An immersive three-dimensional blind aid navigation and interaction method and system

Abstract

The application discloses an immersive three-dimensional blind aid navigation and interaction method and system, and the method comprises the following steps: S1, laser radar point cloud data and visual images are acquired, visual image-laser radar point cloud registration is carried out through an improved iterative closest point algorithm, and fused point cloud is generated; S2, according to the fused point cloud, an octree map modeling is carried out, and a three-dimensional environment map is generated; S3, IMU data is acquired, and the pose of a user is estimated; S4, according to the pose of the user and the three-dimensional environment map, a navigation path is planned, and the user is guided to walk along the navigation path. The application adopts the ICPS algorithm to carry out visual-lightweight laser radar point cloud registration, adopts the octree map to dynamically model the environment based on the global target point determined by the user, combines the current position of the user, and plans a navigation path based on a multi-target optimization function, so that the navigation path can safely pass through the modeled three-dimensional environment and reach the navigation target point.

Description

Technical Field

[0001] This invention belongs to the field of intelligent assistive navigation technology, specifically relating to an immersive three-dimensional assistive navigation and interaction method and system. Background Technology

[0002] Current intelligent navigation technologies for the blind still have shortcomings: rapid and accurate 3D spatial modeling of the environment in which the blind person is located is the primary prerequisite for achieving intelligent and accurate navigation for the blind. Patent CN117427278A proposes a path navigation system and method for the blind. However, in terms of 3D environment modeling, existing solutions rely on low-dimensional raster maps, which cannot reconstruct key spatial details such as step height and column width. Patent CN118298170A proposes an intelligent navigation system and method for the blind based on image semantic segmentation, but this is a purely visual solution. Due to the lack of depth estimation information for obstacles, it results in large ranging errors and cannot identify transparent obstacles. Neither of these solutions has been able to establish a refined 3D spatial structure.

[0003] Utilizing multimodal information beyond visual perception for navigation command feedback is crucial for lowering the barrier to entry for blind users of smart devices and improving the effectiveness of assistive navigation. In this regard, patent CN114533503A proposes a smart assistive glasses system and interaction method for blind users, using a voice playback module to transmit navigation commands; patent CN118196347A proposes a mixed reality-based assistive navigation device and method for the blind, using bone conduction headphones, tactile feedback devices, and thermal feedback devices to construct a navigation prompting device for transmitting navigation commands. The former is too simple, easily leading to inaccurate command execution; the latter's prompting device is too complex and costly. Furthermore, neither of these solutions includes a feedback mechanism; whether the pedestrian accurately executes the current navigation command directly affects the effectiveness of subsequent commands in navigation.

[0004] The fundamental problems with these technical solutions are: insufficient three-dimensional spatial modeling of the environment in which blind people live; lack of global and local coordination mechanism in path planning, which is prone to failure when switching between indoor and outdoor scenes; and the failure of the interactive system to form a closed loop of "perception-decision-feedback", resulting in the ineffective coordination of multi-channel information such as touch and hearing. Summary of the Invention

[0005] In view of the above-mentioned shortcomings in the prior art, the present invention provides an immersive three-dimensional assisted navigation and interaction method and system for the blind, which solves the problems that existing smart assisted navigation technologies for the blind still have, such as failing to establish a refined three-dimensional spatial structure, having overly simple navigation command interaction methods or overly complex and costly devices, and lacking feedback mechanisms.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: an immersive three-dimensional assisted navigation and interaction method for the blind, comprising the following steps:

[0007] S1. Acquire lidar point cloud data and visual images, and perform visual image-lidar point cloud registration using an improved iterative nearest point algorithm to generate a fused point cloud;

[0008] S2. Based on the fused point cloud, model the octree map to generate a 3D environment map;

[0009] S3. Acquire IMU data and estimate the user's pose;

[0010] S4. Based on the user's pose and the 3D environment map, plan the navigation path and guide the user to walk along the navigation path.

