Portable system and computer-implemented method for assisting a person with visual impairment

Description

[0001] The invention relates to a portable system and a computer-implemented method for supporting a person with visual impairments.

[0002] The ability to perceive the world visually is very important for people. Consequently, it is very difficult for people with a visual impairment or visual impairment to fully participate in everyday life.

[0003] US 2008 / 170118 A1 describes a wearable system for assisting a person with visual impairments. The system includes an optical sensor, a processing unit, and an actuator. It generates a 3D model of the environment, determines motion trajectories, and, for example, predicts a collision between the person and a moving object.

[0004] The invention is based on the objective of offering a system and a method that supports people with visual impairments in perceiving their environment.

[0005] This problem is solved by a system having the features of claim 1 or by a method having the features of claim 15. Advantageous embodiments are described in the dependent claims.

[0006] According to the invention, the wearable system for supporting a person with visual impairments comprises at least one optical sensor, at least one actuator, and a processing unit that is functionally connected to the sensor and the actuator. The optical sensor is worn on the person's head and is configured to optically detect the person's surroundings. The actuator is configured to provide tactile and / or acoustic feedback to the person in response to control signals. The processing unit is configured to generate a three-dimensional model of the surroundings from the detected environment, to distinguish between moving objects, stationary objects, and scene elements within the model, to classify the moving and stationary objects, and to determine the motion trajectories of the moving objects relative to the stationary objects in the model.to determine the position of the person and a movement trajectory of the person relative to the stationary objects of the environment model, based on a relative movement between the person on the one hand and the moving objects, the stationary objects and scene components on the other, to predict interactions, to assign a priority attribute to the objects, whereby either the object with which a temporally nearest interaction was predicted is assigned the highest priority attribute or the object that is in the person's line of sight is assigned a higher priority attribute than an object that is outside the line of sight, and to generate the control signals for the actuator based on the predicted interactions for the object(s) with the highest priority attribute(s).

[0007] The system generates tactile and / or auditory feedback derived from the person's environment, which can be perceived by the person with visual impairments. This improves the orientation, information, and interaction possibilities of the person with visual impairments within their surroundings.

[0008] The system is designed as a wearable system. This means that the optical sensor, actuator, and processing unit are designed to be worn by a person with visual impairments. It is therefore a mobile system. Advantageously, the processing unit can be worn, for example, on a belt, using a chest strap, or on a backpack. The sensor can be integrated into a head covering, headband, or headset. The actuator can also be integrated into a head covering, headband, headset, or even a wristband. To ensure the system's portability, the system or individual system components are preferably equipped with one or more independent power sources (e.g., rechargeable batteries).

[0009] The at least one optical sensor, worn on the head, serves to optically capture the surroundings of the person with visual impairments. This sensor captures both two-dimensional (2D) image information and three-dimensional (3D) geometry of the environment. The optical sensor can, for example, be designed as a 3D camera system. Continuous environmental capture occurs, for instance, at a frame rate of 25 frames per second. However, it is also possible to use a LiDAR camera system or similar technology in combination with 2D image sensors.

[0010] Furthermore, the optical sensor may be a previously known 2D image sensor. By moving the head, the sensor can capture the environment from different perspectives. The 3D information can then be determined, for example, by correlating the image data and triangulation. If several 2D image sensors with an overlapping image area and different positions are used, the 3D information can also be determined, for example, by correlating the image data and triangulation. In all sensor configurations, it is particularly advantageous to combine the 3D information from several individual measurements, thereby increasing the accuracy of the model.

[0011] The environment captured three-dimensionally by the optical sensor is the immediate surroundings (near field) of the person with visual impairments, specifically the immediate vicinity up to a distance of several meters. The optical sensor is designed to detect a room in which the person is located and the objects within that room. It is also designed to capture the surroundings in three dimensions outdoors, within a radius of approximately 20 meters around the person with visual impairments. The 2D capture can also cover a wider area around the person.

