Apparatus and computer implemented method for determining behavior of a target user

Abstract

The invention relates to an apparatus (10) for determining a behavior of a target user, the apparatus comprising: - a sensing unit (20) configured to face at least one eye region (44) of a target user, the sensing unit (20) being configured to acquire a plurality of target signals representing a change of at least one feature of the at least one eye region (44) of the target user, the sensing unit comprising at least one sensor and the at least one sensor being oriented towards the eye region (44); and - a controller configured to: - provide a machine learning algorithm, - provide a plurality of objective target data related to the acquired target signals as input to the machine learning algorithm, - determine a behavior of the target user as output of the machine learning algorithm.

Description

Technical Field

[0001] This invention relates to the field of determining a user's behavior, particularly by considering at least one eye region. Specifically, this invention relates to an apparatus for determining the behavior of a target user. The invention further relates to a computer-implemented method for determining the behavior of a target user. Background Technology

[0002] The eye region is primarily composed of the eye, eyelids, and the muscles surrounding the eye. Depending on the conditions a user encounters or the actions they intend to perform, a user may engage in behaviors that alter one or more of these components of the eye region. Such behaviors can be changes in visual comfort (e.g., glare), eye movements, changes in pupil size, eyelid movements, or muscle strain. Typically, a user's behavior is a response to external stimuli (e.g., light), spontaneous actions performed by the user (e.g., focusing their gaze on an object), or psychological reactions (e.g., emotions experienced by the user). Determining the link between changes in the eye region and user behavior is highly complex because the eye region comprises numerous muscles and can be in various states.

[0003] It is known in the art to use sensors to measure specific physiological characteristics of a user's eye region in order to correlate changes in these specific physiological characteristics with behavior. An example of this method is provided in EP 19306731. This known method implies that the components or parts of the eye region measured by the sensors are predetermined, and behavior is determined based on these specific measurements. This method requires first determining the correlation between changes in the physiological characteristics of the eye region and the user's behavior. As an example, the correlation between eye closure or changes in eyelid position and glare must first be determined. However, it is difficult to correlate changes in the activity of specific muscles around the eyes with the user's behavior. Therefore, it is even more difficult to correlate the various changes encountered in the eye region with the user's behavior. Furthermore, the same changes in the eye region for two different users may not necessarily be interpreted as the same behavior for those two users. In fact, for one user, glare may mean eye closure, while for another user, glare may only mean changes in the activity of muscles around the eyes. Therefore, even basic behaviors may be difficult to correlate with physical changes in the eye region.

[0004] Therefore, more complex behaviors involving limited variations or complex combinations of variations in the eye region may be even more difficult to determine using analysis.

[0005] Furthermore, identifying the complex behaviors that cause numerous changes in the eye region can mean using many sensors and powerful computing systems. However, even with powerful computing systems, it is difficult to accurately identify complex user behaviors using this analytical approach.

[0006] Therefore, there is a need for a device that can more accurately correlate changes in a user's eye region with that user's behavior. Summary of the Invention

[0007] Therefore, the present invention provides an apparatus for determining the behavior of a target user, the apparatus comprising:

[0008] - A sensing unit configured to face at least one eye region of a target user, the sensing unit being configured to acquire a plurality of target signals representing a change in at least one feature of the at least one eye region of the target user.

[0009] The sensing unit includes at least one sensor, and

[0010] The at least one sensor is oriented toward the eye region; and

[0011] - The controller, which is configured as follows:

[0012] -Provide machine learning algorithms,

[0013] - Provide multiple objective target data related to these acquired target signals as input to the machine learning algorithm.

[0014] - Determine the behavior of the target user as the output of the machine learning algorithm.

[0015] The device uses machine learning algorithms to correlate objective target data related to the acquired target signal with behavior. Machine learning algorithms are capable of processing highly complex signals and then associating these complex signals with certain behaviors. Therefore, an objective determination of the behavior can be obtained.

[0016] In doing so, behavior can be determined without any assumptions about the state the eye region should be in. This allows for the avoidance of using very rigorous and demanding analytical signal processing of such feature variations (e.g., subtle and varied changes in the muscles around the eyes). In particular, this avoids the use of subjective target data from the target user to determine behavior, which is a significant benefit compared to known methods. In other words, behavior determination can be automated, i.e., without any actions or data from the target user after the method is initiated.

[0017] Furthermore, because some data is too complex or too much to be detected using typical analysis methods, the acquired target signals are utilized more fully.

[0018] At least one sensor may be a non-imaging sensor.

[0019] Another advantage is that behavior determination using machine learning algorithms does not require complex signals in the form of images or videos to determine the target user's behavior. These complex signals require significant power, and providing large data streams demands substantial computational power. Therefore, complex behaviors can be determined using basic signals acquired through a sensing unit. The sensor in the sensing unit can be a simple pixel (such as a non-imaging sensor, radiometer, or unique photodiode). Thus, imaging systems or complex signal systems can be obtained using non-imaging sensors. Using basic signals allows these drawbacks to be avoided.

