A method and system for simulating a jump rope count

Abstract

This application provides a simulated rope skipping counting method and system, belonging to the field of computer vision and artificial intelligence technology. The method includes: acquiring a video frame sequence containing the user's movements and extracting the coordinates of key points on the wrist and ankle; identifying jumping events based on the coordinate sequence of the ankle key points and determining their characteristic time points; identifying rope-swinging movements by analyzing the periodicity and closure of the wrist's movement trajectory based on the coordinate sequence of the wrist key points; for each jumping event, determining whether rope-swinging movements exist within a preset time window centered on its characteristic time point; if so, classifying the jumping event as a valid rope skipping movement and counting it. This application, through collaborative verification of upper and lower limb movements, can effectively distinguish between real rope skipping and interfering movements such as jumping in place, significantly improving counting accuracy and anti-interference capability.

Description

Technical Field

[0001] This application relates to the fields of computer vision and artificial intelligence, and in particular to a simulated rope skipping counting method and system based on upper and lower limb collaborative verification. Background Technology

[0002] With the popularization of national fitness and smart sports, rope skipping has become a widely used and highly efficient aerobic exercise. Among these, cordless rope skipping has become a popular indoor training method due to its advantages such as low space requirements and high safety. To automatically count cordless rope skips, existing technologies typically employ computer vision-based solutions.

[0003] For example, a common approach is to capture video of a user's movement using a camera and then use a human pose estimation algorithm to obtain the coordinates of key points on the user's body. This approach then analyzes the coordinate changes of lower limb key points, such as the hip and ankle key points, particularly their vertical displacement, to determine whether the user has completed a jump, and uses this as a basis for counting.

[0004] However, the above scheme only analyzes the user's lower limb jumping behavior, neglecting the equally important upper limb rope-swinging motion in rope skipping. This makes it impossible for the scheme to effectively distinguish between real simulated rope skipping movements and other periodic movements that only involve lower limb jumping, such as high knees and vertical jumps. Therefore, in practical applications, these non-rope skipping movements are often incorrectly counted as valid repetitions, resulting in low accuracy and poor resistance to interference, making it difficult to meet the high accuracy requirements of scenarios such as physical education teaching or physical fitness testing. Summary of the Invention

[0005] The purpose of this application is to provide a simulated rope skipping counting method, system, electronic device and storage medium, which aims to solve the technical problem that the existing pure visual rope skipping counting scheme relies only on lower limb jumping analysis, which makes it impossible to distinguish real simulated rope skipping movements from other similar jumping movements, resulting in inaccurate counting and weak anti-interference ability.

[0006] To achieve the above objectives, this application provides a simulated rope skipping counting method, comprising the following steps: acquiring a video frame sequence containing a user exercising; applying a human posture estimation algorithm to the images in the video frame sequence to extract coordinates of at least wrist and ankle key points; identifying jump events by monitoring the periodic changes in the vertical displacement of the ankle key points based on the coordinate sequence of the ankle key points, and determining the characteristic time point of each jump event; identifying rope-swinging actions by analyzing the periodicity and closure of the movement trajectory of the wrist key points based on the coordinate sequence of the wrist key points; determining whether the rope-swinging action exists within a preset time window centered on its characteristic time point for each jump event; and, when it exists, determining the jump event as a valid rope skipping action and counting it.

[0007] Optionally, analyzing the periodicity of the motion trajectory includes: performing frequency domain analysis on the coordinate sequence of the wrist key points to determine whether the main frequency of the coordinate sequence falls within a preset frequency range of 2 Hz to 4 Hz.

[0008] Optionally, analyzing the closure of the motion trajectory includes: calculating the ratio of the minimum enclosing circle diameter of the motion trajectory formed by the wrist key points in consecutive frames to the length of the motion trajectory; and determining that the motion trajectory satisfies the closure condition when the ratio is greater than a preset closure threshold.

[0009] Optionally, the step of identifying a jump event specifically includes: extracting the coordinates of the hip center key point output by the human posture estimation algorithm; and identifying the jump event by analyzing whether there is a downward-upward pulse pattern in the acceleration sequence of the hip center key point in the vertical direction.

[0010] Optionally, the method further includes: before recognizing the jumping event and recognizing the rope-swinging action, performing a validity judgment on each frame image, the validity judgment including: obtaining the confidence scores of the wrist key points and the ankle key points, or the hip center key points; calculating the average value of the confidence scores; and when the average value is lower than a preset frame confidence threshold, determining the current frame as an invalid frame and discarding it.