[0012] The at least one actuator, in response to corresponding control signals transmitted to it by the processing unit, provides tactile and / or acoustic feedback to the person with visual impairments. Tactile feedback, in this context, refers to any feedback or stimulus that can be perceived by the person with visual impairments through their skin. For example, this stimulus generated by the actuator could be a pressure stimulus. However, it is also possible for the actuator to generate an electrical stimulus.

[0013] The processing unit is functionally connected to at least one sensor and at least one actuator. "Functionally connected" here refers to any connection that enables communication or signal and / or data exchange between the processing unit and the sensor or actuator. This connection can be wired or wireless (e.g., using Bluetooth, LTE, or 5G).

[0014] The computing unit is designed as an electronic computing unit, for example as a microcomputer or microcontroller with appropriate input and output interfaces as well as internal processor and memory units, so that the computing unit can receive and process the data provided by the optical sensor and other input devices and output the resulting data to the actuator and other output devices. Through the presence of appropriate programming (e.g., software modules), the computing unit is capable (i.e., designed) to perform the functionalities described below.

[0015] The processing unit is designed in such a way that it first generates a three-dimensional environmental model from the captured environment. This could, for example, be a voxel-based environmental model.

[0016] Furthermore, the processing unit is designed to distinguish between moving objects, stationary objects, and scene elements within the environment model. Moving objects include both objects that are currently moving and objects that are capable of moving by their very nature. Stationary objects are defined as objects that are neither currently moving nor capable of moving by their very nature. Scene elements can include, for example, the background of the environment or other components of the surroundings, such as—depending on the scene—the ground, walls, stairs, green spaces, sidewalk, etc.

[0017] Furthermore, the processing unit is designed to classify at least movable and immovable objects, i.e., to assign them to specific classes. The classes used for this purpose can vary depending on the environment. Typical classes would be, for example: automobile, bicycle, traffic sign, table, chair, door, drinking glass, or unknown. An object is assigned to the class "unknown" if no classification within the system is possible using known classes. A confidence level can also be assigned to the classification. This confidence level can be used to quantitatively represent how reliably objects have been classified.

[0018] The computing unit is further equipped to determine the motion trajectories of the moving objects relative to the stationary objects or scene components of the environment model, i.e., the computing unit determines the direction and speed of movement of the individual moving objects.

[0019] Furthermore, the computing unit is designed to determine the position of the person with visual impairments within the environment or within the environment model and to determine a movement trajectory of the person relative to the stationary objects and scene components of the environment model.

[0020] In addition, the computing unit is designed to predict interactions based on a relative movement between the person with visual impairments on the one hand and the moving objects, the stationary objects and scene elements on the other.

[0021] Interactions in this context include, for example, collisions between a person with visual impairments and moving or stationary objects. Furthermore, interactions can also include, for example, a hazard in the environment (e.g., a step) that the person with visual impairments encounters. Another example of an interaction would be the person with visual impairments reaching for an object in their environment.

[0022] The prediction period for which the interactions are predicted depends on the existing environment, but is generally between 0.5 seconds and a few minutes, especially between 1 second and 1 minute.

[0023] Finally, the processing unit is designed to generate control signals for the actuator based on the predicted interactions. In other words, depending on the predicted interaction, the processing unit generates a control signal for the actuator, so that the actuator, in response to these control signals, provides tactile and / or auditory feedback to the person with visual impairments. Depending on the interaction, the person thus receives different tactile and / or auditory feedback.

[0024] In this way, the system makes it possible to support the person with visual impairments through situation-adapted tactile and / or acoustic feedback.

[0025] The processing unit is further configured to assign a priority attribute to objects and to generate control signals for the actuator based on the predicted interactions for the object(s) with the highest priority attribute(s). In other words, the processing unit prioritizes the objects in the environment and generates the actuator control signals for the object(s) deemed most important. Either the object with which the next interaction was predicted is assigned the highest priority attribute, or the priority attribute is determined based on the person's line of sight. Specifically, objects located within the person's line of sight are assigned a higher priority than objects located outside that line of sight.

