply in this field.

[0012] In general, it appears from conventional methods and the literature that a choice must be made between the performance of the method and its coverage, that is to say the quantity and variability of the subjects on which the algorithm of the method has been trained.

[0013] Finally, it should be noted that the method must be implementable in an aircraft and therefore with limited onboard computing resources, while determining the neurophysiological state of the operator in near "real time". For example, detecting a pilot's loss of consciousness during takeoff must be possible in less than a few seconds, typically less than two seconds.

[0014] One object of the invention is then to propose an electronic device for monitoring the neurophysiological state of an operator in an aircraft, making it possible to overcome the difficulties explained above, in particular by offering accurate, reliable and robust monitoring of the variability of the monitored operator.

[0015] To this end, the invention relates to an electronic device for monitoring the neurophysiological state of an operator in an aircraft, the monitoring device comprising:

[0016] - a receiving module configured to receive data from at least one sensor onboard the aircraft, each sensor is configured to measure at least one piece of information relating to the operator;

[0017] - a categorization module configured to associate a category with the operator from a list of predetermined categories based on the data received;

[0018] - a processing module configured to extract from each data point at least one parameter representative of the neurophysiological state of the operator;

[0019] - a detection module configured to receive the category associated with the operator and the representative parameter(s), the detection module being further configured to apply a model from a machine learning method to determine, based on the representative parameters, whether the operator is in a nominal neurophysiological state or in an altered neurophysiological state, the model being chosen from a list of predetermined models based on the category associated with the operator.

[0020] Thus, the present invention addresses the problem of the trade-off between performance and monitoring coverage. Indeed, the prior categorization of the operator, despite the possible combinations of variability, ensures sufficient monitoring coverage while offering higher performance for each specific model. Overall, by averaging the performance of the models specific to each operator category, a higher detection performance is obtained than with a single model that attempts to cover all operator variability. In effect, operator monitoring is performed by a specific model for each category, while ensuring coverage of all combinations through the separation into different categories.

[0021] According to other advantageous aspects of the invention, the electronic monitoring device comprises one or more of the following features, taken individually or in all technically possible combinations:

[0022] - the monitoring device further includes an alert module configured to emit an alert signal when the detection module determines that the operator is in an altered neurophysiological state; - Each sensor is chosen from the group consisting of:

[0023] - a cardiac sensor, in particular an electrocardiograph;

[0024] - a pulse oximeter, in particular a photoplethysmography sensor;

[0025] - a breathing sensor;

[0026] - an accelerometer;

[0027] - a cranial electrode, for example an electroencephalograph;

[0028] - a pressure sensor arranged in an operator's seat;

[0029] - a pressure sensor arranged in a control device suitable for being activated by the operator;

[0030] - an operator perspiration sensor;

[0031] - an electrodermal response sensor;

[0032] - a camera configured to capture at least one image comprising at least one part of the operator;

[0033] - a microphone;

[0034] - an infrared sensor for the operator's skin temperature;

[0035] - an internal temperature sensor for the operator;

[0036] - a near-infrared spectroscopy headband.

[0037] - the categorization module is configured to associate a category based on at least one so-called individual attribute chosen from the group consisting of: gender, age, ethnic origin, body hair, hair length, presence of elements on the skin of the face; - the categorization module is configured to associate a category based on at least one attribute called an accessory worn, chosen from the group consisting of: wearing glasses, polarized or non-polarized glasses, contact lenses, surgical mask, gas mask, headphones, cap; - the processing module is configured to extract from each data at least one parameter representative of the neurophysiological state of the operator according to the category associated with the operator; - The processing module is configured to extract from each data point at least one parameter representative of the operator's neurophysiological state by implementing for each data point an algorithm chosen from the group consisting of:

[0038] - an extraction of a predetermined characteristic of the associated data followed of a machine learning method;

[0039] - a deep learning method applied directly to the associated data

[0040] - a predetermined model applied to the associated data.

[0041] - the algorithm implemented is chosen according to the category of the operator.