[0011] Optionally, the method further includes: when there are missing data points due to the invalid frame, using linear interpolation or a method based on historical motion trend prediction to complete the coordinates of key points in one or more frames after the invalid frame.

[0012] Optionally, counting includes: after determining the jump event as a valid jump rope action, if the valid jump rope action is the first valid jump rope action to be determined, or the time interval between the valid jump rope action and the last counted valid jump rope action is greater than a minimum time interval threshold of 0.3 seconds, then a counting operation is performed.

[0013] To achieve the above objectives, this application also provides a simulated rope skipping counting method, comprising the following steps: acquiring a video frame sequence containing a user exercising; applying a human pose estimation algorithm to the images in the video frame sequence to extract coordinates of at least wrist and ankle keypoints; identifying candidate jump events based on the coordinate sequence of the ankle or hip center keypoints; for each candidate jump event, extracting comprehensive motion features within a time window around the time of the candidate jump event, the comprehensive motion features including at least one of the following: the velocity of the wrist keypoint, the vertical displacement of the ankle keypoint, and the rate of change of the relative distance between the wrist and ankle keypoints; inputting the comprehensive motion features into a pre-trained machine learning classifier; and determining whether the candidate jump event is a valid rope skipping action and counting it based on the output of the machine learning classifier.

[0014] Optionally, when counting, if the candidate jump event currently judged as a valid jump rope action is the first one, or the time interval between it and the last counted valid jump rope action is greater than a minimum time interval threshold, then the counting operation is performed.

[0015] To achieve the above objectives, this application also provides a simulated jump rope counting system, comprising: one or more processors; and a memory storing instructions, which, when executed by the one or more processors, cause the system to perform the method as described in any one of claims 1 to 9.

[0016] Compared with existing technologies, the technical solution provided in this application has the following beneficial effects: By establishing models of upper limb rope-swinging movements and lower limb jumping events, and performing temporal co-verification of the two, it ensures that counting is only performed when the rope-swinging and jumping movements are coordinated and matched. This effectively distinguishes real simulated rope-jumping movements from similar interference movements such as stationary jumping and high knees, which only involve lower limb jumping. This fundamentally solves the misjudgment problem caused by existing technologies that only analyze lower limb movements, significantly improving the accuracy and anti-interference capability of counting. Furthermore, the method of this application only requires a regular camera, eliminating the need for users to wear any sensors or use special jump ropes, thus freeing it from dependence on dedicated hardware. It has strong universality and is suitable for a wide range of application scenarios such as home fitness and online sports testing. Simultaneously, because the algorithm can run efficiently locally on the terminal device, it effectively protects user privacy and data security while meeting real-time feedback requirements. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a simulated rope skipping counting method based on upper and lower limb coordination verification provided in this application embodiment; Figure 2 This application provides a schematic diagram of the architecture of a simulated jump rope counting system. Figure 3 A schematic diagram of the trajectory for upper limb rope-swinging motion analysis provided in an embodiment of this application; Figure 4 A timing diagram illustrating the verification of upper and lower limb movement coordination provided in an embodiment of this application; Figure 5 The system signaling interaction timing diagram provided for the embodiments of this application.

[0019] Figure Label Explanation: 1-User; 10-Terminal Device; 11-Camera; 12-Display; 20-Processing Unit; 21-Video Acquisition Module; 22-Posture Estimation Module; 23-Upper Limb Analysis Module; 24-Lower Limb Analysis Module; 25-Collaborative Validation Module; 26-Counting and Output Module; 31-Effective Rope Swing Trajectory; 31a-Minimum Enclosing Circle; 32-Invalid Movement Trajectory; 41-Ankle Vertical Height Curve; 41a-Candidate Jump Event; 42-Rope Swing Effectiveness Score Curve; 42a-Effective Rope Swing Time Period; 43-Collaborative Validation Time Window; 44-Effective Count Event; 45-Invalid Jump Event; 501-User; 502-Camera; 503-Processing Module; 504-Display Module; 511-Executed Action; 512-Acquired Video Frame; 513-Executed Action Analysis; 514-Update Display Results; 515-Play Prompt Sound; S10 - Video frame acquisition; S20 - Human body key point detection; S30 - Valid frame selection and posture completion; S40 - Upper limb rope swinging motion analysis; S50 - Lower limb jumping event detection; S60 - Upper and lower limb coordination verification; S70 - Valid rope skipping motion confirmation; S80 - Anti-interference counting; S90 - Result output. Detailed Implementation