[0026] In a further preferred embodiment, the processing unit can also be configured to assign a scene attribute to the environment model and additionally generate the control signals for the actuator based on this scene attribute. For example, the processing unit can determine whether the current environment is an indoor space or an outdoor situation. Accordingly, the environment model would be assigned either "Indoor" or "Outdoor" as a scene attribute. The control signals generated by the actuator then differ depending on the scene attribute. In this way, feedback to the person can be tailored to the specific situation. The scene attributes can also be further refined and used for more specific assistance functions of the system. For example, there could be additional scene attributes for shops, train stations, or similar environments.

[0027] In addition to at least one optical sensor, the system can include an acoustic sensor that acoustically detects the environment of the person with visual impairments. In this case, the processing unit also considers the acoustic environment detected by the acoustic sensor when generating the environmental model. This acoustic environment can also be taken into account when generating the priority attribute and / or the scene attribute. This allows for even more precise and reliable modeling of the environment, the priority attribute, and / or the scene attribute. Consequently, the control signals generated for the actuator, and thus the feedback to the person with visual impairments, are also more accurate.

[0028] The tactile actuator can be designed as a forehead display, for example, integrated into a headband worn by a person with visual impairments. This forehead display is designed such that the display facing the person's forehead generates stimuli at specific grid points, thus providing information (feedback) to the person with visual impairments. The tactile actuator can also be designed as a forearm display, for example, integrated into a wristband worn by a person with visual impairments. This forearm display can also be designed such that the display facing the arm (especially the forearm) generates stimuli at specific grid points, thus providing information to the person with visual impairments.

[0029] The actuator can also be a directional actuator, which, in response to control signals from the processing unit, moves a movable reaction element in a predetermined direction. This movement, in turn, can be perceived by a person with visual impairments, for example, as a pressure stimulus if they are wearing a wristband with such a directional actuator integrated into it, or if they otherwise touch the directional actuator. Designing the actuator as a directional actuator is particularly advantageous when the predicted interaction involves a grasping movement by the person. Appropriate feedback from the directional actuator to the person allows the grasping movement (e.g., the position of the person's hand) to be influenced, corrected, and supported.

[0030] Furthermore, the actuator can be designed in such a way that information about the environment of the person with visual impairments is transmitted.

[0031] In a preferred embodiment, the computing unit can be configured to perform object discrimination, object classification, interaction prediction and / or the generation of control signals for the actuator using one or more machine learning methods, in particular using so-called neural networks.

[0032] In another preferred embodiment, the processing unit can be configured to generate identical control signals for objects classified in the same object class. For example, the same symbol (e.g., a stylized silhouette of a car or a dot-and-dash pattern) can be displayed by the actuator for each object assigned to the class "automobile". In this way, a person with a visual impairment can quickly identify the object without significant cognitive effort.

[0033] The complexity and resolution of the symbols that can be transmitted to a person with visual impairments are generally limited to a relatively small number of grid points that can be simultaneously perceived by the person and generated by the actuator. Accordingly, either very simplified and abstract pictorial symbols are displayed, or a dot-and-dash pattern is presented, the meaning of which the person with visual impairments may then have to learn.

[0034] Both pictorial symbols and dot-dash patterns can be displayed as sequences (i.e., changing over time). In particular, the symbol or dot-dash pattern can be displayed in a treadmill-like fashion, pulsating, and / or with changing size. In this way, the information content can be increased without requiring a higher resolution of raster dots.

[0035] InIn another preferred embodiment, the system further comprises a remote access unit. This remote access unit is configured to access the environment model and, if necessary, adapt or modify it. Furthermore, the remote access unit is configured to communicate with the visually impaired person and / or generate control signals for the actuator. The remote access unit is preferably connected to the processing unit via a wireless communication link.

[0036] InIn another preferred embodiment, the computing unit can further be configured to determine a path to reach the target position based on the desired target position of the person with visual impairments and the environmental model, and to generate control signals for the actuator based on this path. In this way, the system can serve as a navigation aid and thus provide even greater support to the person with visual impairments.