[0020] Non-imaging sensors refer to sensor types that provide a single pixel as output. This type of sensor includes, but is not limited to, radiometers and photodiodes.

[0021] Furthermore, the use of non-imaging sensors allows for systems that are very lightweight in terms of power supply and have very small hardware.

[0022] Non-imaging sensors can be completely different from imaging sensors, which provide an output consisting of multiple pixels that form an image or multiple images in the form of a video.

[0023] At least one of the sensors can be sensitive to infrared, visible, and UV light. In other words, the sensor can acquire signals from infrared, visible, and UV light. The advantage of using an infrared-sensitive sensor is that it does not interfere with the target user and improves the signal-to-noise ratio. Each sensor can be sensitive to the same range of light (e.g., infrared light). Alternatively, each sensor can be sensitive to different ranges of light (e.g., one light in the visible light range and another in the infrared light range).

[0024] This device can be configured to be worn by a user. In particular, the device is preferably configured to be positioned and supported on the user's head so as to face at least one eye area of ​​the target user.

[0025] The device can be used to determine parameters that indicate eye fatigue.

[0026] The device can be used to determine the transmission of electrochromic lenses, particularly electrochromic lenses.

[0027] The device can be used in a comprehensive refractive examination instrument or to identify electrochromic lenses.

[0028] Depending on the embodiment of the device, which can be considered individually or in combination, the sensing unit includes at least two non-imaging sensors.

[0029] According to embodiments of the device that can be considered individually or in combination, at least one non-imaging sensor is associated with at least one light source.

[0030] Depending on the embodiment of the device, which can be considered individually or in combination, the sensing unit includes at least one light source.

[0031] "Integrated" means that the sensor and the light source are synchronized in frequency / time, and / or the light source is oriented toward the eye area.

[0032] The advantage of this embodiment is that it reduces the percentage of signal obscured by noise.

[0033] The at least one light source is configured to emit a light signal toward at least one of the user's eyes. In other words, the light source acts as a transmitter of the light signal. The light signal is reflected by the eye region and then received by the at least two sensors for processing. Thus, the sensors act as receivers of the light signal. The device may include a light source associated with multiple sensors. Alternatively, the light source may be associated with only one sensor to form a transmitter / receiver pair. In the latter case, the light source and sensor may be housed together in the same housing to form a sensing element configured to emit and receive light signals.

[0034] Comparing the emitted and received light signals allows the controller to determine changes in at least one of the signals. In doing so, changes in the physical characteristics of the eye region can be determined. For example, when a light signal is emitted toward the eyelid and the eyelid moves, the light signal may no longer be reflected by the skin of the eyelid, but rather by the eye itself. The reflection of the light signal varies depending on the surface from which it is reflected. Therefore, it can be determined when the position of the eyelid has changed.

[0035] The at least one light source can be configured to emit visible light, infrared light, or UV light. Preferably, the at least one light source is configured to emit invisible light signals (e.g., infrared light signals).

[0036] Preferably, the at least one light source and the at least one sensor are configured to remotely emit and receive light signals, respectively. In other words, the sensing unit can be configured to remotely acquire multiple target signals, which represent changes in at least one feature of at least one eye region of the target user.

[0037] According to an embodiment of the device, at least one light source is oriented towards the eye area.

[0038] According to an embodiment of the device, at least one light source may be placed around the eye area in order to avoid disturbing the target user.

[0039] The sensing unit is designed to be positioned in front of the target user's eye area, with the sensor and light source oriented towards the eye area. The sensor and light source are positioned both in front of and around the eye area. In other words, the sensor and / or light source can be positioned away from the target user's face, above and below the eye area.

[0040] According to an embodiment of the device, the device further includes at least one light stimulation source for stimulating at least one eye.

[0041] According to embodiments of the device, the light stimulation source can be a light source.

[0042] The present invention also provides a computer-implemented method for determining the behavior of a target user, the method comprising the following steps:

[0043] -Provide machine learning algorithms,

[0044] - Acquire multiple target signals, which represent changes in at least one feature of at least one eye region of a target user.

[0045] - Provide multiple objective target data related to these acquired target signals as input to the machine learning algorithm.

[0046] - Determine the behavior of the target user as the output of the machine learning algorithm.

[0047] The computer-implemented method allows for the enjoyment of the same advantages and technical effects as those described above for the device.

[0048] According to an embodiment of the method, the machine learning algorithm is based on multiple initial data related to a set of initial users, the initial data including multiple acquired learning signals representing changes in at least one feature for at least one eye region for each initial user in the set.

[0049] According to an embodiment of the method, the plurality of initial data associated with a group of initial users includes subjective data and objective data, the subjective data including the perceptions of the initial users in the group of behaviors caused by the changes in at least one feature for at least one eye region of each initial user in the group.

[0050] According to an embodiment of the method, the method further includes:

[0051] - Provide the machine learning algorithm with the aforementioned initial data related to an initial set of users.