[0020] To better understand the technical solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0021] Example 1 This application provides a simulated rope skipping counting method based on upper and lower limb coordination verification. This method models and detects the user's upper limb rope-swinging motion and lower limb jumping motion separately, and verifies their coordination relationship in the time dimension, thereby achieving accurate counting of rope skipping motions without a rope. It can effectively eliminate interference from non-rope skipping motions such as vertical jumps and high knees.

[0022] Please see Figure 2 This illustration shows the architecture of a simulated jump rope counting system provided in an embodiment of this application. As an optional implementation, the system can be embodied as a terminal device 10, such as a smartphone, tablet, or personal computer equipped with a camera. The terminal device 10 integrates a camera 11, a display screen 12, and a processing unit 20 as the computing core. The processing unit 20 is used to execute the method described in this application and can be logically divided into multiple functional modules, including: a video acquisition module 21, a posture estimation module 22, an upper limb analysis module 23, a lower limb analysis module 24, a collaborative verification module 25, and a counting and output module 26.

[0023] The following will combine Figure 1 The flowchart shown illustrates in detail the specific steps of the simulated jump rope counting method provided in this embodiment.

[0024] In step S10, when user 1 begins to perform simulated rope skipping in front of camera 11 on terminal device 10, video acquisition module 21 acquires a sequence of video frames containing user 1's movements in real time through camera 11. It is understood that, to ensure the accuracy of subsequent attitude estimation, the video stream parameters can be set, for example, a resolution of 1280x720 pixels and a frame rate of 30 frames per second.

[0025] In step S20, for each frame of image acquired by the video acquisition module 21, the pose estimation module 22 calls a preset human pose estimation algorithm for processing to detect and locate multiple key points of the user 1's body. In one embodiment of this application, a lightweight algorithm suitable for real-time operation on mobile terminals, such as MediaPipe Pose, can be used. This algorithm can output two-dimensional or three-dimensional coordinates of at least 13 key points, including at least the left wrist, right wrist, left ankle, and right ankle, which are crucial to this application, as well as the left shoulder, right shoulder, left elbow, right elbow, hip center, left knee, and right knee for auxiliary analysis. In addition to coordinate information, the pose estimation algorithm typically provides a confidence score for each detected key point, with a value typically ranging from 0 to 1, to characterize the reliability of the algorithm's judgment of the key point's location.

[0026] In step S30, to improve the robustness and accuracy of subsequent analysis, the system preprocesses the data output by the pose estimation module 22. This preprocessing process can be completed by the general preprocessing module in the processing unit 20 or jointly by the upper limb analysis module 23 and the lower limb analysis module 24, mainly including valid frame screening and pose completion. Specifically, the system checks the confidence of core key points (such as wrists and ankles) in each frame. For example, after calculating the average confidence of the four key points of the left and right wrists and ankles, if the average is lower than a preset frame confidence threshold (e.g., 0.6), the pose estimation result of the current frame is considered to be of low quality due to motion blur, occlusion, or sudden changes in illumination. At this time, the system determines the frame as an invalid frame and discards it, so that it does not participate in subsequent motion analysis. This screening mechanism can effectively avoid noise introduced by low-quality data. After discarding invalid frames, or in the case of missing individual non-core key points (such as elbows) in valid frames, the system can adopt a data completion strategy to ensure the continuity of the data sequence. One possible implementation is linear interpolation, which uses the coordinates of the keypoint in the nearest valid frames to estimate the current position; another approach is to predict based on historical motion trends, such as using a Kalman filter to predict the coordinates of missing points. These completion methods provide a relatively complete and smooth data foundation for subsequent steps such as periodic analysis.

[0027] After preprocessing, the method flow enters two parallel analysis branches: upper limb rope swinging motion analysis in step S40 and lower limb jumping event detection in step S50.