[0042] The invention also relates to a method for monitoring the neurophysiological state of an operator in an aircraft control station, the method comprising at least the following steps:

[0043] - receiving data from at least one sensor on board the aircraft, each sensor being configured to measure at least one piece of information relating to the operator; - association to the operator of a category from a list of predetermined categories based on the data received; - extraction from each data at least one parameter representative of the neurophysiological state of the operator; - reception of the category associated with the operator and the representative parameter(s) and application of a model from a learning method to automatic to determine based on representative parameters whether the operator is in a nominal neurophysiological state or in an altered neurophysiological state, the model being chosen from a list of predetermined models based on the category associated with the operator.

[0044] The invention also relates to a computer program comprising software instructions which, when executed by a computer, implement a monitoring method as defined above.

[0045] These features and advantages of the invention will become clearer upon reading the following description, given solely by way of non-limiting example, and made with reference to the accompanying drawings, in which:

[0046] [Fig-1] [Fig.1] is a schematic representation of an aircraft comprising an monitoring device according to the invention;

[0047] [Fig.2] [Fig.2] is a schematic representation of a control station in the aircraft of [Fig.1],

[0048] [Fig.3] [Fig.3] is a flowchart of a process for constructing and training a model, and

[0049] [Fig.4] [Fig.4] is a flowchart of a monitoring method according to the invention implemented by the electronic device.

[0050] An aircraft 12 is shown in [Fig.1].

[0051] The aircraft 12 is typically an airplane, a helicopter, or even a drone. In other words, the aircraft 12 is a flying machine that can be piloted by an operator 14 via a control station 16. The control station 16 is located inside the aircraft 12 or remotely from the aircraft 12, particularly in the case of a drone.

[0052] The operator 14 is here a pilot but the invention applies similarly to any operator of the aircraft 12 such as a co-pilot or a radar operator.

[0053] As seen in [Fig.2], the control station 16 is here a cockpit of the aircraft 12. As seen in [Fig.1], the control station 16 includes at least one seat 18 for the operator 14, a control device 19 suitable for operation by the operator 14, a windscreen 20 at least partially transparent and separating the interior of the cockpit from the external environment of the aircraft 12, at least one sensor 24 and an electronic monitoring device 22 of a neurophysiological state of the operator 14.

[0054] Each sensor 24 is configured to measure at least one piece of information relating to the operator 14 and in particular to his neurophysiological state, as will be explained in more detail later.

[0055] Each sensor 24 is, in particular, a so-called wearable sensor or a so-called remote sensor.

[0056] A wearable sensor is a sensor designed to be in physical contact with the operator 14. A person skilled in the art will understand that "in contact" means that the sensor 24 touches a part of the operator, possibly with clothing between the sensor and the operator's skin 14. Thus, the wearable sensor 24 is presented for example in the form of a watch on the operator's wrist, a helmet on the operator's head 14 or a sensor integrated into the control device 19 or into the seat 18.

[0057] A remote sensor is a sensor arranged at a distance from the operator 14 during flight operating conditions. Those skilled in the art will understand that "at a distance" means that there is an empty space between the sensor 24 and the operator 14.

[0058] In particular, each sensor 24 is chosen from the group consisting of:

[0059] - a cardiac sensor, in particular an electrocardiograph; - a pulse oximeter, in particular a photoplethysmography sensor; - a breathing sensor; - an accelerometer; - a cranial electrode, for example an electroencephalograph; - a pressure sensor arranged in the seat 18 of the operator 14; - a pressure sensor arranged in a control device 19 specific to to be operated by operator 14; - an operator perspiration sensor; - an electrodermal response sensor; - a camera configured to capture at least one image including at least part of the operator, in particular the eyes for eye tracking; - a microphone; - an infrared sensor for operator skin temperature 14; - an internal temperature sensor of the operator 14; - a near-infrared spectroscopy band (also called a band "fNIRS") using near-infrared light to monitor brain activity.