[0011] Furthermore: In S1, the method for visual image-LiDAR point cloud registration specifically involves calculating the rigid transformation matrix of the registered point cloud. ;

[0012]

[0013] In the formula, arg This represents the parameter that causes the function to reach its extreme value. This indicates a search for the function that yields its minimum value. T , Indicates the accumulation symbol. T This represents the rigid body transformation matrix that needs to be optimized. This indicates the coordinates of the feature points captured by the camera. This indicates the coordinates of the feature points acquired by the lidar. express The corresponding surface normal vector, express The corresponding surface normal vector, and Indicates material information encoding. Indicates the weighting factor. This represents the proportion of the control surface normal vector error in the overall objective function. This indicates the proportion of the control encoding material information error in the overall objective function. This represents the square of the Euclidean distance.

[0014] Furthermore: In S2, the voxel occupancy probability is updated using the following formula during the octree map modeling process;

[0015]

[0016] In the formula, This represents the logarithmic probability value of the voxel at the current moment. This represents the historical state value at the previous moment. This indicates that based on the current sensor observation data z The calculated instantaneous occupancy probability, These are the initial prior values. It is a logarithmic function;

[0017]

[0018] In the formula, Indicates navigation prior values, This represents the first frame observation value from the sensor. This represents the weighting coefficient.

[0019] Furthermore: In S3, the objective function for estimating the user's pose is specifically:

[0020]

[0021] In the formula, This represents the 16-dimensional state vector to be optimized. This represents the 6-dimensional raw measurement data from the IMU sensor. This represents the position observations of a GNSS receiver in three dimensions or higher. express The corresponding motion model, express The corresponding observation model.

[0022] Furthermore: In S4, the method for planning the navigation path based on the user's pose and the 3D environment map is as follows:

[0023] S41. Obtain the global path point sequence based on the global target points determined by the user, and generate local target points based on the user's current location;

[0024] S42. Evaluate the motion scheme of local target points through a multi-objective optimization function, and then generate a navigation path.

[0025] Furthermore: In S41, local target points The specific expression is:

[0026]

[0027] In the formula, Represents a global path point. Indicates the user's current location. Indicates constraints. Represents the L2 norm;

[0028] In S42, the multi-objective optimization function The specific expression is:

[0029]

[0030] In the formula, Indicates to local target The degree of proximity, This indicates the calculation of the safe distance to the nearest obstacle. This indicates the consistency between the direction of movement and the overall objective. v Indicates movement speed. w Indicates the steering angular velocity. This represents the weight that controls the forward efficiency. This indicates the weight used to control obstacle avoidance safety. The weight representing the correct control direction.

[0031] An immersive three-dimensional assisted navigation and interaction system for the blind includes an environmental sensing device, a central processing unit, and a multimodal interaction device, wherein the central processing unit is connected to the environmental sensing device and the multimodal interaction device respectively.

[0032] The environmental sensing device includes a binocular vision module, a lightweight lidar, and an inertial measurement unit;

[0033] The central processing unit includes a multi-sensor fusion module, a dynamic environment modeling module, a state estimation module, and an intelligent path planning module connected in sequence. The multi-sensor fusion module is also connected to a binocular vision module and a lightweight LiDAR, respectively.

[0034] The multimodal interaction device includes a haptic guide wristband and bone conduction headphones, both of which are connected to a central processing unit.

[0035] Furthermore, the system also includes glasses, with a binocular vision module comprising a left vision camera and a right vision camera, located on the sides of the glasses frame, and a lightweight LiDAR located in the center of the glasses frame.

[0036] Furthermore: the central processing unit and the inertial measurement unit are respectively located on the two temples of the glasses, and the bone conduction headphones are located on the two temples of the glasses.

[0037] Furthermore, the tactile guidance wristband is equipped with an integrated piezoelectric ceramic vibration unit.

[0038] The beneficial effects of this invention are as follows: Addressing the problems of existing intelligent assistive navigation technologies for the blind, such as failure to establish a refined three-dimensional spatial structure, overly simplistic navigation command interaction methods, or overly complex and costly devices, and the lack of feedback mechanisms, this invention provides an immersive three-dimensional assistive navigation and interaction method and system, which has the following advantages:

[0039] (1) Refined 3D environment modeling: With the help of multimodal sensors on wearable smart glasses, including binocular vision modules and lightweight lidar, a 3D spatial structure within a range of 3 to 10 meters in front of the blind person is established. For objects that occupy space, the area they occupy in space is modeled.

