Ergonomic virtual shooting handle design method

Abstract

The invention relates to the technical field of virtual shooting equipment design, and discloses an ergonomic virtual shooting handle design method. The method comprises the following steps: acquiring hand operation data of a target user through a multi-source sensor, wherein the hand operation data comprises time sequence posture data, space pressure data and dynamic motion data; performing fusion preprocessing on the hand operation data to generate a standardized hand data set; extracting multi-dimensional features from the standardized hand data set, wherein the multi-dimensional features cover a posture stability index, a pressure distribution mode and a motion fluency coefficient; mode mining is carried out on the multi-dimensional features, and a typical operation mode of the target user is identified; according to the typical operation mode, a preset handle design knowledge base is retrieved, and a basic design parameter set is obtained; iteratively correcting the basic design parameter set to generate an optimized design parameter set; and controlling the three-dimensional modeling tool to generate a digital model of the virtual shooting handle based on the optimization design parameter set.

Description

Technical Field

[0001] This invention relates to the field of virtual shooting equipment design technology, specifically to an ergonomic virtual shooting handle design method. Background Technology

[0002] Current virtual shooting handle designs primarily employ standardized, universal models or experience-based design methods based on limited user surveys. Existing technologies rely heavily on static measurements or subjective interviews to capture hand operation characteristics, failing to obtain real biomechanical data from users during dynamic operations. The design process lacks the ability to quantitatively analyze multi-dimensional characteristics such as user hand posture stability, pressure distribution, and movement smoothness. Handle design parameter adjustments are often based on the designer's experience and judgment, making it difficult to establish an objective correlation between user operation characteristics and design parameters. Existing design methods cannot effectively identify typical operating patterns of different user groups, resulting in designs lacking specificity. The ergonomic design of virtual shooting handles requires addressing the quantitative challenges throughout the entire process, from user data collection to design parameter generation.

[0003] The fusion processing capabilities for multi-source sensor data are insufficient, and data from different sources suffer from temporal asynchrony and scale differences. Feature extraction methods are limited to a single dimension, making it difficult to fully reflect the complex characteristics of hand operations. The retrieval mechanism of the knowledge base is simple and cannot adapt to the specific needs of individual users. The parameter optimization process lacks an effective iterative correction mechanism, limiting the accuracy of the design results. The design of a virtual shooting handle requires a comprehensive design method that can achieve multi-source data fusion, multi-dimensional feature mining, and personalized parameter optimization. Summary of the Invention

[0004] The purpose of this invention is to provide an ergonomic virtual shooting handle design method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an ergonomic virtual shooting handle design method, the method comprising: Hand operation data of the target user is collected through multi-source sensors. The hand operation data includes temporal posture data, spatial pressure data, and dynamic motion data. The hand manipulation data is fused and preprocessed to generate a standardized hand data set, wherein the fusion and preprocessing includes data alignment, noise filtering and scale normalization; Multi-dimensional features are extracted from the standardized hand dataset, which include posture stability index, pressure distribution pattern and motion smoothness coefficient; Pattern mining is performed on the multi-dimensional features to identify the typical operation patterns of the target user; Based on the typical operation mode, a preset handle design knowledge base is retrieved to obtain a set of basic design parameters; The basic design parameter set is iteratively modified to generate an optimized design parameter set; Based on the optimized design parameter set, the 3D modeling tool is controlled to generate a digital model of the virtual shooting handle.

[0006] Preferably, the fusion preprocessing of the hand operation data includes: The temporal pose data, spatial pressure data, and dynamic motion data are synchronized and aligned according to timestamps to form a time-aligned data stream; A Kalman filter is applied to filter noise from the time-aligned data stream to obtain a smooth data stream; The smoothed data stream is subjected to min-max normalization to map all data values ​​to the interval between zero and one, generating a standardized set of hand data.

[0007] Preferably, extracting multi-dimensional features from the standardized hand dataset includes: Calculate the magnitude of change in time-series pose data per unit time and derive pose stability indices; Analyze the centroid and dispersion of spatial pressure data to determine the pressure distribution pattern; Differentiate the dynamic motion data to obtain acceleration and jerk, and calculate their variance as the motion smoothness coefficient.

[0008] Preferably, pattern mining of the multi-dimensional features includes: A self-organizing map neural network is used to reduce the dimensionality and cluster multi-dimensional features to form a feature map. High-density regions are identified on the feature map, and the feature combinations corresponding to the high-density regions are defined as typical operation modes. Profile analysis was used to verify the separation of typical operating patterns and ensure the saliency of the patterns.

[0009] Preferably, the preset handle design knowledge base includes: The handle design knowledge base stores the association rules between operating modes and design parameters; Use graph traversal algorithms to find rule nodes in association rules that match typical operation patterns; Extract the range of design parameter values ​​from the rule nodes and randomly sample to generate a set of basic design parameters.

[0010] Preferably, iteratively correcting the set of basic design parameters includes: Construct an objective function that measures the deviation between the set of design parameters and user comfort. Simulated annealing algorithm is used to randomly perturb and evaluate the set of basic design parameters; Based on the objective function value, the perturbation is accepted or rejected, and the system gradually converges to the optimal design parameter set.

[0011] Preferably, the objective function is constructed as follows: A comfort index is defined, which is based on an ergonomic model to calculate the degree of matching between the handle shape and the anatomical structure of the hand; Add constraints, including material strength and weight limits; The objective function is a weighted combination of comfort index and the degree of constraint violation.