[0060] The electronic monitoring device 22 is configured to monitor a neurophysiological state of the operator 14. The neurophysiological state relates to the nervous system of the operator 14. It is representative of the ability of the operator 14 to act to perform the tasks necessary for the safety of the aircraft 12, for example piloting the aircraft 12 or responding to external communications for a pilot.

[0061] In particular, the neurophysiological state may be a so-called "nominal" neurophysiological state, corresponding to the expected neurophysiological state of the operator 14 during a flight of an aircraft 12, that is to say an awake and lucid state.

[0062] The neurophysiological state may be an "altered" neurophysiological state, corresponding to a neurophysiological state in which the neurophysiological state of the operator 14 impacts their ability to act to ensure the safety of the aircraft 12. An altered neurophysiological state is, for example, a state of stress, fatigue, excessive or insufficient mental workload, hypo- or hypervigilance (or tunnel vision), etc. An altered neurophysiological state can also be a state of loss, at least partial, of awareness of the external world by operator 14, such as a state of drowsiness, sleep, or fainting. In this neurophysiological state, operator 14 has impaired or even nonexistent awareness of their environment and cannot react accordingly. This altered neurophysiological state is problematic during the flight of aircraft 12 because operator 14 is unable to perform the tasks they are required to accomplish in a reactive and appropriate manner.

[0063] For this purpose, the monitoring device 22 is configured to determine whether the operator 14 is in a nominal neurophysiological state or in an altered neurophysiological state.

[0064] In particular, the monitoring device 22 includes a receiving module 30, a categorization module 31, a processing module 32 and a detection module 34.

[0065] The monitoring device 22 advantageously includes further an alert module 36.

[0066] The receiving module 30 is configured to receive data from at least one sensor 24 on board the aircraft 12, in particular in the control station 16.

[0067] The categorization module 31 is configured to associate with the operator 14 a category from a list of predetermined categories from the data received by the receiving module 30.

[0068] A category is a class within which elements are arranged according to a number of criteria. In this case, predetermined categories allow the different operators 14 to be classified according to a list of predetermined attributes. Thus, a category makes it possible to characterize the variabilities of each operator 14 by associating it with a category.

[0069] The list of categories is predetermined before the operational flight phases, notably with the help of experts in the field.

[0070] The categorization module 31 is configured to associate a category with the operator 14 based on attributes relating to the operator 14.

[0071] In particular, the categorization module 31 is configured to associate a category based on at least one so-called individual attribute. An individual attribute is a physical or physiological attribute specific to the operator 14. Each individual attribute is notably chosen from the group consisting of: gender, age, ethnic origin, body hair, hair length, presence of features on the facial skin, such as tattoos, makeup, scars, etc.

[0072] Alternatively or in addition, the categorization module 31 is configured to associated a category based on at least one attribute called a worn accessory. A worn accessory attribute is an attribute relating to a garment or object worn by the operator 14. Each worn accessory attribute is notably chosen from the group consisting of: wearing glasses, polarized or non-polarized glasses, contact lenses, surgical mask, gas mask, headphones, cap.

[0073] The processing module 32 is configured to extract from each data at least one parameter representative of the neurophysiological state of the operator 14.

[0074] Those skilled in the art will understand that the categorization module 31 and the processing module 32 do not necessarily use the same data from the sensors. For example, one sensor such as a camera is used by the categorization module 31 and another sensor such as a heart rate sensor is used by the processing module 32. However, in one possible embodiment, both modules 31 and 32 use the same data, for example, images from a camera in the control station 16.

[0075] A parameter representative of the neurophysiological state is a parameter defined, for example, by experts in the field, that provides information on the neurophysiological state of the operator. For example, a low heart rate, eyes closed for a long period of time, constant pressure applied, a tilted head position, etc., are parameters that allow us to determine that the operator is in an altered neurophysiological state.

[0076] Advantageously, the processing module 32 is configured to extract from each data at least one parameter representative of the neurophysiological state of the operator 14 according to the category associated with the operator 14.

