A desktop lifting control method based on teaching video

Abstract

The application provides a desktop lifting control method based on a teaching video, and the desktop lifting control method comprises the following steps: extracting a first key frame image at the t moment in the teaching video; obtaining reference object information and multiple human body target key points in the first key frame image; analyzing the total length of the teaching video to obtain a target height curve composed of human body target part heights at different moments; and optimizing the target height curve according to the reaction condition of the desktop to obtain an optimized height control curve for the desktop to execute a lifting action. The problem solved by the application is that the process of adjusting the lifting desktop is controlled by a user through a remote controller, so that the user needs to frequently adjust the remote controller according to different types of sports teaching videos, and the lifting reaction efficiency of the lifting desktop cannot keep up with the sports teaching video in the adjustment process, which leads to a decrease in the coherence of fitness and a decrease in the fitness efficiency of the user.

Description

Technical Field

[0001] This invention relates to the field of height-adjustable desk technology, and more specifically, to a desktop height adjustment control method based on instructional videos. Background Technology

[0002] As society progresses, people's awareness of fitness is gradually increasing. However, due to the limitations of fitness venues and time, the demand for doing fitness exercises at home during fragmented time is gradually increasing, and various sharing platforms are launching more and more instructional videos of body movements.

[0003] Specifically, users typically place the playback device in a fixed position before starting their workout. However, while watching the exercise video and performing the exercises, their gaze often shifts away from the screen as they follow the movements, diminishing the workout experience. Furthermore, to keep their eyes on the screen at all times, users often need to adjust their head angle during the exercises, further reducing the effectiveness of the workout.

[0004] Although some related technologies involve placing the playback device on a height-adjustable table so that users can adjust the table's height via remote control to raise or lower the playback device, thus ensuring that the head does not need to make unnecessary movements and improving the user's fitness experience.

[0005] However, the relevant technology has at least one of the following problems: Since the adjustment of the height-adjustable table is done by the user controlling the remote control, the user needs to frequently adjust the remote control according to different types of exercise instruction videos. In addition, the adjustment process may also lead to a decrease in the continuity of fitness and a reduction in the user's fitness efficiency because the height-adjustable table's response efficiency cannot keep up with the exercise instruction videos. Summary of the Invention

[0006] The problem this invention solves is that the adjustment of the height-adjustable table is done by the user controlling the remote control, which requires the user to frequently adjust the remote control according to different types of exercise instruction videos. Furthermore, the adjustment process can lead to a decrease in the continuity of the workout and a reduction in the user's workout efficiency because the height-adjustable table's response efficiency cannot keep up with the exercise instruction videos.

[0007] To address the aforementioned problems, this invention provides a desktop height control method based on instructional videos, comprising: Step S1: Extracting the first keyframe image at time t in the instructional video; Step S2: Obtaining reference object information and multiple human target key points from the first keyframe image; wherein, the reference object information includes multiple reference object key points and external dimensions; Step S3: Determining the coordinates of multiple corner points corresponding to the multiple reference object key points; Step S4: Obtaining the world coordinates of the corresponding multiple reference object key points by quantitatively determining camera intrinsic parameters and external dimensions; Step S5: Solving for camera extrinsic parameters on the multiple corner point coordinates and world coordinates; Step S6: Obtaining and recording the height of the human target body parts corresponding to the multiple human target key points by matrix calculation based on camera imaging principles, combined with the detection of multiple human target key points; wherein, the height of the human target body parts includes the height of the human shoulder or the height of the human eye; Step S7: Analyzing the total duration of the instructional video to obtain a target height curve composed of the heights of the human target body parts at different times; Step S8: Optimizing the target height curve according to the desktop's reaction conditions to obtain an optimized height control curve for the desktop to perform height adjustment actions; wherein, the reaction conditions include the height adjustment rate.

[0008] Compared with existing technologies, the technical effects achieved by this solution are as follows: Considering the actual hardware conditions of the height-adjustable table, it is difficult for the height-adjustable table to perfectly match the height-adjusting actions in the teaching video. Often, the height changes caused by the character in the teaching video jumping multiple times in a short period of time will occur. For example, due to the limited lifting rate, when the height-adjustable table rises from one height to the height that matches the first moment in the teaching video, the character in the teaching video is already at a different height. This results in a lag in the height-adjusting table's response compared to the teaching video, leading to inconsistencies in the height and lifting actions of the two at the same moment. Therefore, by combining this technical solution with the inherent limitations of the height-adjustable desk, and by optimizing the target height curve, on the one hand, it ensures that the frequency of the desk's tabletop lifting motion is consistent with the optimized height control curve, avoiding a decrease in the user's fitness experience due to inconsistencies in the lifting frequencies. On the other hand, since the human shoulder or eye level is used as a key point to represent changes in human height, it is easier to accurately identify and capture compared to other parts of the human body. This ensures that when the lifting frequency of the desk is consistent with the height changes of the target human body parts in the instructional video, the user's line of sight height changes, the desk's tabletop height, and the height changes of the target human body parts are all matched when performing fitness movements according to the instructional video. This improves the user's comfort during the fitness process and prevents the user from triggering movements that are harmful to their health by trying to keep pace with the desk's tabletop height changes.