[0028] In step S40, the upper limb analysis module 23 is responsible for analyzing whether the user has a valid rope-swinging motion. This module maintains a fixed-length sliding window to store the wrist key point coordinate sequence over a recent period. For example, the length of this window can be set to 15 frames, corresponding to a time length of 0.5 seconds at a frame rate of 30 frames per second, which is sufficient to cover a complete rope-skipping motion cycle. For the left and right wrists, their two-dimensional coordinate sequences {(xt, yt)} within the window are stored respectively, where t is an integer from 1 to 15. The upper limb analysis module 23 makes a judgment by analyzing the periodicity and closure of these two trajectory sequences.

[0029] Periodic analysis aims to confirm whether wrist movements conform to the frequency of rope skipping. Typical rope skippers operate at a frequency between 2 and 4 times per second, or 2 to 4 Hz. In this embodiment, the upper limb analysis module 23 extracts the vertical coordinate sequence {yt} of the wrist and applies a Fast Fourier Transform to it. The transformed spectrum is analyzed to find the dominant frequency with the highest energy. If this dominant frequency falls within the preset frequency range of 2 to 4 Hz, the wrist movement is considered to meet the periodicity condition.

[0030] Closure analysis aims to determine whether the wrist's movement trajectory approximates a circle or ellipse, a typical characteristic of simulated rope-swinging movements, distinguishing it from non-rope-swinging movements such as forward and backward arm swings. Please refer to... Figure 3 This visually demonstrates the valid and invalid trajectories. A valid rope-swinging trajectory 31 is typically closed or nearly closed, while an invalid movement trajectory 32 (such as random swaying) may be chaotic. In this embodiment, the upper limb analysis module 23 quantifies the closure of the trajectory by calculating the ratio (D / L) of the diameter D of the minimum enclosing circle 31a of the trajectory to the total length L of the trajectory itself. For a perfect circular trajectory, this ratio is theoretically close to 1 / π; however, for a chaotic, non-closed polygonal trajectory, its length L will be much larger than its enclosing circle diameter D, resulting in a very small ratio. Therefore, a preset closure threshold can be set; when the calculated ratio is greater than this threshold, the wrist movement trajectory is determined to meet the closure condition.

[0031] When the movement trajectory of one or both wrists simultaneously meets the above periodicity and closure conditions within the current sliding window, the upper limb analysis module 23 determines that there is a valid rope-swinging action within the current time period and generates a "rope-swinging valid" signal.

[0032] In parallel, in step S50, the lower limb analysis module 24 is responsible for detecting the user's lower limb jumping events. This process is achieved by monitoring the changes in the vertical coordinates (y-coordinates) of key points on both ankles over time. Before or at the beginning of the user's movement, the system may prompt the user to stand still for a few seconds. During this period, the lower limb analysis module 24 records the average vertical coordinates of both ankles and uses them as a personalized static baseline H_base. During the movement, the module continuously tracks the average vertical coordinates Y_ankle of both ankles. When Y_ankle simultaneously satisfies the condition Y_ankle > H_base + Δh (where Δh is a small positive value, such as 5 pixels, used to filter out minor fluctuations caused by instability), the system marks the current state as "off the ground"; subsequently, when Y_ankle falls back to the vicinity of H_base, the system marks the state as "landed". Whenever the system detects a complete transition from the "off the ground" state to the "landed" state, it considers a candidate jump event to have occurred. Please refer to [link to relevant documentation]. Figure 4 Each peak on the ankle vertical height curve 41 represents a candidate jump event 41a. The lower limb analysis module 24 records the timestamp of each candidate jump event, which can be defined as the moment when the ankle reaches its highest point.

[0033] Subsequently, the method proceeds to the core step of this application, namely step S60: upper and lower limb coordination verification. This step is performed by the coordination verification module 25. For each candidate jumping event 41a detected by the lower limb analysis module 24, the coordination verification module 25 does not immediately count it as a valid rope jump, but checks whether there is a valid rope swinging action determined by the upper limb analysis module 23 near the time point of the event.

[0034] Please refer to it again. Figure 4 This diagram vividly illustrates the collaborative verification process. The horizontal axis represents time, the upper axis represents the ankle vertical height curve 41, and the lower axis represents the schematic rope-swinging effectiveness score curve 42. When the upper limb analysis module 23 determines that a valid rope-swinging action exists, the rope-swinging effectiveness score curve 42 is at a high level, forming a valid rope-swinging time period 42a. When the collaborative verification module 25 receives a candidate jump event 41a (e.g., the first peak in the diagram) and its timestamp t_jump, it defines a preset collaborative verification time window 43 centered on t_jump, for example, [t_jump - 0.2s, t_jump + 0.2s]. Then, the module checks whether the rope-swinging effectiveness score curve 42 is at a high level within this 0.4-second time window, i.e., whether there is an overlapping valid rope-swinging time period 42a.

