Intelligent dumbbell based on muscle-bone model action chain type correction

Abstract

The invention relates to the technical field of intelligent fitness and exercise rehabilitation, in particular to an intelligent dumbbell based on muscle-bone model action chain type correction. According to the intelligent dumbbell, the motion trail of the dumbbell during training of a user is collected through a built-in inertial measurement unit, human body parameters of the user are combined, the motion trail is input into a preset human body muscle-bone model, and a real-time angle sequence of main joints of the whole body of the user is reversely deduced through inverse kinematics calculation; the sequence is compared with a standard action library, the initial compensation joint with the earliest abnormal deviation in the action chain is accurately recognized, and then a targeted correction instruction is provided for a user in a voice mode or an augmented reality mode and the like. A chained correction mechanism can guide correction from a biomechanical source, so that the action normalization and the training efficiency are effectively improved; meanwhile, the functions of bilateral balance evaluation, dynamic resistance adjustment and long-term stability tracking jointly form a personalized training system integrating accurate evaluation, real-time feedback and active intervention.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent dumbbells, in particular to an intelligent dumbbell based on action chain correction of a muscle-bone model. BACKGROUND

[0002] With the popularization of the national fitness consciousness and the rapid development of the Internet of Things technology, intelligent fitness equipment is increasingly integrated into daily life. As a representative device in the field of strength training, intelligent dumbbells have evolved from simple weight-adjustable dumbbells to intelligent terminals integrating sensors, processors and communication modules. The core value lies in monitoring the training data of users and providing basic action specification feedback to make up for the lack of immediate feedback in traditional dumbbell training.

[0003] The prior art mainly relies on the inertial measurement unit (IMU) built in the intelligent dumbbell to collect motion data such as acceleration and angular velocity of the body in space, and uses a specific algorithm through the processor to restore the motion trajectory of the dumbbell, and then compares the real-time trajectory with the pre-stored standard action trajectory template. When the trajectory deviation exceeds the preset threshold, the system determines that the action is not standardized and issues a sound, light or vibration warning.

[0004] However, the main technical problem of the prior art is that it cannot effectively attribute the action error, that is, the system can detect macroscopic trajectory deviation, but due to the single analysis dimension, it cannot trace back to the specific root of the deviation on the human motion chain, for example, it cannot determine which joint of the ankle, knee, hip, etc. Abnormality is caused, so it cannot provide targeted correction guidance. SUMMARY

[0005] The present application provides an intelligent dumbbell based on action chain correction of a muscle-bone model, which solves the above-mentioned problems of the prior art.

[0006] To achieve the above-mentioned purpose, the embodiments of the present application disclose the following technical solutions:

[0007] The present application discloses an intelligent dumbbell based on action chain correction of a muscle-bone model, comprising a dumbbell body, an inertial measurement unit arranged in the dumbbell body, a processor, a memory and a feedback module; the memory stores a pre-trained human muscle-bone model, a standard action library and user's human parameters and a preset deviation threshold for different joints, and the standard action library contains a standard joint activity data sequence corresponding to at least one standard action;

[0008] The processor is configured to perform the following steps:

[0009] determining a target action to be performed by the user;

[0010] collecting real-time motion data of the dumbbell body during the execution of the target action by the user through the inertial measurement unit;

[0011] The real-time motion data is processed to obtain the real-time motion trajectory of the dumbbell body;

[0012] The real-time motion trajectory and the user's human body parameters are input into the human musculoskeletal model, which is used to simulate the dynamic relationship of various joints in the human body.

[0013] Inverse kinematics calculations are performed using a human musculoskeletal model to output a real-time joint angle data sequence representing the user's main joints during the execution of the target action;

[0014] Retrieve the standard joint movement data sequence corresponding to the target movement from the standard movement library;

[0015] The real-time joint angle data sequence is compared with the standard joint activity data sequence, and the angle deviation between the real-time joint angle data sequence and the standard joint activity data sequence at at least one joint is calculated.

[0016] Based on the calculated angle deviation, the initial compensatory joint that first shows an angle deviation exceeding the preset deviation threshold for the corresponding joint is identified.