[0077] In particular, the processing module 32 is configured to adapt which parameter from a list of possible parameters is extracted according to the category of the operator 14.

[0078] By way of example, the processing module 32 can adapt the image extracted parameter according to whether or not the operator 14 is wearing glasses. For example, if the operator 14 is not wearing glasses, the processing module 32 extracts eye blinks from the images, whereas if the operator 14 is not wearing glasses, the processing module 32 extracts a head movement.

[0079] Alternatively or in addition, the processing module 32 is configured to adjust the way a parameter is extracted according to the category associated with the operator 14.

[0080] By way of example, the detection threshold for a blink of the operator 14's eye is adjusted according to the category of the operator 14. Indeed, the detection threshold for a blink is not the same between an Asian person and a Caucasian person, due to the average opening of the eyelid or the characteristics of the epicanthic fold.

[0081] Advantageously, the processing module 32 is configured to extract from each data at least one parameter representative of the neurophysiological state of the operator 14 by performing an extraction of a predetermined characteristic of the associated data followed by a machine learning method.

[0082] By way of example, the characteristic is the head position of operator 14 extracted from a video obtained by a camera. Then a machine learning method is implemented to process the head position over time and deduce the parameter representative of the neurophysiological state of operator 14.

[0083] A machine learning method makes it possible to obtain a model capable of solving tasks without being explicitly programmed for each of these tasks. Machine learning comprises two phases. The first phase consists of defining a model from data present in a training database, also called observations. Defining the model here notably involves recognizing a deleterious neurophysiological state. This so-called learning phase is generally carried out prior to the practical use of the model. The second phase corresponds to the use of the model: once the model is defined, new data can then be submitted to the model in order to determine the neurophysiological state of the operator 14.

[0084] Alternatively, the processing module 32 is configured to extract from each data at least one parameter representative of the neurophysiological state of the operator 14 by implementing a deep learning method applied directly to the associated data.

[0085] A deep learning method is a technique based on the neural network model: dozens or even hundreds of layers of neurons are stacked to increase the model's complexity. In particular, a neural network is generally composed of a succession of layers, each taking its inputs from the outputs of the previous one. Each layer is composed of a plurality of neurons, taking their inputs from the neurons in the preceding layer. Each synapse between neurons is associated with a synaptic weight, so that the inputs received by a neuron are multiplied by this weight and then summed by that neuron. The neural network is optimized by adjusting the various synaptic weights during its training based on the data in the training database. The optimized neural network is then the model.A new dataset can then be given as input to the neural network, which then provides the result of the task for which it was trained.

[0086] Alternatively, the processing module 32 is configured to extract from each data at least one parameter representative of the neurophysiological state of operator 14 by implementing a predetermined modeling applied to the associated data.

[0087] Predetermined modeling is, by example, a physical model comprising a set of rules predetermined by a domain expert.

[0088] In an advantageous embodiment, the processing module 32 is configured to extract from each data point at least one parameter representative of the neurophysiological state of the operator 14 by implementing an algorithm chosen according to the category of the operator 14 from the group consisting of:

[0089] - an extraction of a predetermined feature from the associated data followed by a machine learning method; - a deep learning method applied directly to the associated data; - a predetermined model applied to the associated data.

[0090] For example, for a category in which the associated deep learning method has been trained on a large amount of data with good performance results, this method will be preferred. Conversely, for a category with little experience data and where machine learning methods are less effective, a predetermined model is preferred to avoid false positives.

[0091] The detection module 34 is configured to receive the category associated with the operator 14 determined by the categorization module 31 and the representative parameter(s) determined by the processing module 32.

[0092] The detection module 34 is further configured to apply a model from a machine learning method to determine, based on representative parameters, whether the operator 14 is in a nominal neurophysiological state or in an altered neurophysiological state.

[0093] The machine learning method used by the detection module 34 is different from, or possibly similar to, that used by the processing module 32.