[0052] - The machine learning algorithm is trained using the multiple initial data sets.

[0053] According to an embodiment of the method, the method further includes:

[0054] - Determine subjective data related to the target user, including the target user's perception of the behavior.

[0055] - Provide the subjective data related to the target user as input to the machine learning algorithm.

[0056] According to an embodiment of the method, the behavior is a change in the visual comfort of the target user, and the change in the at least one feature is caused by light stimulation provided to the at least one eye region.

[0057] According to an embodiment of the method, the change in the user's visual comfort is glare.

[0058] According to an embodiment of the method, the method further includes determining a plurality of glare categories to classify the light sensitivity of an initial user, the behavior determination step including determining, among the plurality of glare categories, the glare category corresponding to the behavior of the target user.

[0059] According to an embodiment of the method, the behavior is the movement of at least one eye of the target user, or the size of at least one pupil of the target user, or muscle strain around the eye, or movement of the eyelid.

[0060] According to an embodiment of the method, the at least one feature includes at least one of the following: the position of at least one eyelid, the position of the pupil, the size of the pupil, and muscle contraction in the at least one eye region.

[0061] According to an embodiment of the method, the method further includes determining at least one filter for the transparent support, the at least one filter being capable of improving or maintaining the visual comfort and / or visual function of the target user based on the behavior.

[0062] According to an embodiment, the determination method is a computer-implemented method. Attached Figure Description

[0063] To gain a more complete understanding of the descriptions and advantages provided herein, please now refer to the following brief description in conjunction with the accompanying drawings and detailed description, wherein the same reference numerals denote the same parts.

[0064] Figure 1 A schematic perspective view of one side of the binoculars electro-optical device is shown.

[0065] Figure 2 schematically shownFigure 1 A three-dimensional view of the other side of the binoculars.

[0066] Figure 3 schematically shown Figure 1 A front view of a binocular electro-optical device.

[0067] Figure 4 The illustration shows the device worn by the user. Figure 1 A side view of a binocular electro-optical device.

[0068] Figure 5 The diagram schematically illustrates a first infrared signal emitted from an infrared sensor toward the user's eye region and a second infrared signal reflected from the eye region toward the infrared sensor.

[0069] Figure 6 The diagram illustrates a scale for quantifying the level of discomfort a user experiences when exposed to light stimulation.

[0070] Figure 7 schematically shown Figure 1 A front view of a binoculars electro-optical device, in which the sensing elements are numbered.

[0071] Figure 8 It shows the representation Figure 7 A graph of the signal data acquired by the numbered sensing element relative to the frame axis. Detailed Implementation

[0072] In the following description, the drawings are not necessarily drawn to scale, and some features may be shown in a generalized or schematic form for clarity and brevity or for informational purposes. Furthermore, although various embodiments of manufacture and use are discussed in detail below, it should be understood that many inventive concepts that can be practiced in a variety of contexts are provided herein. The embodiments discussed herein are merely representative and do not limit the scope of the invention. It will also be apparent to those skilled in the art that all technical features defined relative to the method can be transposed individually or in combination to the apparatus, and conversely, all technical features relative to the apparatus can be transposed individually or in combination to the method.

[0073] The terms “comprise” (and any of its grammatical variations, such as “comprises” and “comprising”), “have” (and any of its grammatical variations, such as “has” and “having”), “contain” (and any of its grammatical variations, such as “contains” and “containing”), and “include” (and any of its grammatical variations, such as “includes” and “including”) are open-ended connecting verbs. They are used to indicate the presence of a feature, integer, step, or component or group thereof, but do not exclude the presence or inclusion of one or more other features, integers, steps, or components or groups thereof. Therefore, a method or a step in a method that “comprises,” “has,” “contains,” or “includes” one or more steps or elements possesses, but is not limited to, only possessing that one or more steps or elements.

[0074] This invention provides an apparatus for determining a user's behavior. This apparatus may be an eye-worn device (e.g., a head-mounted display).

[0075] The device can be a binocular device, such that it is configured to face each of the user's eye regions during use. Alternatively, the device can be monocular and configured to face only one of the user's eye regions.

[0076] The device can be configured for wear by a user. Preferably, the device is configured to be positioned and supported on the user's head so as to face at least one eye area of ​​the user. In other words, the size and weight of the device are configured to allow the user to manipulate the device in front of their eyes using a support device. The support device can be the user's hand, allowing the user to manipulate the device as binoculars. Alternatively, the support device can be a device for securing the device to the user's head, as a strap that can wrap around the user's head or an eyeglass arm positioned on the user's ear. Alternatively, the support device can be a support leg configured to sit on a table or the ground. Furthermore, the device may include a rechargeable battery to achieve energy self-sufficiency.

[0077] A user's "behavior" refers to the physical, physiological, or psychological sensations experienced by the user. This behavior causes changes in the user's physical characteristics, particularly those of the eye area. For example, when a user experiences glare, there may be muscle activity in the eye area, as well as changes in pupil size. The device is configured to determine the user's behavior based on these physical changes in the characteristics of the eye area.