[0017] Generate correction instructions for the initial compensatory joint;

[0018] The control feedback module outputs correction commands to the user.

[0019] This invention relates to a chain-like correction mechanism based on a musculoskeletal model. It fundamentally solves the technical bottleneck of existing smart fitness equipment, which can only monitor the trajectory of the end-effectors but cannot detect the root causes of user movement errors. This solution uses an inertial measurement unit to collect the dumbbell's movement trajectory and innovatively inputs the trajectory along with the user's body parameters into a pre-stored human musculoskeletal model. This model, as a digital biomechanical simulation system, can accurately deduce the real-time angle data of the major joints throughout the body during user movements through inverse kinematics calculations. This design achieves a leap from external equipment movement to internal joint activity, laying the foundation for in-depth movement analysis. Furthermore, the solution does not simply list all joint deviations but uses a dynamic time warping algorithm for precise comparison, identifying the initial compensatory joint where the abnormal deviation first appears in chronological order. This chain-like correction logic can accurately locate the biomechanical root causes of movement disorders, solving the problem of vague and general feedback in traditional methods. Based on this, targeted voice commands, such as stabilizing the scapula, can guide the user to correct the source of the problem. Furthermore, the dual dumbbell synergy design can detect and alert to bilateral muscle imbalances; it provides intuitive visual feedback when combined with augmented reality devices; and the dynamic resistance adjustment mechanism can proactively intervene before movement becomes uncontrollable. These designs collectively achieve a leap from single-track monitoring to whole-body compensatory chain monitoring, from general to precise root cause correction, and from information prompts to biomechanical assistance, ultimately effectively improving training quality and preventing sports injuries. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the process of an embodiment of the present invention;

[0021] Figure 2 This is a system block diagram according to an embodiment of the present invention;

[0022] Figure 3 This is an interaction diagram of the modules in an embodiment of the present invention. Detailed Implementation

[0023] Specific embodiments of the invention will now be described in detail. Although the invention is described in conjunction with these specific embodiments, it should be understood that the invention is not intended to be limited to these specific embodiments. Rather, these embodiments are intended to cover alternative, modified, or equivalent embodiments that may be included within the spirit and scope of the invention as defined by the claims. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. The invention may be practiced without some or all of these specific details. In other instances, well-known processes have not been described in detail so as not to unnecessarily obscure the invention.

[0024] When used in conjunction with the terms "comprising," "method comprising," or similar language in this specification and appended claims, the singular forms "a," "some," and "the" include plural references unless the context clearly indicates otherwise. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] Example 1

[0026] It includes a dumbbell body, an inertial measurement unit, a processor, a memory, and a feedback module installed inside the dumbbell body; the memory stores a pre-trained human musculoskeletal model, a standard movement library, the user's human body parameters, and preset deviation thresholds for different joints; the standard movement library contains at least one standard joint activity data sequence corresponding to a standard movement.

[0027] The processor is configured to perform the following steps:

[0028] Determine a target action that the user needs to perform;

[0029] The inertial measurement unit collects real-time motion data of the dumbbell body during the user's execution of the target action.

[0030] The real-time motion data is processed to obtain the real-time motion trajectory of the dumbbell body;

[0031] The real-time motion trajectory and the user's human body parameters are input into the human musculoskeletal model, which is used to simulate the dynamic relationship of various joints in the human body.

[0032] Inverse kinematics calculations are performed using a human musculoskeletal model to output a real-time joint angle data sequence representing the user's main joints during the execution of the target action;

[0033] Retrieve the standard joint movement data sequence corresponding to the target movement from the standard movement library;

[0034] The real-time joint angle data sequence is compared with the standard joint activity data sequence, and the angle deviation between the real-time joint angle data sequence and the standard joint activity data sequence at at least one joint is calculated.

[0035] Based on the calculated angle deviation, the initial compensatory joint that first shows an angle deviation exceeding the preset deviation threshold for the corresponding joint is identified.

[0036] Generate correction instructions for the initial compensatory joint;

[0037] The control feedback module outputs correction commands to the user.