[0037] The computer-implemented method according to the invention for supporting a person with visual impairments comprises the following process steps: (1) capturing the person's environment using an optical sensor; (2) generating a three-dimensional environment model; (3) distinguishing between moving objects, stationary objects, and scene elements in the environment model; (4) classifying the moving objects and the stationary objects; (5) determining the motion trajectories of the moving objects relative to the stationary objects of the environment model; (6) determining the position and motion trajectory of the person relative to the stationary objects of the environment model; (7) predicting interactions based on relative movement between the person on the one hand and the moving objects, the stationary objects, and scene elements on the other.(8) Assigning a priority attribute to the objects, whereby either the object with which a temporally nearest interaction has been predicted is assigned the highest priority attribute, or the object located in a line of sight of the person (P) is assigned a higher priority attribute than an object located outside the line of sight; and (9) generating control signals for a tactile and / or acoustic actuator, based on the predicted interactions for the object(s) with the highest priority attribute(s).

[0038] The invention is further explained with reference to an exemplary embodiment in the drawing figures, where identical reference numerals denote identical or equivalently acting components. The figures show: Fig. 1 a schematic representation of a portable system to support a person with visual impairments; Fig. 2 a schematic block diagram of the system made of Fig. 1 ; and Fig. 3 a flowchart of partial steps of a procedure to support a person with visual impairments, which is performed by the system Fig. 1 is executed.

[0039] Fig. 1 Figure 1 schematically shows a portable system 1 for supporting a person P with visual impairments. System 1 consists of the main components: computing unit 10, operating controller 20, headband 30, wristband 40, and - in Fig. 1 Not shown - Remote access unit 50.

[0040] System 1 will now be referred to Fig. 1 and Fig. 2 described in more detail.

[0041] The computing unit 10 is designed as a portable microcomputer and can be worn by person P, for example, on a belt, carrying strap, or in a backpack. The computing unit 10 is equipped with software modules that function as an image processor 101, a sound and speech processor 102, an environment model generator 103, an object identifier 104, an object classifier 105, a priority attribute generator 106, a scene attribute generator 107, a trajectory determiner 108, a predictor 109, and a control signal generator 110. By executing these software modules, the computing unit 10 can perform the respective functionalities described in more detail below. The software modules can be designed as individual, separate modules. However, it is also possible to integrate several or all modules into a single, shared module.For example, the object identifier 104 and the object classifier 105 can be combined in a common object identification and object classification module.

[0042] The operating controller 20 is functionally connected to the computing unit 10 and serves to control the computing unit 10 by person P or another person. For this purpose, the operating controller 20 has one or more human-machine interfaces.

[0043] Headband 30 is worn on the head by person P and functions as a headset. InIntegrated into the headband 30 are a 3D camera system 301, a microphone 302, a haptic forehead display 303, a loudspeaker 304, and a headset accelerometer 305. These components of the headband 30 are functionally connected to the processing unit 10. The 3D camera system optically detects the surroundings of person P and constitutes an "optical sensor" within the meaning of the present invention. The microphone 302 acoustically detects the surroundings of person P and constitutes an "acoustic sensor" within the meaning of the present invention. The haptic forehead display 303 is designed such that, in response to control signals from the processing unit 10, it generates tactile stimuli at specific grid points on a display facing the forehead of person P. The haptic forehead display 303 thus constitutes an "actuator configured to provide tactile feedback to the person in response to control signals."The loudspeaker 304 is integrated into an earpiece and is designed to generate acoustic signals in response to control signals from the processing unit 10. The loudspeaker 304 thus constitutes an "actuator designed to provide acoustic feedback to the person in response to control signals." The data acquired by the headset accelerometer 305, together with the data from the 3D camera system, are used by the processing unit 10 to determine the position of person P and the movement of person P relative to their environment.

[0044] The wristband 40 incorporates a haptic forearm display 401 and a wristband accelerometer 402. These components of the wristband 40 are functionally connected to the processing unit 10. The haptic forearm display 401 is designed to generate tactile stimuli at specific grid points on a display facing the forearm of person P in response to control signals from the processing unit 10. The haptic forearm display 401 thus represents a further "actuator designed to provide tactile feedback to the person in response to control signals." The data acquired by the wristband accelerometer 402, together with the data from the 3D camera system, are used by the processing unit 10 to determine the position of person P and the movement of person P relative to their environment and / or to determine the position of person P's arm / hand.