[0040] (2) Navigation path planning: The ICPS algorithm is used for visual-lightweight lidar point cloud registration, and IMU / GNSS tightly coupled positioning is used for pedestrian pose estimation. Based on the global target point determined by the user, the navigation software system generates a global navigation path and local target points. The innovative Dynamic-Log-Odds octree map dynamic modeling environment is adopted. Based on the user's current position and a multi-objective optimization function, a navigation path that can safely pass through the modeled 3D environment and reach the navigation target point is planned.

[0041] (3) Immersive navigation and interaction for the blind: Based on the planned local navigation path, the navigation instructions corresponding to the local navigation path are transformed into dual-mode guidance of wristband tactile vibration and bone conduction speech. The blind person is guided to move forward according to the planned path through sound and touch multimodal methods. During the process of moving forward, the planned path is adjusted according to the walking deviation. Based on the new observations of the multimodal sensor, the three-dimensional environment model is updated and the planned path is adjusted to realize immersive visual and tactile guided intelligent navigation. Attached Figure Description

[0042] Figure 1 This is a flowchart of an immersive three-dimensional assisted navigation and interaction method for the blind according to the present invention.

[0043] Figure 2 This is a schematic diagram of an immersive three-dimensional assisted navigation and interaction system for the blind according to the present invention. Figure 1 .

[0044] Figure 3 This is a schematic diagram of an immersive three-dimensional assisted navigation and interaction system for the blind according to the present invention. Figure 2 .

[0045] Figure 4 This is a schematic diagram illustrating the application scenario of the immersive three-dimensional blind-aid navigation and interaction system and method of the present invention. Detailed Implementation

[0046] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, in one embodiment of the present invention, an immersive three-dimensional assisted navigation and interaction method for the blind includes the following steps:

[0048] S1. Acquire lidar point cloud data and visual images, and perform visual image-lidar point cloud registration using an improved iterative nearest point algorithm to generate a fused point cloud;

[0049] S2. Based on the fused point cloud, model the octree map to generate a 3D environment map;

[0050] S3. Acquire IMU data and estimate the user's pose;

[0051] S4. Based on the user's pose and the 3D environment map, plan the navigation path and guide the user to walk along the navigation path.

[0052] In S1, the specific method for visual image-LiDAR point cloud registration is as follows: calculate the rigid transformation matrix of the registered point cloud. ;

[0053]

[0054] In the formula, arg This represents the parameter that causes the function to reach its extreme value. This indicates a search for the function that yields its minimum value. T , Indicates the accumulation symbol. T This represents the rigid body transformation matrix that needs optimization, used to align point clouds in different coordinate systems. This indicates the coordinates of the feature points captured by the camera. This indicates the coordinates of the feature points acquired by the lidar. express The corresponding surface normal vector, express The corresponding surface normal vector is used to describe the local geometric orientation. and This indicates encoded material information, such as visual RGB values ​​or radar reflectance. This represents a weighting factor used to distinguish objects with similar appearances. The matching confidence is dynamically adjusted based on the distance between point pairs and the similarity of their materials. This represents the proportion of the control surface normal vector error in the overall objective function. This indicates the proportion of the control encoding material information error in the overall objective function. This represents the squared Euclidean distance, used to quantify the matching error across the three dimensions of position, normal, and material. By fusing geometric and material features, the registration robustness in complex scenes (such as transparent obstacles and low-texture areas) is significantly improved.

[0055] In this embodiment, the present invention uses the ICPS algorithm for visual image-LiDAR point cloud registration. Based on the position error and normal vector error of the traditional ICP, a material consistency error term is added. By using the RGB / texture information captured by the camera and the reflection intensity information of the lightweight LiDAR, the following effects are achieved: (1) Distinguishing objects with similar appearance but different materials, such as glass doors and mirror walls; (2) Improving the registration robustness of low-texture areas.

[0056] In S2, the voxel occupancy probability is updated using the following formula during the octet map modeling process. In this embodiment, the voxel occupancy probability update adopts the innovative Dynamic-Log-Odds method: the dynamic update of the voxel occupancy status is realized through the logarithmic transformation of the probability, as shown in the following formula.