[0038] In this embodiment, a touch screen is provided on the dumbbell body, and the touch screen is connected to the processor to display a standard exercise library and receive the user's selection input for the target exercise; a wireless communication module is provided in the dumbbell body; the processor establishes a connection with an external mobile terminal through the wireless communication module to receive the user's selection instruction for the target exercise from the external mobile terminal;

[0039] In practice, the user first selects a target exercise, such as a dumbbell bench press, from a library of standard exercises via a touchscreen display or a connected mobile app. The processor then loads the corresponding standard joint activity data sequence, which describes the ideal curves of the angles of major joints such as the shoulder, elbow, and wrist joints over time when completing a standard bench press.

[0040] The user begins performing a bench press. The inertial measurement unit inside the dumbbell, including a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, starts continuously acquiring linear acceleration and angular velocity data of the dumbbell in three-dimensional space at a high-frequency sampling rate. The processor performs sensor fusion processing on the linear acceleration and angular velocity data, employing a Kalman filter algorithm to eliminate noise and accurately estimate the real-time attitude and position of the dumbbell in space, thereby reconstructing its complete real-time motion trajectory.

[0041] The processor then inputs this real-time motion trajectory along with the user's pre-entered human body parameters into a pre-stored human musculoskeletal model. The human musculoskeletal model is a digital human simulator based on biomechanical principles. Through inverse kinematics calculations, the processor deduces the most likely angle combinations of all the user's joints to generate this specific dumbbell trajectory, thus obtaining a sequence of real-time joint angle data corresponding to the real-time motion trajectory.

[0042] Next, the processor compares the real-time joint angle data sequence with the standard joint activity data sequence, calculating the angle deviation of each joint in each phase of the movement. The key to chain-like correction is that it doesn't simply list all joints with deviations, but rather analyzes the order in which the deviations occur. The system identifies the joint that first exhibits an angle deviation exceeding a preset safety threshold for that joint and determines it as the initial compensating joint. In the bench press, insufficient scapular stability might cause the shoulder joint to abduct abnormally before the elbow joint. After identifying the initial compensating joint, the processor generates specific corrective instructions, such as "Please stabilize your scapula," and then broadcasts these instructions to the user in real time through a feedback module, such as a voice playback unit, guiding the user to correct the movement from its source.

[0043] This solution further proposes that the real-time motion trajectory and the user's human body parameters be input into the human musculoskeletal model in the following steps:

[0044] The user's body parameters include at least one or more combinations of height, arm length, leg length, and shoulder width;

[0045] The human musculoskeletal model is a mathematical model constructed based on the principle of multi-rigid-body dynamics. This human musculoskeletal model simplifies the human body into a chain system composed of multiple rigid segments connected by joints; the dumbbell body is defined as the end effector of the chain system.

[0046] The specific steps for performing inverse kinematic calculations using a human musculoskeletal model include:

[0047] The real-time motion trajectory of the dumbbell body is used as the trajectory constraint for the end effector of the chain system;

[0048] The length of each rigid segment in the chain system is determined by combining the user's human body parameters;

[0049] The orientation of the pelvis in space is used as the root node constraint;

[0050] An optimal combination of joint angles is obtained by using an iterative optimization algorithm. This optimal combination minimizes the error between the end effector trajectory and the real-time motion trajectory of the chain system. The optimal combination of joint angles constitutes the real-time joint angle data sequence.

[0051] In this embodiment, the human musculoskeletal model is constructed as a multi-rigid-body dynamic chain system. This model simplifies the human body into an open-loop or closed-loop kinematic chain composed of multiple rigid segments connected by ideal joints. For example, during dumbbell curls, a kinematic chain can be constructed from the hand to the dumbbell and then to the torso. The real-time trajectory of the dumbbell is used as the trajectory constraint for the end effector of this kinematic chain.

[0052] In practical implementation, inverse kinematics calculation uses the pelvis's posture and position in space as the root node constraint of the entire chain system, fixing the model's baseline. The goal is to find an optimal combination of joint angles that minimizes the error between the trajectory calculated by the musculoskeletal model's end effector and the actual dumbbell trajectory acquired and processed by the IMU. Since this problem typically lacks an analytical solution, iterative optimization algorithms, such as gradient descent or the Levenberg-Marquardt algorithm, are used for numerical solutions. The algorithm continuously adjusts the assumed angles of each joint in the model until the difference between the simulated trajectory and the actual trajectory is below a preset tolerance. The resulting combination of joint angles at this point is the inversely calculated real-time joint angle data sequence.