[0035] In step S70, the system makes a decision based on the results of the collaborative verification. If the verification passes, meaning that a valid rope-swinging action does exist within the collaborative verification time window 43, the system determines that the candidate jump event 41a is a valid rope-jumping action. Figure 4 The valid count event 44 is shown in the data. Conversely, if, as shown in the second candidate jump event, the rope-swinging validity score curve 42 is at a low level within its corresponding time window (indicating that the user is not shaking their wrist or is shaking it improperly), then this jump is judged as an invalid jump event 45, such as a simple jumping in place. Accordingly, the system will ignore this event and not count it.

[0036] In step S80, an anti-interference count is performed. For each confirmed valid jump rope action, the counting and output module 26 performs an anti-interference check before updating the total count. This check aims to prevent repeated counting within a very short time due to minor body tremors or rapid flips in the algorithm at critical states. Specifically, the system records the timestamp of the last successful count. The counter is incremented only if the time interval between the current valid jump rope action and the previous valid jump rope action is greater than a preset minimum time interval threshold (e.g., 0.3 seconds, which corresponds to a maximum jump rope frequency of approximately 200 times per minute). If the time interval is too short, it is considered a duplicate detection and is not counted.

[0037] Finally, in step S90, the result is output. The counting and output module 26 updates and displays the final count value, after collaborative verification and anti-interference processing, on the display screen 12 of the terminal device 10 in real time.

[0038] Through the steps described above, the method in this embodiment decomposes the user's jump rope movements into two sub-tasks: upper limb and lower limb. It also strictly requires the two sub-tasks to coordinate in time, thereby achieving high-precision counting. Experimental data shows that, using the solution provided in this embodiment, tested on a self-built test dataset containing 200 video clips, the counting error rate can be stably controlled within 3% compared to manually labeled results. Furthermore, compared to baseline methods that rely solely on lower limb jump analysis, the miscount rate is significantly reduced by 68%.

[0039] Example 2 This embodiment is a variation of Embodiment 1. Its core idea is also based on upper and lower limb coordination verification, but different technical means are used in the two key steps of upper limb rope-swinging motion analysis and lower limb jumping event detection to demonstrate the universality and scalability of the technical solution of this application. The system architecture and overall process of this embodiment are similar to those of Embodiment 1, so they will not be repeated here; only the differences will be described.

[0040] In this embodiment, for the upper limb rope swinging action analysis in step S40, the upper limb analysis module 23 uses a substitution algorithm to determine periodicity and closure.

[0041] For periodicity analysis, as an alternative to the Fast Fourier Transform in Example 1, this example uses the autocorrelation function to analyze the periodicity of the wrist y-coordinate sequence. The autocorrelation function measures the similarity between a signal and itself at different time delays. For a periodic signal, its autocorrelation function will also exhibit the same periodicity. The upper limb analysis module 23 calculates the autocorrelation function of the wrist y-coordinate sequence and finds the first significant peak at a non-zero delay; the delay time at which this peak occurs is the main period of the signal. If the detected main period falls between 0.25 seconds and 0.5 seconds (corresponding precisely to the frequency range of 4 Hz to 2 Hz), the wrist movement is determined to satisfy the periodicity condition. It should be noted that this method may be more stable when dealing with non-stationary or noisy short sequences.

[0042] For closure analysis, as an alternative to the method of calculating the ratio of the minimum enclosing circle to the trajectory length in Example 1, this embodiment fits the set of motion trajectory points of the wrist within 15 consecutive frames into an ellipse. The upper limb analysis module 23 can use techniques such as least squares to find the equation of the ellipse that best fits these trajectory points. Then, the average distance from all trajectory points to the fitted ellipse is calculated, i.e., the fitting residual. It is understood that if the trajectory itself is close to an ellipse, the residual will be small; conversely, if the trajectory is chaotic, the residual will be large. When the calculated fitting residual is less than a preset threshold, the wrist motion trajectory is considered to have good closure.