[0078] The eye area includes at least one of the following: the lower and upper eyelids, eyebrows, eyelashes, eyes, the skin around the eyes, and the muscles around the eyes.

[0079] This behavior can be a change in visual comfort (such as glare), eye movement, or a change in pupil size.

[0080] "Changes in visual comfort" refers to changes in visual comfort experienced by the user, which may take the form of visual discomfort or alterations in visual function.

[0081] Visual comfort can be linked to a user's light sensitivity. Therefore, the device can be configured to determine a user's light sensitivity threshold by monitoring the response of the user's eye region to a given lighting environment.

[0082] A user's "sensitivity to light" refers to any relatively strong and prolonged response, or change in comfort or visual function, related to a temporary or continuous luminous flux or stimulus. The quantity representing a user's eye's sensitivity to said characteristic luminous flux is the light sensitivity threshold. The light sensitivity threshold can be determined by measuring any action of the user that indicates discomfort or visual perception. The light sensitivity threshold allows for the objective determination of a user's experienced visual function and / or visual discomfort.

[0083] Determining eye movement allows for tracking eye position, which can be used in various fields (e.g., cognitive science experiments).

[0084] The device includes a sensing unit configured to face at least one eye region of a user. In other words, the sensing unit is intended to be positioned in front of the user's face. The sensing unit is also configured to acquire a plurality of target signals representing changes in at least one feature of the user's at least one eye region. The feature may include at least one of the following: the position of at least one eyelid, the position of the pupil, the size of the pupil, and muscle contraction in the at least one eye region.

[0085] The sensing unit is configured to acquire a plurality of target signals representing changes in at least one feature of at least one eye region of the user without contact with the user. "Acquiring without contact with the user" means acquiring the signal without positioning electrodes or measuring elements on the user's eye region or skin. In other words, signal acquisition is contactless between the eye region and the sensing unit. Specifically, the acquisition of the at least one signal can be performed at a distance of 1 cm or greater. In a preferred embodiment, only the housing containing the sensing unit contacts the user to position the device and support it against the user's head.

[0086] The sensor of the sensing unit can be a simple pixel or multiple pixels (preferably, a single pixel). In fact, using machine learning algorithms makes it possible to significantly simplify the device by allowing the use of simple detectors (such as pixels).

[0087] At least one sensor can be sensitive to infrared, visible, and UV light. In other words, the sensor can acquire signals from infrared, visible, and UV light. The advantage of using an infrared-sensitive sensor is that it avoids interfering with the user. Furthermore, invisible light is often used to stimulate the eyes of the target user without interfering with the determination process and to improve the signal-to-noise ratio.

[0088] Each sensor can be sensitive to the same range of light (such as infrared light). Alternatively, each sensor can be sensitive to different ranges of light (for example, one light in the visible light range and another light in the infrared light range).

[0089] For clarity, see reference now. Figures 1 to 4 The device is described using an embodiment shown. Figures 1 to 4 Each feature described in this embodiment can be considered individually as a potential feature of the invention.

[0090] Device 10 may include a housing 31 that forms an outer enclosure of device 10. Housing 31 forms a cavity 16 intended to be positioned in front of a user's face. Preferably, the side of housing 31 forming the cavity 16 may further include a contour configured to engage with the user's face to position the sensing unit in front of the user's eyes. This contour may, for example, be configured to contact the user's nose and / or forehead.

[0091] like Figure 2 and Figure 3 As shown, the device 10 may also include a sensing unit 20 having a plurality of sensing elements 21.

[0092] Each sensing element 21 includes a sensor for receiving optical signals.

[0093] Alternatively, each sensing element 21 includes a sensor for receiving optical signals, a light source for emitting optical signals, and a sensor for receiving optical signals.

[0094] Each sensing element 21 is oriented toward the user's eyes. For this purpose, the housing 31 of the device 10 is configured to position the user's eyes at a predetermined location. The sensing elements 21 are positioned with respect to this predetermined location facing the user's eye area.

[0095] like Figure 2 and Figure 3 As shown, for example, device 10 may include two sensing units, each including three sensing elements 21 for each of the user's eyes. This allows for the determination of the target user's behavior for each of the user's eyes.

[0096] Alternatively, for example, device 10 may include a sensing unit comprising six sensing elements 21. This allows, but is not specifically targeted at, determining the behavior of a target user for a single eye. The advantage is that it simplifies the operation of the controller.

[0097] Preferably, the sensing elements 21 are positioned around the eye at different viewpoints and angles to provide complementary and redundant data. This helps determine the user's behavior to find the correlation between the acquired signals and the user's actions. Having multiple sensing elements 21 instead of a single sensing element gives the device 10 more opportunities to correlate with various user postures.