[0053] This solution further proposes comparing real-time joint angle data sequences with standard joint activity data sequences. The specific steps include:

[0054] Time alignment processing is performed on the real-time joint angle data sequence and the standard joint activity data sequence. The time alignment processing adopts a dynamic time warping algorithm to eliminate the influence of the difference in execution speed between the two sides on the comparison results.

[0055] Based on time alignment, the difference between the angle values ​​of each joint in the real-time joint angle data sequence and the standard angle values ​​of the corresponding joints in the standard joint activity data sequence is calculated under the same action phase of the target action, so as to obtain the real-time angle deviation of each joint at each moment.

[0056] In this embodiment, time alignment preprocessing must be performed before comparing the real-time joint angle data sequence with the standard joint activity data sequence. This is because even if the user's movements are very standard, the speed at which they complete a movement may differ from the preset rhythm of the standard sequence. Directly comparing at the same time point will lead to phase misalignment and misjudgment.

[0057] Dynamic time warping is a nonlinear warping technique that finds the optimal matching path between two time series of different lengths. In implementation, it aligns the two sequences at key motion events by stretching or compressing the time axis. For example, in bench press, it aligns the characteristic phase points such as the descent to the lowest point and the push-up to the highest point. Based on this time alignment, it calculates the difference between the actual angle values ​​and standard angle values ​​of each joint in the same motion phase, thus obtaining a true real-time angle deviation sequence after eliminating the influence of execution speed. This provides an accurate data foundation for subsequent deviation analysis and initial compensatory joint identification.

[0058] This solution further proposes a step to identify the initial compensatory joint where the angle deviation first exceeds a preset deviation threshold, specifically including:

[0059] Traverse the real-time joint angle data sequence after time alignment in chronological order;

[0060] For each joint, the time point at which its real-time angle deviation first exceeds the preset deviation threshold for that joint is detected;

[0061] By comparing the time points of all joints, the joint with the earliest time point is identified as the initial compensatory joint.

[0062] In this embodiment, after completing time alignment and calculating the real-time angle deviation sequence of each joint, the processor begins to scan the entire motion process from front to back in chronological order.

[0063] For each monitored joint, the system continuously monitors its angular deviation value and records the exact time when the deviation value first exceeds a preset deviation threshold. This deviation threshold is preset based on biomechanical knowledge and sports rehabilitation practices. The threshold may vary for different joints in different movements, defining the acceptable error range for that joint. For example, in squatting, the threshold for knee varus / valgus is set very low because even slight varus / valgus can increase the risk of injury.

[0064] After traversing the entire data sequence of all joints, the system compares the time points at which each joint first exceeds its limits. The joint with the earliest time point is identified as the initial compensatory joint. This determination is based on an important biomechanical principle: motor compensation typically exhibits a chain reaction characteristic; dysfunction or poor control in one segment will be the first to manifest abnormalities, subsequently triggering compensatory changes in subsequent segments. Therefore, correcting the joint that first exhibits abnormality is the most efficient and fundamental corrective strategy.

[0065] This solution further proposes that the feedback module includes a voice playback unit;

[0066] The step of generating correction instructions for the initial compensating joint is to generate text instructions containing the anatomical name of the initial compensating joint and the correct direction of movement.

[0067] The control feedback module outputs correction instructions to the user by converting text instructions into speech signals using text-to-speech technology and then broadcasting them through a speech playback unit.

[0068] In this embodiment, the feedback module uses a voice playback unit, which can provide real-time guidance without visual interference, allowing users to receive information while maintaining the continuity of their actions.

[0069] When generating correction instructions, the processor accesses a built-in database of anatomical terminology. The instructions are designed to be specific and actionable. First, it explicitly names the joint that needs correction, such as the scapula instead of vaguely saying "shoulder," or the sacroiliac joint instead of generally saying "lower back." Second, the instructions include clear directional guidance, such as scapular retraction and depression, slight elbow adduction, and slight posterior pelvic tilt.