[0045] The computer-implemented procedure carried out by System 1 to support person P is in Fig. 3 illustrated using a flowchart.

[0046] First, the 3D camera system 301 and the microphone 302 capture the environment of person P. More precisely, the 3D camera system 301 captures the environment optically, and the microphone 302 captures it acoustically. The optical data captured by the 3D camera system 301 is processed by the image processor 101. The sound and speech processor 102, in turn, processes the acoustic data captured by the microphone 302.

[0047] The environment model generator 103 creates a three-dimensional environment model from the processed optical and acoustic data. The object identifier 104 distinguishes between moving objects, stationary objects, and scene components within this three-dimensional environment model. The object classifier 105 classifies the moving and stationary objects, assigning the detected moving and stationary objects to different object classes.

[0048] The trajectory determiner 108 determines the motion trajectories of the moving objects relative to the stationary objects of the environment model. The predictor 109 predicts interactions based on a relative motion between the person P on the one hand and the moving objects, the stationary objects, and scene components on the other. The control signal generator 110 generates control signals for the haptic forehead display 303, the loudspeaker 304, and / or the haptic forearm display 401 based on the predicted interactions.

[0049] Priority attribute generator 106 assigns a priority attribute to the objects. Scene attribute generator 107 assigns a scene attribute to the environment model.

[0050] Depending on the detected environment and the scene at hand, System 1 supports person P in different ways. The following application examples are described as illustrations: Application example 1: Orientation in road traffic

[0051] If person P is in a road traffic environment, the 3D camera system 301 optically captures this environment, and the microphone 302 acoustically captures it. From the captured data, the environment model generator 103 creates a three-dimensional environment model. The object identifier 104 identifies objects in this environment, such as moving objects like various cars, bicycles, trucks, buses, and people, or stationary objects like various traffic signs, traffic lights, sidewalks, curbs, tables, chairs, and waste containers. Based on the identified objects and sounds, and possibly on additional information (e.g., GPS data), the scene attribute generator 107 creates a scene attribute, in this case, the attribute "Outdoor".

[0052] The trajectory determiner 108 determines a motion trajectory for the moving objects, and the predictor 109 predicts interactions based on the relative motion between person P on the one hand and the moving objects, stationary objects, and scene elements on the other. The position of person P and their motion trajectory are determined based on data from the headset accelerometer 305 and / or the wristband accelerometer 402 and the 3D camera system 303.

[0053] For example, predictor 109 might predict that, assuming person P maintains a constant relative motion, a collision with a stationary object (e.g., a trash can on the sidewalk) will occur in 20 seconds, and a collision with a moving object (e.g., another person) will occur in 10 seconds. Priority attribute generator 106 then assigns a higher priority attribute to the moving object, since this interaction would occur before the interaction with the stationary object.

[0054] The control signal generator 110 now generates control signals for the haptic forehead display 303 and the haptic forearm display 401, based on the interaction or object with the highest priority attribute and taking into account the scene attribute "Outdoor". More precisely, the control signal generator 110 generates control signals that create stimuli in the form of a dot-dash pattern on the forehead display 303 and the forearm display 401, or on another tactile signal transmitter. Alternatively, with a sufficiently high resolution, a stylized human silhouette can be generated. In this way, person P receives tactile feedback that a collision with another person is predicted and can change their behavior accordingly (e.g., stop or change direction). Application example 2: Orientation in a building

[0055] If person P is located in a building or in a room within the building, the 3D camera system 301 visually captures this environment, and the microphone 302 acoustically captures it. From the captured data, the environment model generator 103 creates a three-dimensional environment model. The object identifier 104 identifies objects in this environment, such as moving objects like people or stationary objects like various tables, chairs, doors, drinking glasses, and steps. Based on the identified objects and sounds, and possibly on other information, the scene attribute generator 107 creates a scene attribute, in this case, the attribute "Indoor".