[0057]

[0058] In the formula, This represents the logarithmic probability value of a voxel at the current moment, quantifying the confidence level that the spatial cell is occupied by an obstacle. This represents the historical state value at the previous moment. This indicates that based on the current sensor observation data z The calculated instantaneous occupancy probability is obtained when the sensor is a LiDAR or a binocular vision module, such as when the LiDAR detects an obstacle. The value is close to 1, and close to 0 when no detection is performed; The prior values ​​after initialization reflect the initial uncertainties of the system; logarithmic function log The expression (.) transforms the probability ratio into a linearly additive Log-Odds form, avoiding the numerical underflow problem of probability multiplication while allowing information fusion through simple addition and subtraction. This model continuously accumulates sensor evidence; for example, multiple detections of obstacles will... Continue to increase, until when When the threshold is exceeded, the voxel is determined to be occupied, thereby constructing a robust 3D environment map.

[0059]

[0060] In the formula, Indicates navigation prior values, This represents the first frame observation value from the sensor. This represents the weighting coefficient.

[0061] In this embodiment, the weighted fusion navigation prior value The value range is 0-4, such as 3.0 for a wall and 0 for an open space; when an obstacle is detected, the sensor's first frame observation value is 2.5, and when no obstacle is detected, the sensor's first frame observation value is -1.0. The value ranges from 0.7 to 0.9, and is used to control the ratio of the two. Larger sizes rely more on navigation data, making them suitable for high-precision navigation software; smaller sizes rely more on real-time sensor data, making them suitable for dynamic environments. This design leverages the global reliability of navigation information while also enabling local corrections through sensor data, for example, when... When = 0.8, the prior value of the wall region after initialization. While preserving prior knowledge, the observation bias is slightly adjusted to effectively balance system robustness and environmental adaptability.

[0062] In S3, the objective function for estimating the user's pose is specifically:

[0063]

[0064] In the formula, This represents the 16-dimensional state vector to be optimized, which includes parameters such as position, velocity, attitude, and sensor bias. This represents the 6-dimensional raw measurement data from the IMU sensor, which includes an accelerometer and a gyroscope. This represents the position observations of a GNSS receiver in three dimensions or higher. express The corresponding motion model, express The corresponding observation model and motion model are used to convert the state vector x Mapped to predicted sensor data, this is achieved by minimizing the squared error between the predicted and actual observations. This allows for the acquisition of high-precision continuous positioning results.

[0065] In this embodiment, the present invention employs a tightly coupled fusion positioning method of inertial measurement unit (IMU) and global navigation satellite system (GNSS) to deeply fuse the absolute position information of the positioning software with the high-frequency relative motion data of the inertial measurement unit to estimate the user's pose.

[0066] In S4, the method for planning navigation paths based on the user's pose and the 3D environment map is as follows:

[0067] S41. Obtain the global path point sequence based on the global target points determined by the user, and generate local target points based on the user's current location;

[0068] S42. Evaluate the motion scheme of local target points through a multi-objective optimization function, and then generate a navigation path.

[0069] In S41, the local target point The specific expression is:

[0070]

[0071] In the formula, This represents a global waypoint, which contains latitude, longitude, elevation, and orientation information. Indicates the user's current location. Indicates constraints. Represents the L2 norm;

[0072] In S42, the multi-objective optimization function The specific expression is:

[0073]

[0074] In the formula, Indicates to local target The degree of proximity, This indicates the calculation of the safe distance to the nearest obstacle. This indicates the consistency between the direction of movement and the overall objective. v Indicates movement speed. w Indicates the steering angular velocity. This represents the weight that controls the forward efficiency. This indicates the weight used to control obstacle avoidance safety. The weight representing the correct control direction.

[0075] In this embodiment, the navigation path planning method uses a multi-objective optimization function. To evaluate each possible motion scheme, the algorithm also sets... and The algorithm employs kinematic constraints to ensure that the generated path not only matches the user's physical capabilities but also allows for safe and efficient navigation to the target location. This is achieved through a multi-objective optimization function. Three key decision parameters were output: optimal movement speed. Optimal steering angular velocity and overall score .

[0076] In addition, this embodiment also uses bone conduction headphones and wristband vibration to guide the user to walk along a locally planned path, and provides sound-tactile feedback when the path deviates from the planned direction, as shown in Tables 1 and 2.