[0009] In one embodiment of the present invention, step S8 includes: within any first preset time period of the total playback duration, controlling the total travel distance of the desktop to satisfy the following formula 1; Formula 1: |▽x|1≤aTv; where x is the total vector of desktop height change within the first preset time period, |▽x|1 is the total travel distance, T is the total duration of the teaching video, v is the lifting and lowering rate of the desktop at a constant speed, and a is the overheat safety factor.

[0010] Compared with existing technologies, the technical effects achieved by this solution are as follows: Due to the limitations of the internal drive mechanism of the height-adjustable desk, frequent height-adjustment movements cannot be performed within a set time. Frequent height-adjustment movements within a set time would cause the internal drive mechanism to overheat, potentially leading to burnout. Therefore, by constraining the height-adjustment movement of the desk, ensuring that it performs its movements within the constraints of Formula 1, the normal service life of the height-adjustable desk is guaranteed.

[0011] In one embodiment of the present invention, step S8 includes: optimizing the target height curve using the L2 norm.

[0012] Compared with existing technologies, the technical effect achieved by adopting this technical solution is to ensure that the lifting frequency between the two can remain consistent when the height-adjustable desk moves up and down according to the optimized target height curve.

[0013] In one embodiment of the present invention, optimizing the target height curve using the L2 norm includes: determining whether there is a phased rise and fall portion in the target height curve; if so, optimizing the phased rise and fall portion into a first height portion; wherein, the phased rise and fall portion is a portion in the target height curve where the target height rises or falls at least once or alternates between rising and falling at least once within a second preset time period, and the first height portion includes at least a portion where the height remains unchanged.

[0014] Compared with existing technologies, the technical effects achieved by this technical solution are: when the height-adjustable desk moves up and down according to the optimized target height curve, the height adjustment frequency between the two can be kept consistent, and the height adjustment range between the two can also be adapted.

[0015] In one embodiment of the present invention, optimizing the phased lifting and lowering portion into a first height portion includes: obtaining a first target height point within a second preset time period and a second target height point obtained in the next moment; determining whether the height difference between the first target height point and the second target height point is greater than a preset difference; if so, optimizing the height control curve according to the lifting and lowering rate of the desktop; if not, optimizing the portion of the target height curve corresponding to the first target height point to the second target height point to keep the height constant.

[0016] Compared with existing technologies, the technical effects achieved by this solution are as follows: It further improves the consistency of height adjustment frequency and range between the two when the height-adjustable desk moves according to the optimized target height curve. Furthermore, it avoids the increased heat generation caused by frequent reciprocating height adjustments within a short period. If the height difference in the instructional video is less than the preset difference, it is understandable that although there are one or more changes in the height of the target body part during that time period, the magnitude of the change is small and has little impact on the user's actual learning and viewing experience. Considering the overheating issue of the height-adjustable desk, the target height curve can be optimized to maintain a constant height during that time period, thereby reducing the heat generation of the desk during that period.

[0017] In one embodiment of the present invention, optimizing the target height curve using the L2 norm includes: generating a desktop height control curve based on the target height curve; establishing an objective function using the target height curve and the desktop height control curve to obtain the optimized height curve; the objective function is: Where h(t) is the target height curve, x(t) is the desktop height control curve, and t is the time in the total duration.

[0018] Compared with existing technologies, the technical effect achieved by adopting this technical solution is to further ensure that the optimized height control curve obtained after optimization is adapted to the changes in the desktop height of the height-adjustable desk.

[0019] In one embodiment of the present invention, a storage module storing an optimized height control curve is imported into the height-adjustable desk; the control module of the height-adjustable desk reads the storage module and controls the desktop to move up and down according to the optimized height control curve.