[0056] The trajectory determiner 108 defines a motion trajectory for the moving objects, and the predictor 109 predicts interactions based on the relative motion between person P on the one hand and the moving objects, the stationary objects, and scene elements on the other. For example, the predictor 109 might predict that, assuming person P maintains a constant relative motion, a collision with a closed door will occur in 10 seconds, and that, assuming the same relative motion, person P will reach a (downward) step in 5 seconds. The priority attribute generator 106 then assigns a higher priority attribute to the step, since this interaction would occur before the interaction with the closed door.

[0057] The control signal generator 110 now generates control signals for the haptic forehead display 303 and the loudspeaker 304, based on the interaction or object with the highest priority attribute and taking into account the scene attribute "Indoor". More precisely, the control signal generator 110 generates control signals that create stimuli in the form of a dot-dash pattern on the forehead display 303. Alternatively, with a sufficiently high resolution, a stylized silhouette of a downward-leading step can be generated. Furthermore, the control signal generator 110 generates control signals that cause the loudspeaker 304 to emit a warning message, such as "Caution, step down". In this way, person P receives tactile and audible feedback that a downward-leading step is being predicted and can adjust their behavior accordingly. Application example 3: Support during grasping movements

[0058] For example, if person P is standing in front of a table on which a drinking glass is placed and wants to pick up the drinking glass, the 3D camera system 301 optically captures the surroundings, and the microphone 302 acoustically captures these surroundings. From the captured data, the environment model generator 103 creates a three-dimensional environment model. The object identifier 104 identifies objects in this environment, in particular the drinking glass.

[0059] The trajectory determiner 108 determines the relative motion of person P – more precisely, the position and movement of person P's hand, derived from data from the wristband accelerometer 402 – relative to the drinking glass. This depends on the distance and spatial position between the drinking glass and the person's hand. P, Predictor 109 predicts when the drinking glass will be grabbed.

[0060] The control signal generator 110 then generates, based on the distance and spatial position between the drinking glass and the person's hand P, Control signals for the wristband 40. More precisely, the control signal generator produces 110 control signals, which generate 401 stimuli in the form of a directional indication (e.g., a stylized arrow) on the forearm display. In this way, person P receives tactile feedback indicating in which direction they need to move their hand to grasp the drinking glass.

[0061] If system 1 is equipped with a directional actuator (not shown in the figures), the control signal generator 110 can alternatively or additionally generate control signals that move the movable reaction element of the directional actuator in a specific direction. This provides person P with tactile feedback indicating the direction in which they must move their hand to grasp the drinking glass. REFERENCE MARK LIST

[0062] 1 System 10 Computing Unit 20 Control Controller 30 Headset 40 Wristband 50 Remote Access Unit 101 Image Processor 102 Sound and Speech Processor 103 Environment Model Generator 104 Object Identifier 105 Object Classifier 106 Priority Attribute Generator 107 Scene Attribute Generator 108 Trajectory Determiner 109 Predictor 110 Control Signal Generator 301 3D Camera System 302 Microphone 303 Haptic Forehead Display 304 Speaker / Earphone 305 Headset Accelerometer 401 Haptic Forearm Display 402 Wristband Accelerometer P Person

Claims

1. Portable system (1) for assisting a person (P) with visual impairments, comprising: - at least one optical sensor (301) that is worn on the head of the person (P) and designed to optically capture surroundings of the person (P); - at least one actuator (303, 304, 401) that is designed to provide tactile and / or acoustic feedback to the person (P) in response to control signals; - a computing unit (10) that is operatively connected to the sensor (301) and the actuator (303, 304, 401) and designed: ∘ to generate a three-dimensional model of the surroundings from the captured surroundings; ∘ to distinguish between moving objects, stationary objects and scene elements in the model of the surroundings; ∘ to classify the moving objects and the stationary objects; ∘ to determine the movement trajectories of the moving objects relative to the stationary objects in the model of the surroundings; ∘ to determine the position of the person (P) and a movement trajectory of the person relative to the stationary objects in the model of the surroundings; and ∘ to predict interactions on the basis of a relative movement between firstly the person (P) and secondly the moving objects, the stationary objects and scene elements; characterized in that the computing unit (10) is further designed: o to assign a priority attribute to the objects, wherein either the highest priority attribute is assigned to the object predicted to have the next upcoming interaction or the object that is within a line of sight of the person (P) is assigned a higher priority attribute than an object that is away from the line of sight; and ∘ to generate the control signals for the actuator (303, 304, 401) on the basis of the predicted interactions for the object(s) with the highest priority attribute (s) .