[0077] Table 1. Cooperative Boot Command Rules

[0078]

[0079] Table 2 Obstacle Distance Classification Response

[0080]

[0081] Formula 1:

[0082]

[0083] In the formula, d For distance.

[0084] Formula 2:

[0085]

[0086] Formula 3:

[0087]

[0088] like Figure 2 As shown, an immersive 3D assisted navigation and interaction system for the blind includes an environmental sensing device, a central processing unit, and a multimodal interaction device.

[0089] The environmental perception device includes a binocular vision module, a lightweight lidar, and an inertial measurement unit. The binocular vision module is equipped with a high-resolution imaging system and active infrared structured light, which are used to achieve three-dimensional environmental semantic perception by combining point clouds. The lightweight lidar acquires spatial point clouds by scanning, which are used to combine with visual images for three-dimensional environmental spatial perception. The inertial measurement unit integrates a three-axis accelerometer and a gyroscope to estimate the user's pose.

[0090] The central processing unit comprises a multi-sensor fusion module, a dynamic environment modeling module, a state estimation module, and an intelligent path planning module connected in sequence. The multi-sensor fusion module is also connected to a binocular vision module and a lightweight LiDAR. Specifically, the multi-sensor fusion module performs visual image-LiDAR point cloud registration based on an improved iterative nearest-point algorithm; the dynamic environment modeling module uses an innovative Dynamic-Log-Odds method combined with an octree map to dynamically model the environment; the state estimation module integrates inertial measurement unit, visual, and point cloud data to estimate the user's pose; and the intelligent path planning module obtains the global navigation path and local target points through the navigation software system. Based on the 3D environment modeling results, combined with the local navigation target points and the current location, it plans a navigation path that can safely traverse the modeled 3D environment and reach the navigation target point based on a multi-objective optimization function.

[0091] The multimodal interaction device includes a haptic guide wristband and bone conduction headphones, both of which are connected to a central processing unit. The haptic guide wristband guides the user to walk along a navigation route through vibration, while the bone conduction headphones play audio through bone conduction to guide the user to walk along the route announced by the sound.

[0092] like Figure 3As shown, the system also includes glasses, and the binocular vision module includes a left vision camera and a right vision camera, which are located on the sides of the glasses frame, respectively, while a lightweight LiDAR is located in the center of the glasses frame.

[0093] The central processing unit and the inertial measurement unit are respectively located on the two temples of the glasses, and the bone conduction headphones are also located on the two temples of the glasses.

[0094] The tactile guidance wristband is equipped with an integrated piezoelectric ceramic vibration unit.

[0095] like Figure 4 As shown, the usage scenario of this invention's system is as follows: The user sets the red-marked location as the destination. The navigation software generates a global path and local target points. The user then sequentially reaches the local target points, such as starting position -> fountain -> green house -> destination, ultimately arriving at the destination. The user walks along the yellow global navigation path shown in the diagram. The system performs real-time 3D environment modeling. When a rock obstacle is detected, the bone conduction headset issues an obstacle avoidance command, "Obstacle ahead on the right," and the left wristband vibrates, guiding the user to turn left to avoid the obstacle. The user then continues along the global navigation path. Upon encountering another obstacle, at a distance of 3m, the bone conduction headset issues a voice warning, "Obstacle ahead on the right," followed by a left turn command. The left wristband vibrates, guiding the user to turn left to avoid the obstacle. The user does not follow the guidance and gets closer to the obstacle. At a distance of 0.8m, the bone conduction headset issues a voice warning, "Danger! Stop immediately!" and both wristbands vibrate simultaneously. The user then follows the system's guidance and ultimately successfully reaches the destination.

[0096] In the description of this invention, it should be understood that the terms "center," "thickness," "upper," "lower," "horizontal," "top," "bottom," "inner," "outer," and "radial," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying the relative importance or the number of technical features implicitly specified. Therefore, a feature defined by "first," "second," and "third" may explicitly or implicitly include one or more of that feature.