[0020] By adopting the technical solution of the present invention, the following technical effects can be achieved:

[0021] (1) On the one hand, it ensures that the frequency of the tabletop lifting action of the height-adjustable table is consistent with the optimized height control curve, thus avoiding the reduction of the user's fitness experience due to the inconsistency of the lifting frequency of the two; on the other hand, since the human shoulder or human eye is used as the key point to represent the change of human height, it is easier to be accurately identified and obtained compared with other parts of the human body. This ensures that when the lifting frequency of the height-adjustable table is consistent with the change of height of the target human body part in the teaching video, the user's line of sight height change, the tabletop height of the height-adjustable table and the change of height of the target human body part are matched when the user performs fitness movements according to the teaching video, thereby improving the user's comfort during the fitness process and avoiding the user from triggering movements that are harmful to the body by keeping in step with the change of tabletop height of the height-adjustable table.

[0022] (2) By constraining the lifting motion of the tabletop, the lifting motion is required to be performed under the constraint of Formula 1, thus ensuring the normal service life of the lifting table.

[0023] (3) It further improves the height adjustment frequency between the two when the height adjustment table moves according to the optimized target height curve, and the height adjustment range between the two can also be matched. In addition, it avoids the frequent reciprocating height adjustment of the height adjustment table in a short period of time, which would increase its heat generation. If the height difference in the teaching video is less than the preset difference, it is understandable that although there is one or more changes in the height of the target human body part in the teaching video during this period, the change range is small and has little impact on the user's actual learning and viewing of the screen. In combination with the overheating problem of the height adjustment table, the target height curve can be optimized to keep the height constant during this period of height change, thereby reducing the heat generation of the height adjustment table during this period of time. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating a desktop lifting control method based on instructional videos, provided as an embodiment of the present invention.

[0025] Explanation of reference numerals in the attached figures: Detailed Implementation

[0026] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0027] See Figure 1 This is a flowchart illustrating a desktop height control method based on instructional videos provided in an embodiment of the present invention. The desktop height control method includes, for example:

[0028] Step S1: Extract the first keyframe image at time t from the teaching video;

[0029] Step S2: Obtain reference object information and multiple human target key points in the first keyframe image; wherein, the reference object information includes multiple reference object key points and their external dimensions; wherein, the human target key points in the teaching video can be obtained through the MediaPipe algorithm.

[0030] Step S3: Determine the coordinates of multiple corner points based on the key points of multiple reference objects; among them, the coordinates of multiple corner points can be obtained by improving the YOLOv5 algorithm, which can improve the original YOLOv5 algorithm from only detecting the object's bounding box to being able to identify the object's corner points.

[0031] For example, a yoga mat can be used as a reference object. The yoga mat can be regarded as a near-standard cuboid (that is, the four corner points can be regarded as the four corner points of the cuboid). By using the intrinsic parameters of the quantitative camera and the length and width of the yoga mat, the world coordinates of the four corner points of the yoga mat can be obtained respectively (the yoga mat always stays on the ground, so the z-axis coordinate in the world coordinates of the yoga mat is always 0).

[0032] Specifically, the world coordinates of the yoga mat are arranged clockwise around its four corners as follows: bottom left (p0): (0, 0, 0); top left (p1): (0, w, 0); top right (p2): (L, w, 0); bottom right (p3): (L, 0, 0).

[0033] Step S4: Obtain the world coordinates of key points corresponding to multiple reference objects using quantitative camera intrinsic parameters and external dimensions;

[0034] Step S5: Solve for camera extrinsic parameters using the coordinates of multiple corner points and world coordinates;

[0035] Step S6: Through matrix calculation based on camera imaging principle, combined with the detection of multiple human target key points, obtain and record the height of the human target part corresponding to multiple human target key points; wherein, the height of the human target part includes the height of the human shoulder or the height of the human eye.

[0036] Specifically, the Perspective-n-Point (PNP) algorithm is used to solve for camera extrinsic parameters by aligning the coordinate systems of the four point images with the world coordinate system. After obtaining the camera extrinsic parameters, matrix calculations based on camera imaging principles are used, combined with the detection of key points such as human shoulder key points, to obtain and record the height of human shoulder key points in each frame.

[0037] Step S7: Analyze the total duration of the instructional video to obtain a target height curve composed of the height of the target human body parts at different times;

[0038] Step S8: Optimize the target height curve based on the desktop's response conditions to obtain an optimized height control curve for the desktop to perform lifting and lowering actions; wherein, the response conditions include the lifting and lowering rate.

[0039] Specifically, multiple human body key points can be highly correlated with the target body part, such as multiple shoulder key points; or multiple eye key points. Considering the accuracy of key points in actual teaching videos, compared to head and eye key points, the shoulder key points are easier to reliably acquire and less prone to loss because the shoulder area occupies a larger image size in the teaching video than the eye area. If eye key points are needed, simply add the corresponding heights to the obtained data.