[0098] like Figure 4 As shown, the sensing unit 20 is designed to be positioned in front of the target user's eye area, with the sensors and light source oriented towards the eye area. Each sensing element 21 can be oriented towards the eyelid, the eye, or a specific portion of the eye area. The sensors can also be oriented towards different portions of the user's eye area. In other words, a first sensor can be oriented towards a first portion of the user's eye area (e.g., the eye itself), while a second sensor can be oriented towards a second portion of the same eye area (e.g., the eyelid). Preferably, the sensors and light source are positioned around the eye area. In other words, the sensors and light source can be positioned above and below the eye area. In this way, the user's line of sight is clear, minimizing interference from the sensing elements 21.

[0099] Preferably, the sensor and / or light source are configured to emit and receive invisible light signals respectively to avoid interfering with the user. In doing so, the measurement is more accurate and better reflects the user's behavior. More preferably, the sensor and light source are configured to emit and receive infrared light signals respectively. The infrared sensing element 21 is simply a distance sensor used to measure the characteristics of the user's eye region. This infrared reflection measurement is very fast (from 1 kHz to 100 kHz) and allows for the detection of high-speed movements, such as eye movements, changes in pupil diameter, or eyelid blinking.

[0100] like Figure 5 As shown, a light source emits a first signal 42 toward the at least one eye region 44, and a sensor receives a second signal 46 corresponding to the first signal 42 reflected by the at least one eye region 44. Therefore, it is possible to calculate how much of the infrared radiation of the first signal 42 was reflected by an object in front of the infrared element 21. Different materials have different reflectivities, so by comparing the difference between the first signal 42 and the second signal 46, it is possible to determine which materials are positioned in front of the infrared element 21. As an example, the reflectivity of the eye 48 is different from that of the eyelid 50. Therefore, a change occurs between two consecutive second signals 46 when infrared radiation is first reflected by the eye 48 and then by the eyelid 50. The same change occurs when infrared radiation is reflected by different materials. Therefore, a change in the position of an eyelid 50 or pupil 52, and a change in the size of the pupil 52, can be determined. For example, changes in these characteristics can indicate visual discomfort for the user. Therefore, a change in at least one of these characteristics can be correlated with a change in the user's visual comfort.

[0101] The device 10 may also include at least one stimulus source for stimulating at least one eye of the user. The stimulus source is designed to cause a change in at least one characteristic of the eye region. As an example, emitting light at high brightness into the wearer's eye can cause eyelid closure, muscle contraction, and pupillary constriction. Such a stimulus source is particularly useful when the determined behavior is a change in visual comfort or function (such as glare).

[0102] According to one embodiment, which can be considered alone or in combination, the light stimulus source can be a light source. Similarly, the light stimulus source serves both as a stimulus and as a means of being captured by a sensing unit after reflection on the eye region. Preferably, the stimulus source is embedded in a cavity 16 formed by the housing 31 of the device 10. The stimulus source can be combined with a diffuser 12 located in front of the user's eyes and disposed within the cavity 16 to provide diffused light. In this case, the stimulus source emits light toward the diffuser 12. Alternatively or in combination, the stimulus source can be positioned to emit light directly toward one or both of the user's eyes. Thus, the device 10 can be configured to expose the user to uniform light or spot light, or both simultaneously.

[0103] The stimulus source preferably comprises at least one light-emitting diode (LED) capable of having a variable spectrum, such as an RGB LED (red-green-blue light-emitting diode) or an RGB-W LED (red-green-blue-white light-emitting diode). Alternatively, the stimulus source may be configured to provide a predetermined single white light spectrum, or alternatively, to provide a spectrum of all visible radiation having substantially the same intensity as a spectrum with peaks. Preferably, a constant current is used to control the at least one stimulus source to obtain a constant luminous flux emitted from the at least one stimulus source. Providing a constant luminous flux to the user allows for a reduction or avoidance of biological effects compared to stimulus sources controlled by pulse width modulation (PWM).

[0104] The light signal received by the sensor can be a light source included in the sensing element, a light source not included in the sensing element, a light stimulus source, or external light such as ambient light or room light.

[0105] According to embodiments of the device that can be considered individually or in combination, at least one sensor is associated with at least one light source.

[0106] Depending on the embodiment of the device, which can be considered individually or in combination, the sensing unit includes at least one light source.

[0107] "Integrated" means that the sensor and the light source are synchronized in frequency / time, and / or the light source is oriented toward the eye area.

[0108] The advantage of this embodiment is that it reduces the percentage of signal obscured by noise.

[0109] When the acquired signal relates to the position of at least one eyelid, the sensing unit 20 is thus able to acquire a signal representing the closed / open state of the eye. Furthermore, the position of one or both eyelids allows for the determination of the blinking frequency, blinking amplitude, blinking duration, and different blinking patterns.

[0110] When the acquired signal relates to the position of the pupil, the sensing unit 20 can acquire a signal representing the position of the eye itself. Therefore, when the acquired signal relates to the size of the pupil, the sensing unit 20 can acquire a signal representing the level of pupil dilation / contraction.