[0043] For the lower limb jumping event detection in step S50, this embodiment provides an alternative solution that does not directly rely on the ankle keypoint. This solution has better robustness in certain specific scenarios (e.g., the user is wearing loose pants, causing the ankle keypoint to be obscured or inaccurately located). The lower limb analysis module 24 primarily monitors the vertical displacement of the hip center keypoint. Typically, the vertical movement of the body's center of gravity is directly related to jumping, and the hip center is a good approximation of the body's center of gravity. The lower limb analysis module 24 extracts the y-coordinate sequence of the hip center keypoint and calculates its second derivative to obtain the vertical acceleration sequence. From a biomechanical perspective, a typical jumping action involves first squatting (generating downward acceleration) and then exerting force upward (generating significant upward acceleration). Therefore, when the module detects a significant "downward-upward" pulse pattern in the vertical acceleration sequence at the hip center, it identifies it as a candidate jumping event and records the time point of the pulse occurrence as the event's feature time point.

[0044] After completing the above-mentioned alternative analysis and detection, the logic of subsequent steps S60 (upper and lower limb coordination verification), S70 (valid rope skipping action confirmation), S80 (anti-interference counting), and S90 (result output) is exactly the same as in Example 1. The coordination verification module 25 still checks whether the candidate jump events detected in this example (based on hip acceleration) match the valid rope-swinging actions (based on autocorrelation and elliptic fitting) determined in this example in time.

[0045] By employing these alternative algorithms, this embodiment demonstrates that the core idea of ​​this application—upper and lower limb coordinated verification—is not limited to a specific implementation detail, but has broad applicability. Furthermore, the introduction of jump detection based on hip center analysis further enhances the method's applicability and stability under complex clothing conditions.

[0046] Example 3

[0047] This embodiment proposes a variation of the simulated rope skipping counting method. Unlike the previous two embodiments, which use explicit, rule-based logic for collaborative verification, this embodiment transforms the problem of determining "whether it is a valid rope skipping action" into a binary classification problem from the field of machine learning. This data-driven approach enables the system to learn more complex and subtle collaborative patterns from large amounts of data than manually set fixed rules.

[0048] The system architecture and initial steps (video acquisition, pose estimation, data preprocessing) of this embodiment are basically the same as those of embodiments 1 and 2. The core difference lies in the collaborative verification method. In this embodiment, the collaborative verification module 25 is replaced by a pre-trained machine learning classifier.

[0049] The process flow of the method is adjusted accordingly as follows: First, the system still needs to identify candidate jump events through the lower limb analysis module 24. The purpose of this step is to trigger subsequent feature extraction and classification, rather than for direct counting. The identification of candidate jump events can be performed using the methods of Example 1 (based on ankle height) or Example 2 (based on hip acceleration).

[0050] Once a candidate jump event is identified, the system does not independently determine whether there is a "valid rope-swinging action" as in the aforementioned embodiments. Instead, it extracts comprehensive motion features that can fully describe the user's upper and lower limb movement state within a time window (e.g., 0.5 seconds before and after, for a total of 1 second) around the time point when the candidate jump event occurs, and organizes these features into a high-dimensional feature vector.

[0051] This feature vector may include, but is not limited to, information in the following dimensions: 1. Upper limb movement features: for example, the dominant frequency of the Fourier transform of the left and right wrist trajectories within the time window, the proportion of dominant frequency energy, the eccentricity of the trajectory, the curvature variance of the trajectory, the average velocity and maximum velocity of the wrist, etc. 2. Lower limb movement features: for example, the maximum vertical displacement of the hip or ankle (i.e., jump height), the time of takeoff, the maximum horizontal separation distance between the two ankles, etc. 3. Upper and lower limb coordination features: this is the most critical type of feature, used to describe the correlation between upper and lower limb movements. For example, the cross-correlation coefficient between the vertical velocity sequences of the wrist and the vertical velocity sequences of the ankle can be calculated to quantify their phase relationship; the rate of change of the relative distance between key points of the wrist and the key points of the ankle can also be calculated, etc.

[0052] This feature vector, rich in motion information, is then fed into a pre-trained machine learning classifier. This classifier can be a lightweight model, such as a support vector machine, or a small multilayer perceptron neural network. The classifier is trained on a large amount of labeled data, including numerous samples of "valid jump rope" and "invalid jumps" (such as stationary jumps, high knees, and jumping jacks). Through training, the classifier learns how to accurately distinguish between these two types of movements based on the input feature vector.

[0053] The classifier output is a direct judgment result, such as "1" representing "valid jump rope" and "0" representing "invalid jump rope", or outputting a probability value of belonging to "valid jump rope".