[0094] The model is chosen from a list of predetermined models according to the category associated with operator 14.

[0095] Fig. 3 represents a flowchart of the process of constructing and training such predetermined models.

[0096] Initially, the experimental protocol is established. This involves defining the neurophysiological state to be monitored:

[0097] - the method of generating this state in subjects, - the relevant sensors for monitoring it, - the means of collecting the "ground truth", also called the label, i.e. the state in which the subject finds themselves (for example: normal fatigue or fatigue extreme), - the number of subjects required to ensure sufficient representativeness of the model, as well as inter-individual variability or variability in terms of the objects worn that we want to address, - the collection environment (laboratory, flight, etc.) and its representativeness in relation to the final use of the system, - potential biases related to experimentation.

[0098] During this protocol definition phase, the method includes a step of identifying the different combinations of variabilities to be monitored. In particular, this step studies which parameters are significant and relevant for characterizing the neurophysiological state of operator 14 according to each category, notably with the help of experts in the field.

[0099] The combinations of variables may depend on only one parameter, such as creating a model for men and one for women, but they may also contain a large number of parameters, for example, a model for Asian men wearing glasses. The number of combinations determines the number of models that need to be created.

[0100] Following this, data are collected during the campaigns. This results in a labeled physiological database, i.e. associated with a specific physiological state.

[0101] Since this data is raw, it is processed. The data quality is then assessed based on additional criteria. For example, the quality of face detection can be determined using a face-tracking algorithm. Data sampling, the presence of artifacts or sensor noise, are also checked. Signal trends, i.e., low-frequency elements irrelevant to monitoring, can also be removed.

[0102] Following data collection, the data are grouped into identified combinations of attributes to form the different combinations. A model is then developed for each category, i.e., each combination.

[0103] Thus, after collecting training data on a plurality of operators, these different operators are, for example, classified according to "gender" variability, i.e., male or female, or, for example, according to "ethnic origin" variability, i.e., Caucasian, African, Asian, Mediterranean, etc. The categories can also be formed by combining two or more variabilities for each operator, for example, male and Asian or female and Australian, and the model trained on data from only Asian men or Australian women in this example.

[0104] At each stage of the process, the elements are advantageously transmitted to an online platform authorized to host GDPR (General Data Protection Regulation) data data protection) and HDS (health data hosting).

[0105] The alert module 36 is configured to emit an alert signal when the detection module 34 determines that the operator is in an altered neurophysiological state.

[0106] The warning signal is for example an audible signal emitted in the control station 16 with the aim of bringing the operator 14 back to a nominal neurophysiological state.

[0107] Alternatively or in addition, the warning signal is for example a signal sent to an aircraft control system 12 to switch to automatic mode so that the tasks to be performed by the operator 14 are carried out autonomously without the intervention of the operator 14. In particular, when the operator 14 is a pilot, the aircraft 12 switches to autopilot.

[0108] Alternatively or in addition, the warning signal is for example a communication signal to an external control device to the aircraft 12 such as a control tower.

[0109] In the example of [Fig. 1], the electronic monitoring device 22 comprises an information processing unit consisting, for example, of a memory and a processor associated with the memory. The receiving module 30, the categorization module 31, the processing module 32, the detection module 34, and, optionally, the alerting module 36 are each implemented as software, or a software component, executable by the processor. The memory is thus capable of storing receiving software, categorization software, processing software, detection software, and, optionally, alerting software. The processor is then capable of executing each of these software components.

[0110] In an alternative not shown, the receiving module 30, the categorization module 31, the processing module 32, the detection module 34, and as an optional complement, the alert module 36 are each implemented as a programmable logic component, such as an FPGA (Field Programmable Gate Array), or as a dedicated integrated circuit, such as an ASIC (Application Specified Integrated Circuit).