[0111] Changes in at least one or more of the following—eyelid position, pupil position, and pupil size—can indicate different behaviors. Therefore, the lighting conditions experienced when these changes occur can be correlated with the user's behavior.

[0112] The device 10 further includes a controller connected to the sensing unit 20 to receive acquired signals from the sensing unit 20. The controller may be wholly or partially embedded within the housing 31. The controller may be partially located in an external terminal. The controller may be remote.

[0113] The controller can be configured to provide machine learning algorithms. Therefore, device 10 is a machine learning-based device for determining user behavior.

[0114] Machine learning algorithms take a training set of observed data points as input to "learn" equations, a set of rules, or some other data structure. This learned structure or statistical model can then be used to generalize from the training set or to predict new data. As used herein, a "statistical model" refers to any learned data structure and / or statistical data structure that establishes or predicts a relationship between two or more data parameters (e.g., input and output). Although the invention is described below with reference to neural networks, other types of statistical models can also be employed according to the invention.

[0115] For example, each data point in the training dataset may include a set of values ​​that are associated with or predict another value in the dataset. In this invention, the machine learning algorithm is configured to associate objective data related to the acquired target signal provided as input to the machine learning algorithm with the user's behavior.

[0116] The machine learning algorithm of the controller can be based on Long Short-Term Memory (LSTM) technology or Convolutional Neural Network (CNN).

[0117] LSTM technology is a part of recurrent neural networks (RNNs). A typical RNN technique consists of a network of neural nodes organized into successive layers. Each node (neuron) in a given layer is unidirectionally connected to every node in the next layer. This structure allows previous time steps to be taken into account in the neural network, because the first layer at the previous time step t-1 is connected to the second layer at time step t. This second layer is also connected to the third layer at the next time step t+1, and so on, in the case of multiple layers. Therefore, each signal provided as input is processed in a time-series manner, thus taking into account the signal provided at the previous time step.

[0118] CNN technology uses signals as images, rather than as time. Multiple acquired signals are processed simultaneously along with all data acquired over the test duration.

[0119] Machine learning algorithms can include a guiding model that defines rules configured to guide the predictions of the machine learning algorithm. These rules can include sub-correlations between target data and various behaviors. For example, this guiding model could provide that a given change in a feature must be associated with a certain behavior. In another example, the guiding model could provide that a predetermined combination of feature changes implies a certain behavior or a set of potential behaviors. This guiding model allows for simplification of correlations derived through machine learning, and thus reduces the time spent deriving correlations and improves the accuracy of those correlations.

[0120] The controller can use a pre-trained machine learning algorithm, i.e., a neural network of which includes equations or a set of rules configured to provide the correlation between changes in the physical characteristics of the user's eye region and the user's behavior. Alternatively, the controller is configured to train the machine algorithm to determine the correlation.

[0121] Training a machine learning algorithm is preferably performed by providing the algorithm with multiple sets of initial data related to an initial set of users. "Initial users" refers to users participating in the learning process of the machine learning algorithm. In other words, initial users provide objective and / or subjective data that allows the machine learning algorithm to correlate physical changes in the eye region with different behaviors. Conversely, "target users" refers to users whose behavior can be determined based on the machine learning algorithm (i.e., whose behavior can be predicted).

[0122] The initial data includes a plurality of acquired learning signals, which represent changes in at least one feature for at least one eye region for each initial user in the group. The initial data may include subjective and / or objective data. The subjective data may include the initial user's perception of behavior resulting from the changes in at least one feature of at least one eye region.

[0123] This training process is repeated multiple times to make the algorithm more accurate. For example, training the algorithm might involve at least one hundred initial users.

[0124] Then, a process for determining the target user's behavior is performed by acquiring multiple target signals representing changes in at least one feature of at least one eye region of the target user. The target signal refers to the signal representing the target user's eye region. Similarly, the initial signal refers to the signal representing the initial user's eye region.

[0125] The target data may include subjective and / or objective data.

[0126] Objective data refers to any value relating to measurements of at least one parameter characteristic of structural state and ocular function or related structures via optical and / or photometric measurements. The selection of such representative values ​​allows for the quantification of the capacity and function of one or more ocular or related structures relative to the glare process via physical measurements.

[0127] Subjective data refers to any verbal response or user action that indicates discomfort or visual perception experienced by the user. Subjective data can include the target user's perception of behavior, particularly behavior resulting from changes in at least one feature of at least one eye region. This subjective data can be obtained by the target user operating the device 10 via a dedicated interface. Subjective data can also be obtained by the target user providing information about their feelings. This subjective data can improve the accuracy of behavior determination. Objective data refers to target signals that indicate changes in at least one feature of at least one eye region of the target user.

[0128] The controller then provides the multiple target signals to its machine learning algorithm and determines the target user's behavior as the output of the machine learning algorithm.

[0129] The following text is for reference only. Figures 6 to 8 Describe an example of behavior identification. In this example, the identified behavior is glare. To achieve this identification, a photosensitivity test is performed using a stimulus that induces glare in the target user.