[0040] In a specific example, fitness instructors in instructional videos can make significant changes in their movements within a short period, and can also remain stationary for a while. Correspondingly, the lifting rate of an adjustable table is typically constant, ensuring that items on its surface remain relatively stable during ascent or descent. To improve user fitness efficiency, the lifting motion of the table is correlated with the key points of the human body in the instructional video in the direction of lifting, thus simulating the lifting table as the person in the video and mimicking their lifting motion when performing fitness exercises.

[0041] Specifically, ideally, an adjustable desk should rise or fall in a manner consistent with the target body's key points in the vertical direction. This would allow users to avoid engaging in unrelated actions when following instructional videos, such as having to look down to watch if the playback device is placed too low. Alternatively, if a user is in a dynamic environment, such as jumping, while the playback device is stationary, their line of sight may not be synchronized with the device's position, causing inconvenience during workouts.

[0042] Considering the actual hardware limitations of height-adjustable desks, it's difficult for them to perfectly match the height-adjusting movements in instructional videos. Often, the height changes caused by multiple jumps in the instructional video occur within a short period. For example, due to the limited lifting speed, when the desk rises from one height to the height that matches the first moment in the instructional video, the instructional video may already be at a different height. This results in a lag in the height-adjusting response of the desk compared to the instructional video, leading to inconsistencies in the height and movement of the two desks at the same moment.

[0043] Therefore, by combining this technical solution with the inherent limitations of the height-adjustable desk, and by optimizing the target height curve, on the one hand, it ensures that the frequency of the desk's tabletop lifting motion is consistent with the optimized height control curve, avoiding a decrease in the user's fitness experience due to inconsistencies in the lifting frequencies. On the other hand, since the human shoulder or eye level is used as a key point to represent changes in human height, it is easier to accurately identify and capture compared to other parts of the human body. This ensures that when the lifting frequency of the desk is consistent with the height changes of the target human body parts in the instructional video, the user's line of sight height changes, the desk's tabletop height, and the height changes of the target human body parts are all matched when performing fitness movements according to the instructional video. This improves the user's comfort during the fitness process and prevents the user from triggering movements that are harmful to their health by trying to keep pace with the desk's tabletop height changes.

[0044] Preferably, step S8 includes:

[0045] Within any first preset time period of the total playback duration, the total movement of the control desktop satisfies the following formula 1;

[0046] Formula 1: |▽x|1≤aTv;

[0047] Where x is the total vector of desktop height change within the first preset time period, |▽x|1 is the total motion distance, T is the total duration of the teaching video, v is the rising and falling rate of the desktop at a constant speed, and a is the overheat safety factor.

[0048] In another specific example, in conjunction with the above-mentioned details, when the desktop of a height-adjustable desk is raised or lowered, it often generates a lot of heat due to its mechanical movement. The faster the movement frequency or the greater the total amount of movement in a short period of time, the more likely it is to overheat and burn out the internal drive components of the desk.

[0049] Therefore, due to the limitations of the internal drive mechanism of the height-adjustable desk, frequent height adjustments within a set time are not possible. Frequent height adjustments within this timeframe would cause the internal drive mechanism to overheat, potentially leading to burnout. Therefore, by constraining the height adjustment of the desk's desktop, ensuring that it operates within the constraints of Formula 1, the normal lifespan of the desk is guaranteed. The value of 'a' is related to the desk's heat dissipation capacity; a larger capacity allows for a larger value, and vice versa. Ideally, without considering heat dissipation, 'a' is set to 1.

[0050] For example, if the target object in the teaching video moves faster than the table's lifting speed within a set time period, the total travel distance of the target object will be greater than the preset total travel distance. Conversely, considering the actual situation, since the table's lifting speed is uniform (i.e., uniform ascent or descent), it should be noted that during the actual lifting process, such as the transition from a stationary state with a fixed height and zero speed to a dynamic state of uniform ascent or descent, the speed will inevitably increase from zero to the set speed, i.e., there is an acceleration process. However, since this acceleration process occupies a very small portion of the total table movement time, and considering the actual situation of table lifting control, the uniform ascent or descent speed is small, thus further reducing the time occupied by the aforementioned acceleration process. Therefore, in this technical solution, the time occupied by the acceleration process is ignored.

[0051] On the other hand, the same applies when the desktop's rate changes direction in a short period of time, such as changing from an upward to a downward process, which will not be elaborated here.

[0052] Preferably, step S8 includes:

[0053] The target height curve is optimized using the L2 norm.