[0111] When the electronic device 22 is implemented in the form of one or more software programs, i.e., in the form of a computer program, it is also capable of being stored on a computer-readable medium (not shown). The computer-readable medium is, for example, a medium capable of storing electronic instructions and being connected to a bus of a computer system. For example, the readable medium is an optical disc, a magneto-optical disc, a ROM, a RAM, any type of non-volatile memory (e.g., EPROM, EEPROM, FLASH, NVRAM), a magnetic card, or an optical card. A computer program containing software instructions is then stored on the readable medium.

[0112] The operation of the electronic monitoring device 22 according to the invention will now be explained with the help of [Fig.4] representing a flowchart of the method, according to the invention, of monitoring a neurophysiological state of an operator 14 in a control station 16 of an aircraft 12.

[0113] Initially, aircraft 12 is in an operational flight situation, flying for example towards an airport.

[0114] At least one operator 14 is present in the control station 16 of the aircraft 12. The operator 14 is, for example, a pilot, as illustrated here.

[0115] For example, the pilot is, for instance, an Asian man wearing glasses.

[0116] The method includes an initial step 100 of receiving by the receiving module 30 data from at least one sensor 24 on board the aircraft 12.

[0117] Each sensor 24 is configured to measure at least one piece of information relating to the pilot.

[0118] Here, as illustrated by [Fig.2], the receiving module 30 receives, for example, images of the pilot's face from a camera located in the control station 16 and the pilot's heart rate data from the watch on the pilot's wrist.

[0119] Then, the process includes a step 110 of associating to the operator 14 a category from a list of predetermined categories from the data received by the categorization module 31.

[0120] In the preceding example, the categorization module 31 determines from the images that the pilot's gender is "male", that the pilot's "ethnic origin" is "Asian", and that the object worn on the face is "glasses". Based on these attributes, the categorization module 31 associates the pilot with the category "Asian male wearing glasses", which is one of the predetermined categories available to the categorization module 31.

[0121] Then, the process includes a step 120 of extracting each data from at least one parameter representative of the neurophysiological state of the operator 14.

[0122] In the preceding example, the processing module 32 extracts cardiac data from the pilot, including heart rate, and compares it to a resting threshold value below which there is a suspicion of an altered neurophysiological state of the pilot. Alternatively or in addition, the processing module 32 extracts the pilot's head movement from the images.

[0123] Advantageously, the processing module 32 extracts from each data at least one parameter representative of the neurophysiological state of the operator according to the category associated with the operator 14.

[0124] In the previous example, the processing module 32 avoids extracting the blink rate from his eyes because this extraction is less reliable due to the presence of the glasses.

[0125] The method then includes a step 130 of receiving the category associated with operator 14 and the representative parameter(s) by the detection module 34 and applying a model from a machine learning method to determine, based on the representative parameters, whether operator 14 is in a nominal neurophysiological state or in an altered neurophysiological state. The model is chosen from a list of predetermined models based on the category associated with operator 14.

[0126] Continuing with the previous example, the detection module 34 implements the model associated with the category "Asian man with glasses." This model has been specifically trained on training data relating to this pilot category and is therefore particularly effective at determining the pilot's neurophysiological state. Here, based on the pilot's heart rate and / or head movements, the detection module 34 determines whether the pilot is in a nominal or altered neurophysiological state. For example, if the detection module 34 receives a low heart rate and / or a lateral head movement, the model can determine whether the pilot is in an altered neurophysiological state.

[0127] Advantageously, the method includes a step 140 of emitting an alert signal when the detection module 34 determines that the operator 14 is in an altered neurophysiological state.

[0128] It is therefore understandable that the present invention offers a number of advantages.

[0129] Indeed, the device according to the invention allows for effective monitoring of the neurophysiological state of operator 14 while remaining robust to the variability of the operators being monitored. The prior categorization of operator 14 ensures sufficient monitoring coverage, while offering higher performance on each specific model. In fact, the monitoring of operator 14 is carried out by a model associated with each category, specifically trained on data relating to that category. Each specific model therefore offers higher performance than a single model that would attempt to cover all possible operator variability.