Claims

1. A method for immersive 3D assisted navigation and interaction for the blind, characterized in that, Includes the following steps: S1. Acquire lidar point cloud data and visual images, and perform visual image-lidar point cloud registration using an improved iterative nearest point algorithm to generate a fused point cloud; S2. Based on the fused point cloud, model the octree map to generate a 3D environment map; S3. Acquire IMU data and estimate the user's pose; S4. Based on the user's pose and the 3D environment map, plan the navigation path and guide the user to walk along the navigation path. In S1, the specific method for visual image-LiDAR point cloud registration is as follows: calculate the rigid transformation matrix of the registered point cloud. ; In the formula, arg This represents the parameter that causes the function to reach its extreme value. This indicates a search for the function that yields its minimum value. T , Indicates the accumulation symbol. T This represents the rigid body transformation matrix that needs to be optimized. This indicates the coordinates of the feature points captured by the camera. This indicates the coordinates of the feature points acquired by the lidar. express The corresponding surface normal vector, express The corresponding surface normal vector, and Indicates material information encoding. Indicates the weighting factor. This represents the proportion of the control surface normal vector error in the overall objective function. This indicates the proportion of the control encoding material information error in the overall objective function. Represents the square of the Euclidean distance; In S4, the method for planning navigation paths based on the user's pose and the 3D environment map is as follows: S41. Obtain the global path point sequence based on the global target points determined by the user, and generate local target points based on the user's current location; S42. Evaluate the motion scheme of local target points through a multi-objective optimization function, and then generate a navigation path; In S41, the local target point The specific expression is: In the formula, Represents a global path point. Indicates the user's current location. Indicates constraints. Represents the L2 norm; In S42, the multi-objective optimization function The specific expression is: In the formula, Indicates to local target The degree of proximity, This indicates the calculation of the safe distance to the nearest obstacle. This indicates the consistency between the direction of movement and the overall objective. v Indicates movement speed. w Indicates the steering angular velocity. This represents the weight that controls the forward efficiency. This indicates the weight used to control obstacle avoidance safety. The weight representing the correct control direction.

2. The immersive three-dimensional assisted navigation and interaction method for the blind according to claim 1, characterized in that, In S2, the voxel occupancy probability is updated using the following formula during the octet map modeling process; In the formula, This represents the logarithmic probability value of the voxel at the current moment. This represents the historical state value at the previous moment. This indicates that based on the current sensor observation data z The calculated instantaneous occupancy probability, These are the initial prior values. It is a logarithmic function; In the formula, Indicates navigation prior values, This represents the first frame observation value from the sensor. This represents the weighting coefficient.

3. The immersive three-dimensional assisted navigation and interaction method for the blind according to claim 2, characterized in that, In S3, the objective function for estimating the user's pose is specifically: In the formula, This represents the 16-dimensional state vector to be optimized. This represents the 6-dimensional raw measurement data from the IMU sensor. This represents the position observations of a GNSS receiver in three dimensions or higher. express The corresponding motion model, express The corresponding observation model.

4. An immersive three-dimensional assisted navigation and interaction system for the blind, applied to the immersive three-dimensional assisted navigation and interaction method as described in any one of claims 1 to 3, characterized in that, The system includes environmental sensing devices, a central processing unit, and a multimodal interaction device; The environmental sensing device includes a binocular vision module, a lightweight lidar, and an inertial measurement unit; The central processing unit includes a multi-sensor fusion module, a dynamic environment modeling module, a state estimation module, and an intelligent path planning module connected in sequence. The multi-sensor fusion module is also connected to a binocular vision module and a lightweight LiDAR, respectively. The multimodal interaction device includes a haptic guide wristband and bone conduction headphones, both of which are connected to a central processing unit.

5. The immersive three-dimensional assisted navigation and interaction system for the blind according to claim 4, characterized in that, The system also includes glasses, with a binocular vision module consisting of a left vision camera and a right vision camera, located on the sides of the glasses frame, and a lightweight LiDAR located in the center of the glasses frame.

6. The immersive three-dimensional assisted navigation and interaction system for the blind according to claim 4, characterized in that, The central processing unit and the inertial measurement unit are respectively located on the two temples of the glasses, and the bone conduction headphones are also located on the two temples of the glasses.

7. The immersive three-dimensional assisted navigation and interaction system for the blind according to claim 4, characterized in that, The tactile guidance wristband is equipped with an integrated piezoelectric ceramic vibration unit.

Images