2. System according to any of the preceding claims, wherein the computing unit (10) is further designed: o to assign a scene attribute to the model of the surroundings; and ∘ to generate the control signals for the actuator (303, 304, 401) on the basis of the scene attribute as well.

3. System according to any of the preceding claims, further comprising an acoustic sensor (302) that is designed to acoustically record the surroundings of the person (P), and wherein the computing unit (10) is designed to generate the model of the surroundings, the priority attribute and, if referring back on Claim 2, the scene attribute from the surroundings captured by the optical sensor (301) and recorded by the acoustic sensor (302).

4. System according to any of the preceding claims, comprising at least one tactile actuator (303, 401) and at least one acoustic actuator (304), wherein the computing unit (10) is designed to generate control signals for the tactile actuator (303, 401) and / or the acoustic actuator (304) in a manner dependent on predicted interactions.

5. System according to any of the preceding claims, comprising a first tactile actuator in the form of a forehead display (303) which is integrated in a headband (30) that can be worn by the person (P) and a second tactile actuator in the form of a forearm display (401) which is integrated in a bracelet (40) that can be worn by the person and / or in the form of a direction actuator which is integrated in the wearable bracelet (40).

6. System according to Claim 5, wherein the computing unit (10) is designed to generate the control signals for the forearm display (401) and / or the direction actuator on the basis of a distance and / or a spatial position between firstly one of the objects and secondly the bracelet (40), an arm of the person (P) or a hand of the person (P).

7. System according to any of the preceding claims, wherein the computing unit (10) is designed to perform the differentiation of the objects, the classification of the objects, the prediction of the interactions and / or the generation of the control signals for the actuator (303, 304, 401) with the aid of one or more machine learning methods.

8. System according to any of the preceding claims, wherein the computing unit (10) is designed to generate the same control signals for objects that are classified in the same object class.

9. System according to any of the preceding claims, wherein the optical sensor (301) is in the form of a 3-D camera system, in particular a 3-D camera system which is integrated in a headset (30) that can be worn by the person.

10. System according to any of the preceding claims, further comprising a remote access unit (50) that is designed to access the model of the surroundings, to communicate with the person (P) and / or to generate control signals for the actuator (303, 304, 401).

11. System according to any of the preceding claims, wherein information about the surroundings of the person (P) is transmitted by way of the actuator (303, 304, 401).

12. System according to any of the preceding claims, wherein the computing unit is further designed to determine, on the basis of a target position of the person (P) and the model of the surroundings, a preset path for reaching the target position and to generate control signals for the actuator (303, 304, 401) on the basis of the preset path.

13. System according to Claim 12, wherein the target position originates from a navigation system, in particular from a pedestrian navigation system.

14. System according to any of the preceding claims, wherein the computing unit is further designed to determine a confidence for the moving objects and the stationary objects in the classification of the moving objects and the stationary objects.

15. Computer-implemented method for assisting a person (P) with visual impairments, including the following method steps: - capturing surroundings of the person (P) by means of an optical sensor (301); - generating a three-dimensional model of the surroundings; - distinguishing between moving objects, stationary objects and scene elements in the model of the surroundings; - classifying the moving objects and the stationary objects; - determining the movement trajectories of the moving objects relative to the stationary objects in the model of the surroundings; - determining the position and the movement trajectory of the person (P) relative to the stationary objects in the model of the surroundings; - predicting interactions on the basis of a relative movement between firstly the person (P) and secondly the moving objects, the stationary objects and scene elements; - characterized in that a priority attribute is assigned to the objects, wherein either the highest priority attribute is assigned to the object predicted to have the next upcoming interaction or the object that is within a line of sight of the person (P) is assigned a higher priority attribute than an object that is away from the line of sight; and - a control signal for a tactile and / or acoustic actuator (303, 304, 401) is generated on the basis of the predicted interactions for the object(s) with the highest priority attribute(s).