[0130] The invention therefore allows for better detection of altered neurophysiological states in operators and reduces false positives. Thus, the invention improves aircraft safety.

Claims

Demands

1. An electronic monitoring device (22) for the neurophysiological state of an operator (14) in a control station (16) of an aircraft (12), the monitoring device (22) comprising: - a receiving module (30) configured to receive data from at least one sensor (24) on board the aircraft (12), each sensor being configured to measure at least one piece of information relating to the operator (14); - a categorization module (31) configured to associate the operator (14) with a category from a list of predetermined categories based on the received data; - a processing module (32) configured to extract from each piece of data at least one parameter representative of the neurophysiological state of the operator (14);- a detection module (34) configured to receive the category associated with the operator (14) and the representative parameter(s), the detection module (30) being further configured to apply a model from a machine learning method to determine, based on the representative parameters, whether the operator (14) is in a nominal neurophysiological state or in an altered neurophysiological state, the model being chosen from a list of predetermined models based on the category associated with the operator (14).

2. Monitoring device (22) according to claim 1, wherein the monitoring device (22) further comprises an alert module (36) configured to emit an alert signal when the detection module (34) determines that the operator (14) is in an altered neurophysiological state.

3. Monitoring device (22) according to claim 1 or 2, wherein each sensor (24) is selected from the group consisting of: - a cardiac sensor, in particular an electrocardiograph; - a pulse oximeter, in particular a photoplethysmography sensor; - a respiration sensor; - an accelerometer; - a cranial electrode, for example an electroencephalograph; - a pressure sensor arranged in a seat (18) of the operator (14); - a pressure sensor arranged in a control device (19) suitable for operation by the operator (14); - a perspiration sensor of the operator; - an electrodermal response sensor; - a camera configured to capture at least one image including at least a part of the operator (14); - a microphone; - an infrared skin temperature sensor of the operator (14); - an internal temperature sensor of the operator (14); - a near-infrared spectroscopy headband.

4. A monitoring device (22) according to any one of the preceding claims, wherein the categorization module (31) is configured to associate a category based on at least one so-called individual attribute chosen from the group consisting of: gender, age, ethnic origin, body hair, hair length, presence of facial skin features.

5. Monitoring device (22) according to any one of the preceding claims, wherein the categorization module (31) is configured to associate a category based on at least one attribute of an accessory worn, chosen from the group consisting of: wearing glasses, polarized or non-polarized glasses, contact lenses, surgical mask, gas mask, headphones, cap.

6. Monitoring device (22) according to any one of the preceding claims, wherein the processing module (32) is configured to extract from each data at least one parameter representative of the neurophysiological state of the operator according to the category associated with the operator (14).

7. A monitoring device (22) according to any one of the preceding claims, wherein the processing module (32) is configured to extract from each data point at least one parameter representing sensing the neurophysiological state of the operator (14) by implementing for each data point an algorithm chosen from the group consisting of: - an extraction of a predetermined characteristic from the associated data followed by a machine learning method; - a deep learning method applied directly to the associated data; - a predetermined model applied to the associated data.

8. Monitoring device (22) according to claim 7, wherein the implemented algorithm is chosen according to the category of the operator (14).

9. A method for monitoring the neurophysiological state of an operator (14) in a control station (16) of an aircraft (12), the monitoring method comprising at least the following steps: - receiving (100) data from at least one sensor (24) on board the aircraft (12), each sensor being configured to measure at least one piece of information relating to the operator (14); - associating (110) with the operator (14) a category from a list of predetermined categories from the received data; - extracting (120) from each piece of data at least one parameter representative of the neurophysiological state of the operator (14);- reception (130) of the category associated with the operator and of the representative parameter(s) and application of a model from a machine learning method to determine, according to the representative parameters, whether the operator (14) is in a nominal neurophysiological state or in an altered neurophysiological state, the model being chosen from a list of predetermined models according to the category associated with the operator (14).;

10. A computer program comprising software instructions that, when executed by a computer, implement a monitoring method according to the preceding claim.

Images