[0054] Finally, the counting and output module 26 counts based on the classifier's output. When the classifier output determines that it is a "valid jump rope," the counter is incremented by one. Similar to the previous embodiment, a minimum time interval constraint can also be introduced here to prevent duplicate counting. That is, the counting operation is only performed when the time interval between the jump rope action determined to be valid by the classifier and the previous successful count is greater than a preset threshold (such as 0.3 seconds).

[0055] The advantage of this embodiment lies in its data-driven approach, which allows the classifier to discover and utilize subtle combinations of features that are difficult for humans to precisely describe. For example, for some beginners whose movements are not entirely standard, their rope-swinging trajectories may not perfectly meet the strict requirements of periodicity and closure, but the machine learning model can still correctly identify their rope-swinging intentions by learning other cooperating features. This makes the solution in this embodiment potentially have better generalization ability and tolerance for non-standard movements, thus exhibiting higher overall accuracy in practical applications.

[0056] Example 4 This embodiment, based on the basic solution described in Embodiment 1, further adds a series of intelligent interactive and adaptive adjustment functions. These functions aim to enhance the user experience during actual use and enable the system to better adapt to changing real-world scenarios, transforming it from a simple counting tool into a smart fitness assistant with guidance capabilities.

[0057] Please see Figure 5 This is a system signaling interaction timing diagram provided in this embodiment, illustrating the interaction flow between various entities in this embodiment. Entities include user 501, camera 502, and processing module 503 (corresponding to...). Figure 2 The processing unit 20 and the display module 504 (corresponding to) Figure 2 The display screen 12 in the middle may also include a speaker.

[0058] This embodiment specifically adds the following functions: 1. Adaptive baseline calibration: In many practical applications, users' heights and distances from the camera vary, leading to significant differences in the initial vertical position (y-coordinate) of the ankles in the image. Using a fixed baseline would be difficult to adapt to all situations. Therefore, this embodiment introduces an adaptive calibration mechanism. When the program starts, the system prompts the user 501 to stand still in front of the camera for 3 seconds via a display module 504 (e.g., on-screen text). During this period, the processing module 503 continuously monitors the stability of the user's posture. Once it is confirmed that the user is in a stable standing position, the processing module 503 automatically records the average vertical coordinates of both ankles at this time and sets it as the personalized baseline H_base for this user's current movement. In this way, the subsequent jump detection (step S50) is based on a reference standard tailored to the current user, thereby greatly improving the accuracy of the ground clearance judgment.

[0059] 2. Multi-person independent counting support: In scenarios such as gyms or physical education classes, multiple people may move simultaneously within the camera's range. In this embodiment, when the pose estimation module 22 detects multiple human skeletons in the frame, it assigns a unique tracking ID to each individual skeleton. Subsequently, the processing module 503 instantiates an independent analysis process for each ID, i.e., independently running upper limb analysis, lower limb analysis, and collaborative verification for each user. Therefore, the system can maintain an independent counter for each user in the frame and display the counting results separately on the display module 504, easily supporting scenarios where multiple people train or compete simultaneously.

[0060] 3. Real-time feedback on failed actions: To help users improve the standardization of their actions, the collaborative verification module 25 in this embodiment incorporates a simple state machine to identify typical error patterns. For example... Figure 5As shown, when user 501 performs action 511 and camera 502 captures video frames 512, processing module 503 performs action analysis 513. If processing module 503 detects more than three consecutive candidate jump events (i.e., the user is jumping), but all are judged as invalid jumps due to the lack of matching rope-swinging actions, the system determines that the user may have made the error of "jumping without shaking their arms." At this time, processing module 503 will trigger a feedback operation, such as playing a prompt sound 515 or displaying the text "Please cooperate by shaking your arms!" through display module 504. Conversely, if the system only detects valid rope-swinging actions within a period of time, but does not detect corresponding jump events, it will prompt "Please cooperate by jumping!" This real-time, targeted feedback greatly enhances the interactivity and guidance value of the product.