[0130] In use, the device 10 is positioned on the head of the target user such that the sensing unit 20 faces at least one eye region of the target user. Different levels of light intensity are provided to the target user's eyes. At the start of the test, the light intensity is very low and then gradually increased to measure the patient's sensitivity threshold.

[0131] The preferred order of light stimulation is described below.

[0132] Continuous light emission can be provided, increasing illuminance in stages to cause the illuminance to increase from a minimum to a maximum, for example, from 25 lux to 10211 lux. For instance, light emission could begin at 25 lux, continue for 5 seconds to allow the eye to adjust to the light conditions, cancel all previous exposures before measurement, and then continue increasing the illuminance at 20% per second to the maximum illuminance. More generally, light can be emitted to cause a change in illuminance from 25 lux to 10000 lux. This continuous light emission can be achieved using warm light. Another continuous light emission can be achieved using cool light.

[0133] Then, a flash emission is performed, gradually increasing the illuminance to cause it to increase from a minimum to a maximum value, for example, from 25 lux to 8509 lux. The illuminance of the flash emission is preferably increased by at least 30%, preferably by 40%, and most preferably by at least 44%. Before and between each flash emission, the user is subjected to light emission below the minimum illuminance of the flash emission (e.g., 10 lux). The duration of each flash emission is preferably 0.5 seconds, and the time between each flash emission is preferably 2 seconds.

[0134] During this light stimulation, the sensing unit 20 acquires target signals representing changes in characteristics of the target user's eye region. These target signals can be acquired continuously or intermittently and transmitted to the controller.

[0135] Target data is determined based on these target signals. This target data is then fed into the controller's machine learning algorithm to determine potential glare for the target user. Specifically, the controller is able to determine the light conditions causing the glare when the target user experiences it. The target user's light sensitivity can then be determined using an automated process that does not necessarily involve subjective data from the target user.

[0136] refer to Figure 6 Subjective data related to the target user can be obtained. The target user can be asked to press a switch on device 10 to indicate their perception of light stimulation. For example, the target user can perceive discomfort caused by light as "mildly perceptible" (see [reference needed]). Figure 6 The switch is pressed once when the user experiences discomfort or "excessive interference," and a second time when the user presses the switch. The light stimulation is preferably turned off after the target user presses the switch a second time.

[0137] The target signal recorded during the light sensitivity test is complex. Each sensing element 21 acquires approximately 200,000 data points per second. Figure 8 An example of the target signal acquired by sensing element 21 is shown. For example... Figure 7 As shown, each sensing element 21 is numbered 1 to 6.

[0138] refer to Figure 8 The data displayed is approximately 40 seconds long. In reality, sensing element 21 acquires data at approximately 200Hz, therefore 8500 frames correspond to approximately 40 seconds of measurement. The light intensity increases from frame 0 to frame 8000. First reference 30 corresponds to the moment the target user first presses the switch, while second reference 32 corresponds to the moment the target user presses the switch a second time. The same type of data is used to train the machine learning algorithm.

[0139] Based on these results, target users are categorized into multiple glare categories. These glare categories allow for the classification of users' light sensitivity. Therefore, these glare categories can be used, for example, to determine filters for target users, particularly filters for transparent supports, that can improve or maintain the visual comfort and / or visual function of the target users.

[0140] Glare classification for target users can be done through either indirect or direct classification.

[0141] In indirect classification, the machine learning algorithm is trained to detect when the initial user presses the switch. The algorithm then determines the glare category by reducing the time interval between the predicted and actual clicks on the switch. In this way, the algorithm outputs the number of frames in which it predicts the initial user will press the switch. With this number of frames, the controller is able to determine the illuminance of the light stimulus at the predicted time and then classify the target user accordingly.

[0142] In direct classification, the algorithm is trained to directly determine the glare category for the target user.

[0143] One possible classification is to divide people into three groups: highly sensitive, sensitive, and insensitive. The first group typically has a light sensitivity threshold below 1000 lux and corresponds to approximately 25% of the entire population. Next, the second group typically has a light sensitivity threshold between 1000 and 5000 lux and corresponds to approximately 50% of the entire population. Finally, the third group typically has a light sensitivity threshold above 5000 lux and corresponds to approximately 25% of the entire population.

[0144] Then, a second classification was determined through a first trial involving 500 initial users. Based on this second classification, people could be divided into four categories: Category 1, whose photosensitivity threshold was below 600 lux (approximately 25% of the entire group); Category 2, whose photosensitivity threshold was between 600 lux and 2000 lux (approximately 25% of the entire group); Category 3, whose photosensitivity threshold was between 2000 lux and 4000 lux (approximately 25% of the entire group); and Category 4, whose photosensitivity threshold was above 4000 lux (approximately 25% of the entire group).

[0145] Then, a second trial involving 200 initial users was conducted to identify the relevant glare category for each initial user in the second trial based on the first and second classifications. The results of this second trial are shown in the table below.