[0054] Preferably, optimizing the target height curve using the L2 norm includes:

[0055] Determine whether there are phased rises and falls in the target height curve;

[0056] If so, the phased rise and fall section will be optimized into the first height section;

[0057] The phased rise and fall section refers to the target height curve in which the target height rises or falls at least once or alternates between rising and falling at least once within a second preset time period, and the first height section includes at least a part that maintains a constant height.

[0058] Preferably, the phased lifting section is optimized into the first height section, including:

[0059] Obtain the first target elevation point within the second preset time period and the second target elevation point obtained in the next time period;

[0060] Determine whether the height difference between the first target height point and the second target height point is greater than a preset difference value;

[0061] If so, the optimized height control curve is obtained by optimizing the desktop's lifting rate.

[0062] If not, then optimize the portion of the target height curve from the first target height point to the second target height point to keep the height constant.

[0063] Preferably, optimizing the target height curve using the L2 norm includes:

[0064] Based on the target height curve, a desktop height control curve is generated for the desktop.

[0065] An objective function is established using the target height curve and the desktop height control curve to obtain the optimal height curve;

[0066] The objective function is:

[0067] Where h(t) is the target height curve, x(t) is the desktop height control curve, and t is the time in the total duration. It is the sum of the squares of the height differences between the target height curve and the desktop height control curve at each moment.

[0068] Preferably, the storage module containing the optimized height control curve is imported into the height-adjustable table;

[0069] The control module of the height-adjustable desk reads from the storage module and controls the desktop to move up and down according to the optimized height control curve.

[0070] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for controlling a desk lift based on a teaching video, characterized by, The method comprises the following steps: Step S1: extracting a first key frame image at the t-th moment in the teaching video; Step S2: obtaining reference object information and a plurality of human target key points in the first key frame image; wherein the reference object information comprises a plurality of reference object key points and an outline size; Step S3: determining a plurality of corresponding corner point coordinates according to the plurality of reference object key points; Step S4: obtaining world coordinates corresponding to the plurality of reference object key points by quantifying the camera intrinsic parameters and the outline size; Step S5: solving the camera extrinsic parameters for the plurality of corner point coordinates and the world coordinates; Step S6: obtaining and recording human target part heights corresponding to the plurality of human target key points through matrix calculation of the camera imaging principle and in combination with the detection of the plurality of human target key points; wherein the human target part heights comprise human shoulder heights or human eye heights; Step S7: analyzing the total length of the teaching video to obtain a target height curve composed of the human target part heights at different moments; Step S8: optimizing the target height curve according to the reaction conditions of the desktop to obtain an optimized height control curve for the desktop to perform lifting actions; Wherein the reaction conditions include lifting speed.

2. The table lift control method of claim 1, wherein, The step S8 comprises: In any first preset time period in the total length, the total travel of the desktop is controlled to satisfy the following formula 1; Formula 1: |▽x|1≤aTv; Wherein x is the total height change vector of the desktop in the first preset time period, |▽x|1 is the total travel, T is the total length of the teaching video, v is the lifting speed of the uniform motion of the desktop, and a is the overheating safety factor.

3. The table lift control method of claim 2, wherein, The step S8 comprises: Optimizing the target height curve by two-norm.

4. The table lift control method of claim 3, wherein, The optimization of the target height curve by two-norm comprises: Judging whether there is a stage lifting part in the target height curve; If yes, the stage lifting part is optimized to a first height part; Wherein the stage lifting part is a target height that appears at least once in a second preset time period, and the first height part at least includes a part that maintains the height unchanged.

5. The table lift control method of claim 4, wherein, The optimization of the stage lifting part to the first height part comprises: Obtaining a first target height point in the second preset time period and a second target height point obtained at the next moment; Judging whether the height difference between the first target height point and the second target height point is greater than a preset difference value; If yes, the optimized height control curve is obtained according to the lifting speed of the desktop; If not, the part of the target height curve corresponding to the first target height point to the second target height point is optimized to keep the height unchanged.

6. The desktop lifting control method according to any one of claims 3-5, wherein The optimization of the target height curve by two-norm comprises: Generating a desktop height control curve of the desktop according to the target height curve; a target function is established by the target height curve and the desktop height control curve to obtain the optimized height curve; The objective function is: ; wherein h(t) is the target height curve, x(t) is the desktop height control curve, and t is the time in the total time length.

7. The desktop lifting control method according to claim 1, characterized in that, a storage module storing the optimized height control curve is imported into the lifting table; a control module of the lifting table reads the storage module to control the desktop to perform lifting movement according to the optimized height control curve.

Images