[0070] The generated text instructions are converted into clear, natural speech signals by an integrated text-to-speech engine. The voice playback unit announces the commands at appropriate intervals or between repetitions, avoiding interruption during the exertion process. This precise voice feedback, based on anatomical knowledge, effectively guides users to engage the correct muscle groups, achieving re-education in neuromuscular control.

[0071] This solution further proposes that the feedback module also includes a vibration motor integrated inside the dumbbell handle;

[0072] The processor is also configured to perform the following additional steps:

[0073] Based on the real-time angle deviation between the real-time joint angle data sequence and the standard joint activity data sequence, as well as the identified initial compensatory joint information, a corresponding vibration control signal is generated.

[0074] Control the vibration motor to execute specific vibration patterns to provide users with instant, intuitive tactile feedback.

[0075] In this embodiment, in addition to voice feedback, a tactile vibration feedback mechanism is introduced to enrich the output channels of correction commands and adapt to different training environments and user preferences. Specifically, a miniature high-precision vibration motor is integrated and packaged inside the grip area of ​​the dumbbell body, and its vibration intensity and frequency can be precisely controlled by the pulse width modulation signal output by the processor.

[0076] The triggering logic for vibration feedback is deeply coupled with motion analysis. While the processor calculates the angular deviation of each joint in real time, it runs a vibration strategy engine. Based on the severity of the deviation, the joint where it occurs, and the trend of the deviation, this engine selects the most suitable pattern from a pre-set vibration pattern library and generates control signals.

[0077] In one embodiment, when the system detects that the angular deviation of a joint is continuously increasing but has not yet exceeded its preset threshold, the processor can control the vibration motor to output a series of gentle, intermittent short vibrations. This early warning allows the user to perceive the abnormal trend before the movement becomes completely deformed, thus giving them the opportunity to make proactive adjustments.

[0078] Precise positioning prompt: Once the system identifies the initial compensating joint, the processor generates a voice correction command while simultaneously driving a vibration motor to produce a characteristic set of strong vibrations, such as two consecutive rapid vibrations. This unique tactile signal helps users quickly focus their attention on the problem joint when they cannot listen to voice commands or need to maintain a high level of concentration.

[0079] Bilateral Balance Guidance: In dual-dumbbell collaborative training mode, if the system determines that the dominant side is compensating for the weaker side, the processor can control the vibration motor of the weaker dumbbell to emit continuous and stable vibrations, while the dominant dumbbell remains still or vibrates only slightly. This asymmetrical tactile experience can intuitively guide the user to adjust the balance of force exertion, promoting bilateral coordinated development.

[0080] Synergistic with dynamic resistance adjustment: When the resistance adjustment mechanism dynamically reduces the weight of the dumbbell to assist the user in correcting the trajectory, the processor can simultaneously trigger a noticeable, long vibration. This combined tactile and mechanical feedback enhances the user's perception, clearly indicating that the load has changed and urging them to focus on motion control.

[0081] The vibration patterns are highly configurable and can be personalized via a connected app. Key parameters include vibration intensity, frequency, duration, and timing. Different pattern codes are associated with specific corrective intentions, collectively constructing a silent yet highly effective language of body communication.

[0082] This solution further proposes that there is a pair of smart dumbbells, namely a first smart dumbbell and a second smart dumbbell. Both the first smart dumbbell and the second smart dumbbell contain independent inertial measurement units, processors, memory and feedback modules.

[0083] The processors of the first and second smart dumbbells establish a low-latency data connection through a wireless communication module to synchronize their real-time motion data.

[0084] The processor is also configured to perform the following steps:

[0085] Compare the joint range of motion or angular velocity of symmetrical joints in the real-time joint angle data sequence when the first and second smart dumbbells perform symmetrical movements;

[0086] When it is detected that the joint range of motion or angular velocity of one side is consistently better than that of the corresponding joint on the other side and the difference exceeds the asymmetry threshold stored in the memory, the joint with the consistent advantage in range of motion or angular velocity is the dominant side joint, and the other side is the weak side joint. It is determined that the dominant side joint has compensated for the weak side joint.