[0061] 4. Enhanced Result Output and Analysis: In addition to updating the basic total number of jumps on display module 504 (updating display result 514), the counting and output module 26 of this embodiment provides richer analysis results. For example, it can calculate and display the user's average jump rope frequency over the past 5 seconds (converted to "times / minute") in real time to help the user control exercise intensity. Furthermore, this module can provide a comprehensive movement standardization score (e.g., 0-100 points) based on intermediate data provided by upper limb analysis module 23 and lower limb analysis module 24. This score can be based on multiple indicators, such as: the roundness of the rope trajectory (based on the results of closure analysis), the consistency of jump height (based on the peak height variance of candidate jump events), and the synchronicity of upper and lower limb movements (based on the phase difference during co-validation). These enhanced output data provide users with a more comprehensive assessment of their athletic performance.

[0062] By integrating the above functions, the solution provided in this embodiment can not only complete the core counting task, but also significantly improve the user experience and expand the application scenarios through intelligent and user-friendly design, making it more practical in actual home fitness or teaching environments.

[0063] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of simulating a skipping rope count, characterized by, Includes the following steps: Obtain a video frame sequence containing moving users; The human pose estimation algorithm is called on the images in the video frame sequence to extract the coordinates of at least wrist and ankle key points; Based on the coordinate sequence of the ankle key points, jump events are identified by monitoring the periodic changes in their vertical displacement, and the characteristic time points of each jump event are determined. Based on the coordinate sequence of the key wrist points, the rope-swinging action is identified by analyzing the periodicity and closure of its movement trajectory. For each jump event, determine whether the rope-swinging action exists within a preset time window centered on its characteristic time point; and When a jump event occurs, it is determined as a valid jump rope action and counted.

2. The method of claim 1, wherein, The analysis of the periodicity of the motion trajectory includes: Frequency domain analysis is performed on the coordinate sequence of the key points of the wrist to determine whether the main frequency of the coordinate sequence falls within the preset frequency range of 2 Hz to 4 Hz.

3. The method according to claim 1 or 2, characterized in that, The analysis of the closure of the motion trajectory includes: Calculate the ratio of the minimum enclosing circle diameter of the motion trajectory formed by the wrist key points in consecutive frames to the length of the motion trajectory; and When the ratio is greater than a preset closure threshold, the motion trajectory is determined to meet the closure condition.

4. The method according to claim 1, characterized in that, The steps for identifying jump events are as follows: Extract the coordinates of the hip center key point output by the human pose estimation algorithm; and The jump event is identified by analyzing the vertical acceleration sequence of the hip center key point to see if there is a downward-upward pulse pattern.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: Before identifying the jumping event and the rope-swinging action, a validity determination is performed on each frame of the image. The validity determination includes: Obtain the confidence scores of the wrist key points and the ankle key points, or the hip center key points; Calculate the average of the confidence scores; and When the average value is lower than a preset frame confidence threshold, the current frame is determined to be an invalid frame and discarded.

6. The method according to claim 5, characterized in that, The method further includes: When there are missing data points due to the invalid frame, linear interpolation or a method based on historical motion trend prediction is used to complete the coordinates of key points in one or more frames after the invalid frame.

7. The method according to any one of claims 1 to 6, characterized in that, The counting process includes: After determining the jump event as a valid jump rope action, if the valid jump rope action is the first valid jump rope action to be determined, or if the time interval between the valid jump rope action and the last counted valid jump rope action is greater than a minimum time interval threshold of 0.3 seconds, then a counting operation is performed.

8. A method for simulating rope skipping counting, characterized in that, Includes the following steps: Obtain a video frame sequence containing moving users; The human pose estimation algorithm is called on the images in the video frame sequence to extract the coordinates of at least wrist and ankle key points; Candidate jump events are identified based on the coordinate sequence of the ankle key points or the hip center key points; For each candidate jump event, extract the comprehensive motion features within a time window around the time of the candidate jump event. The comprehensive motion features include at least one of the following: the velocity of the wrist key point, the vertical displacement of the ankle key point, and the rate of change of the relative distance between the wrist and ankle key points. The integrated motion features are then input into a pre-trained machine learning classifier; as well as Based on the output of the machine learning classifier, it is determined whether the candidate jump event is a valid rope skipping action and counted.

9. The method according to claim 8, characterized in that, The method further includes: When counting, if the candidate jump event currently judged as a valid jump is the first one, or if the time interval between it and the last counted valid jump is greater than a minimum time interval threshold, then the counting operation is performed.

10. A simulated jump rope counting system, characterized in that, include: One or more processors; as well as A memory storing instructions that, when executed by the one or more processors, cause the system to perform the method as described in any one of claims 1 to 9.

Images