[0146]

[0147] The baselines are given in the first three rows, labeled “Random,” “Mean,” and “Main Category.” These percentages indicate reference values ​​for the accuracy of the glare category for the initial user. These baselines correspond to either the prediction for the random category or the classification for the mean category, which were not obtained using machine learning algorithms.

[0148] Values ​​obtained through recurrent neural network techniques and fusion layer techniques belong to indirect classification. Values ​​obtained through convolutional network techniques belong to direct classification.

[0149] The values ​​obtained through machine learning algorithms using recurrent neural network techniques, fusion layer techniques, and convolutional network techniques are compared with a baseline. It can be seen that the results obtained through machine learning algorithms are more accurate than the baseline. In particular, the fusion layer technique allows for even more accurate results.

[0150] The method according to the invention, executed by the controller of device 10, is implemented by a computer. That is, the computer program product includes one or more sequences of instructions accessible to a processor, and when executed by the processor, causes the processor to perform steps for determining the spectral transmittance of the ocular media of at least one eye of the user and for determining at least one filter as described above.

[0151] (Multiple) instruction sequences may be stored in one or more computer-readable storage media (including predetermined locations in the cloud).

[0152] Although representative methods and apparatuses have been described in detail herein, those skilled in the art will recognize that various substitutions and modifications may be made without departing from the scope described and defined by the appended claims.

Claims

1. An apparatus (10) for determining the behavior of a target user, comprising: - A sensing unit (20) configured to face at least one eye region (44) of a target user, the sensing unit (20) being configured to acquire a plurality of target signals representing a change in at least one feature of the at least one eye region (44) of the target user. The sensing unit includes at least one sensor, which is a non-imaging sensor. The at least one sensor is oriented toward the at least one eye region (44); At least one light source configured to stimulate at least one eye, the behavior being a change in the visual comfort of the target user, the change in the at least one feature being caused by light stimulation provided to the at least one eye region (44), and the change in the user's visual comfort being glare; as well as - Controller, the controller is configured to: -Provide machine learning algorithms, - Stimulate the at least one eye using the at least one light source. - Provide multiple objective target data related to the acquired target signal as input to the machine learning algorithm. - Determine the behavior of the target user as the output of the machine learning algorithm.

2. The apparatus (10) according to claim 1, wherein, The at least one sensor is placed around the eye region (44).

3. The apparatus (10) according to claim 1 or 2, wherein, The at least one sensor is associated with at least one light source.

4. A computer-implemented method for determining the behavior of a target user, the method comprising the following steps: -Provide machine learning algorithms, - Stimulate at least one eye with at least one light stimulus. - Multiple target signals are acquired by a sensing unit (20), the multiple target signals representing a change in at least one feature of at least one eye region (44) of a target user, the sensing unit being oriented toward at least one eye region (44) of the target user, the sensing unit including at least one sensor, the at least one sensor being a non-imaging sensor, the behavior being a change in the visual comfort of the target user, the change in the at least one feature being caused by light stimulation provided to the at least one eye region (44), and the change in the user's visual comfort being glare. - Provide multiple objective target data related to the acquired target signal as input to the machine learning algorithm. - Determine the behavior of the target user as the output of the machine learning algorithm.

5. The computer-implemented method according to claim 4, wherein, The machine learning algorithm is based on multiple initial data related to a set of initial users, the initial data including multiple acquired learning signals representing changes in at least one feature of at least one eye region (44) for each initial user in the set.

6. The computer-implemented method according to claim 5, wherein, The plurality of initial data associated with a group of initial users includes subjective data and objective data, the subjective data including the perceptions of the initial users in the group of behavior resulting from changes in at least one feature of at least one eye region (44) for each initial user in the group.

7. The computer-implemented method according to claim 5, further comprising: - Provide the machine learning algorithm with the plurality of initial data related to an initial set of users. - The machine learning algorithm is trained using the multiple initial data sets.

8. The computer-implemented method according to any one of claims 4 to 7, further comprising: - Determine subjective data related to the target user, including the target user's perception of the behavior. - Provide the subjective data related to the target user as input to the machine learning algorithm.

9. The computer-implemented method according to claim 4, further comprising: The step of determining multiple glare categories to classify an initial user based on light sensitivity, and determining the behavior of the target user, includes identifying a glare category among the multiple glare categories that corresponds to the behavior of the target user.

10. The computer-implemented method according to any one of claims 4 to 7, wherein, The behavior also includes the movement of at least one eye of the target user or the size of at least one pupil of the target user.

11. The method according to any one of claims 4 to 7, wherein, The at least one feature includes at least one of the following: the position of at least one eyelid, the position of the pupil, the size of the pupil, and muscle contraction in the at least one eye region (44).

12. The method according to any one of claims 4 to 7, further comprising determining at least one filter for the transparent support, said at least one filter being capable of improving the visual comfort and / or visual function of the target user based on said behavior, or maintaining the visual comfort and / or visual function of the target user based on said behavior.

Images