[0087] The step of generating correction instructions also includes generating instructions to prompt adjustments to the bilateral balance.

[0088] In this embodiment, the system is extended from single-dumbbell applications to dual-dumbbell collaborative scenarios to address the issue of bilateral body asymmetry. In practice, the user uses a pair of smart dumbbells for symmetrical training exercises such as bench presses and bicep curls. Both the first and second smart dumbbells contain complete processing units and wireless communication modules.

[0089] During the movement, the two dumbbells synchronize their inertial measurement unit data in real time via a low-latency wireless communication module. This allows the processor to obtain a complete, synchronized bilateral motion dataset. After performing inverse kinematics calculations, the system obtains a real-time joint angle data sequence for the symmetrical joints on both sides.

[0090] The processor specifically compares the range of motion or angular velocity of bilaterally symmetrical joints. For example, during the upward push phase of a shoulder press, it compares the abduction angular velocity of the left and right shoulder joints; or throughout the entire movement, it compares the maximum flexion angle of the left and right elbow joints. When the system detects that the range of motion or angular velocity of one joint is consistently better than the corresponding joint on the other side, and this difference exceeds a preset asymmetry threshold, it determines that the dominant side is compensating for the weaker side. At this point, the processor generates corrective instructions not only targeting specific joints but also including prompts to adjust bilateral balance, such as focusing on left-side force generation, reducing right-side compensation, or trying to reach the highest point with both arms simultaneously, guiding the user to consciously improve bilateral coordination.

[0091] This solution further proposes that the processor is also configured to periodically execute the following steps:

[0092] Record multiple real-time joint angle data sequences generated when the user performs the target action multiple times;

[0093] Calculate the dispersion among multiple real-time joint angle data sequences. The dispersion is used to quantify the repeatability stability of the user performing the target action.

[0094] Discreteness is stored and tracked as a neuromuscular control index.

[0095] In this embodiment, a long-term quantitative assessment of the stability of user movements is introduced. Specifically, the processor records multiple sets of real-time joint angle data sequences generated by the user when performing the same target movement multiple times on different training days.

[0096] To quantify the repetitive stability of the movements, the system calculates the dispersion between these sequences. A common implementation involves aligning the joint angle sequences of multiple movements over time and calculating the standard deviation of the angle values ​​for all sequences at each identical time point. The average or integral of these standard deviations at each time point can then be used to obtain a comprehensive dispersion index. This index is defined as the neuromuscular control index.

[0097] The smaller the index value, the more consistent the joint movement trajectory is when the user performs the action multiple times, indicating good repetition stability and reflecting the excellent control of the motor units by the nervous system. Conversely, a large index value indicates unstable movement and high variability. This index is stored and displayed to the user or coach along with historical data to objectively track training effects, fatigue levels, or rehabilitation progress, providing data support for adjusting training plans.

[0098] This solution further proposes that the smart dumbbell also includes an interface for communicating with an external augmented reality display device;

[0099] The processor is also configured to perform the following additional steps:

[0100] Generate visualization information for display on an augmented reality display device, the visualization information including at least: a dynamic image of a simplified human skeleton model driven by a real-time joint angle data sequence, and a dynamic image of a standard skeleton model corresponding to a standard joint activity data sequence superimposed on the dynamic image;

[0101] The simplified human skeleton model and the standard skeleton model are rendered with different colors or transparency to achieve overlapping comparison of motion trajectories.

[0102] In this embodiment, intuitive visual feedback of movement is provided through augmented reality technology. During implementation, the smart dumbbell establishes a communication connection with an external augmented reality display device via Bluetooth or a wireless network.

[0103] The processor generates two types of core visualization information. The first type drives a simplified human skeletal model. This model consists of simple lines and joints, and its movement is directly driven by a sequence of user joint angle data calculated in real time, thus reproducing the user's movements in three-dimensional space in real time.

[0104] The second type involves overlaying a dynamic image of a standard skeletal model corresponding to a standard joint motion data sequence onto the user's real-time model. To facilitate differentiation, the two models are rendered using different colors or transparency. For example, the user's real-time model is displayed with a semi-transparent blue, while the standard model is outlined with opaque green lines.

[0105] Users can see their real-time skeletal movement posture superimposed on a standard posture using AR devices. Any deviations are presented as intuitive geometric differences, such as differences in the elbow flexion angle or the shoulder extension range. This immersive visual feedback greatly lowers the barrier for users to understand the key points of the movement and identify their own mistakes, improving the efficiency of correction.

[0106] This solution further proposes that the standard joint activity data sequence is obtained by statistical modeling based on the standard movement data of professional athletes. The standard joint activity data sequence not only includes the mean curve of joint angles, but also the variance curve representing the safety range.

[0107] In this embodiment, the standard joint activity data sequence is not derived from a single idealized movement, but rather from the collection of joint angle data by performing three-dimensional motion capture on a large group of professional athletes performing specific movements.

[0108] Based on this massive amount of data, a standard sequence is generated using statistical modeling methods. This sequence not only includes the mean curve of joint angle changes over time, i.e., the most typical ideal movement pattern, but also includes a variance curve characterizing the safety range. Specifically, the variance curve can be represented as a standard deviation band or percentile band within a certain range above and below the mean curve.

[0109] In practical comparisons, the system not only checks whether the user's real-time joint angle sequence approximates the mean curve, but more importantly, it determines whether it consistently falls within the safety boundary defined by the variance curve. This safety boundary takes into account reasonable individual differences and minor variations in movement. As long as the user's movement trajectory falls within this safety range, even if there are slight differences from the mean curve, it is considered safe and effective. This approach is more inclusive and practical than using a single ideal curve, and it is also more in line with biomechanical principles.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation methods of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. An intelligent dumbbell based on myoskeletal model action chain correction, characterized in that, The dumbbell body, an inertial measurement unit arranged in the dumbbell body, a processor, a memory, and a feedback module are included; the memory stores a pre-trained human musculoskeletal model, a standard action library, and user's human parameters and a preset deviation threshold for different joints; the standard action library contains a standard joint activity data sequence corresponding to at least one standard action; The processor is configured to perform the following steps: determining a target action to be performed by the user; collecting real-time motion data of the dumbbell body during the execution of the target action by the user through the inertial measurement unit; processing the real-time motion data to obtain a real-time motion trajectory of the dumbbell body; inputting the real-time motion trajectory and the user's human parameters into the human musculoskeletal model, which is used to simulate the dynamics of each joint of the human body; performing inverse kinematics calculation through the human musculoskeletal model to output real-time joint angle data sequence representing the real-time joint angles of the user's main joints during the execution of the target action; calling the standard joint activity data sequence corresponding to the target action from the standard action library; comparing the real-time joint angle data sequence with the standard joint activity data sequence to calculate the angle deviation of the real-time joint angle data sequence and the standard joint activity data sequence at at least one joint; based on the calculated angle deviation, identifying the initial compensatory joint that first appears with an angle deviation exceeding the preset deviation threshold for the corresponding joint; generating a correction instruction for the initial compensatory joint; controlling the feedback module to output the correction instruction to the user.

2. The intelligent dumbbell based on the musculoskeletal model action chain correction according to claim 1, characterized in that, In the step of inputting the real-time motion trajectory and the user's human parameters into the human musculoskeletal model: The user's human parameters at least include one or more combinations of height, arm length, leg length, and shoulder width; The human musculoskeletal model is a mathematical model constructed based on the principle of multi-rigid-body dynamics, which simplifies the human body into a chain system connected by joints; The dumbbell body is defined as the end effector of the chain system; The step of performing inverse kinematics calculation through the human musculoskeletal model specifically includes: using the real-time motion trajectory of the dumbbell body as the trajectory constraint of the end effector of the chain system; determining the length of each rigid segment in the chain system in combination with the user's human parameters; using the posture of the pelvis in space as the root node constraint; solving a set of optimal joint angle combinations through an iterative optimization algorithm, which minimizes the error between the trajectory of the end effector of the chain system and the real-time motion trajectory, and the optimal joint angle combination constitutes the real-time joint angle data sequence.

3. The intelligent dumbbell based on the musculoskeletal model action chain correction according to claim 1, characterized in that, The specific steps of comparing the real-time joint angle data sequence with the standard joint activity data sequence include: performing time alignment processing on the real-time joint angle data sequence and the standard joint activity data sequence, which uses the dynamic time warping algorithm to eliminate the influence of execution speed difference on the comparison result; on the basis of time alignment, calculating the difference between the angle value of each joint in the real-time joint angle data sequence and the standard angle value of the corresponding joint in the standard joint activity data sequence at the same action phase of the target action to obtain the real-time angle deviation of each joint at each time.

4. The intelligent dumbbell based on the musculoskeletal model action chain correction according to claim 1, characterized in that, The step of identifying the initial compensatory joint whose angle deviation first exceeds the preset deviation threshold value specifically comprises: sequentially traversing the real-time joint angle data sequence after time alignment processing; for each joint, detecting the time point when its real-time angle deviation first exceeds the preset deviation threshold value for the joint; comparing the time points of all joints, and determining the joint with the earliest time point as the initial compensatory joint.

5. The intelligent dumbbell based on the musculoskeletal model action chain correction according to claim 1, characterized in that, The feedback module comprises a voice playing unit; The step of generating the correction instruction for the initial compensatory joint specifically comprises generating a text instruction containing the anatomical name of the initial compensatory joint and the correct movement direction; The step of controlling the feedback module to output the correction instruction to the user specifically comprises converting the text instruction into a voice signal through text-to-speech technology and playing the voice signal through the voice playing unit.

6. The intelligent dumbbell based on musculoskeletal model action chain correction according to claim 1, characterized in that, The number of the intelligent dumbbells is one pair, which are a first intelligent dumbbell and a second intelligent dumbbell. Both the first intelligent dumbbell and the second intelligent dumbbell comprise an independent inertial measurement unit, a processor, a memory and a feedback module. The processor of the first intelligent dumbbell and the processor of the second intelligent dumbbell establish a low-delay data connection through a wireless communication module for real-time synchronization of real-time movement data of both sides. The processor is further configured to perform the following steps: comparing the joint range of motion or angular velocity of the symmetrical joint in the real-time joint angle data sequence when the first intelligent dumbbell and the second intelligent dumbbell perform a symmetrical action; when it is detected that the joint range of motion or angular velocity of one side joint continuously outperforms that of the corresponding joint on the other side and the difference exceeds a pre-stored asymmetry threshold value in the memory, the side joint with the continuously superior joint range of motion or angular velocity is the dominant side joint, and the other side is the weak side joint, and it is determined that the dominant side joint compensates for the weak side joint; The step of generating the correction instruction further comprises generating an instruction for prompting adjustment of bilateral balance.

7. The intelligent dumbbell based on musculoskeletal model action chain correction according to claim 1, characterized in that, The processor is further configured to periodically perform the following steps: recording a plurality of real-time joint angle data sequences generated when the user performs the target action multiple times; calculating the dispersion between the plurality of real-time joint angle data sequences, the dispersion being used to quantify the repeat stability of the user performing the target action; storing and tracking the dispersion as a neuromuscular control index.

8. The intelligent dumbbell based on musculoskeletal model action chain correction according to claim 1, characterized in that, The intelligent dumbbell further comprises an interface for communicating with an external augmented reality display device; The processor is further configured to perform the following additional steps: generating visual information for display on the augmented reality display device, the visual information at least including a dynamic image of a simplified human skeletal model driven by the real-time joint angle data sequence, and a dynamic image of a standard skeletal model corresponding to a standard joint activity data sequence superimposed on the dynamic image; wherein the simplified human skeletal model and the standard skeletal model are rendered in different colors or transparencies to achieve superimposition and comparison of movement trajectories.

9. The intelligent dumbbell based on myo-skeletal model action chain correction of claim 1, wherein, The standard joint activity data sequence is modeled according to the standard action data statistics of a professional athlete group. The standard joint activity data sequence not only contains the mean curve of the joint angle, but also contains the variance curve representing the safe range.

Images