[0012] Preferably, the acquisition of hand operation data of the target user through multi-source sensors includes: Use an optical motion capture system to collect temporal posture data of hand joints; Spatial pressure data at the hand contact point is collected using a flexible pressure sensor matrix; Dynamic motion data of hand movements are collected using an inertial measurement unit.

[0013] Preferably, the indices for calculating posture stability include: A sliding window is applied to the time-series pose data to calculate the standard deviation of the pose angles within the window; Exponential smoothing of the standard deviation yields the posture stability index; When the posture stability index is below the threshold, it is marked as a stable operating segment.

[0014] Preferably, controlling the generation of digital models by 3D modeling tools includes: Convert the set of optimized design parameters into script commands for 3D modeling software; Execute script commands to build the initial surface model of the handle; The initial surface model is parametrically adjusted and detailed, and the final digital model is output.

[0015] Compared with the prior art, the beneficial effects of the present invention are: Temporal posture data records the time-series changes in hand joint angles and positions, spatial pressure data measures the pressure distribution between the hand and the contact surface, and dynamic motion data captures the hand's movement trajectory and velocity characteristics. This multi-source synchronous data acquisition method overcomes the limited representational capabilities of single data types. The raw hand manipulation data undergoes fusion preprocessing, including data alignment, noise filtering, and scale normalization. Data alignment addresses the time synchronization issue from multiple sensor acquisitions, noise filtering eliminates environmental interference and measurement errors, and scale normalization eliminates the influence of individual hand size differences. The preprocessed, standardized hand dataset provides a high-quality data foundation for subsequent feature extraction.

[0016] Multi-dimensional features were extracted from a standardized hand dataset, encompassing postural stability indices, pressure distribution patterns, and motion smoothness coefficients. Postural stability indices quantify the degree of hand tremor and the ability to maintain hand position during operation; pressure distribution patterns analyze the magnitude and variation of pressure in different areas of the hand; and motion smoothness coefficients assess the continuity and coordination of hand movements. These features comprehensively describe users' operating habits and comfort needs from various perspectives. Pattern mining was performed on these multi-dimensional features to identify typical operating patterns of target users. Through cluster analysis and pattern recognition algorithms, stable behavioral characteristics of users in different operating scenarios were discovered, and representative operating patterns were extracted. These typical operating patterns reflect users' core needs and usage habits, providing a basis for personalized design.

[0017] The design knowledge base stores validated handle design rules and parameter mapping relationships, enabling the translation of user operation modes into specific design parameters. The basic design parameter set is iteratively refined, and parameter values ​​are continuously optimized through simulation analysis and user feedback to generate an optimized design parameter set that better meets user needs. Based on this optimized design parameter set, a 3D modeling tool is used to generate a digital model of the virtual shooting handle, realizing a digital design process from data to physical object. This entire method, through the synergistic effect of multi-source data acquisition, feature extraction, pattern recognition, and parameter optimization, achieves personalized ergonomic design of the virtual shooting handle, significantly improving product comfort and usability. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the ergonomic virtual shooting handle design method described in this invention. Figure 2 A flowchart for preprocessing hand manipulation data fusion; Figure 3 A flowchart for extracting multi-dimensional features from a standardized hand dataset; Figure 4 The result of pattern mining from a self-organizing map neural network is shown in the figure. Figure 5 The result of the annealing algorithm optimization process is shown in the figure. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1This invention provides an ergonomic virtual shooting handle design method. The method includes: collecting hand operation data of a target user in a simulated operation scenario by deploying a multi-source sensor system. This data comprehensively covers the temporal posture data, spatial pressure data, and dynamic motion data of the hand during operation. The raw hand operation data from different sensors undergoes fusion preprocessing. This process aims to solve the time synchronization problem between data, eliminate random noise interference, and unify the data scale, thereby generating a high-quality, comparable, standardized hand data set. From the standardized data set, multi-dimensional features reflecting user operation characteristics are systematically extracted. These features mainly include posture stability indices that quantify posture stability, spatial pressure distribution patterns describing grip force, and motion smoothness coefficients that evaluate the smoothness of movement. Based on the extracted multi-dimensional features, an unsupervised learning algorithm is used for deep pattern mining to identify the most frequent and representative typical operation patterns of the target user. The identified typical operation patterns are used as query conditions to retrieve a preset handle design knowledge base. This knowledge base stores a large number of verified mapping relationships between operation patterns and handle design parameters. By searching, a set of basic design parameters matching the user's patterns can be obtained. After obtaining the basic design parameters, they are not used directly. Instead, they are used as a starting point for iterative refinement through optimization algorithms to further approach the optimal ergonomic design goal, ultimately generating an optimized set of design parameters. This optimized set of design parameters is then used as input to drive the parametric modeling engine of the 3D modeling tool, automatically constructing and outputting a precise digital model of the virtual shooting handle.

[0021] Example 1: See Figure 2In practice, the process of acquiring hand operation data from multiple sensors needs to be conducted in a controlled simulated shooting environment. The target user is required to hold a standard-shaped simulated handle and perform a series of virtual shooting tasks, such as panning, rotating, zooming, and tracking and locking onto virtual targets. The optical motion capture system consists of more than eight high-speed infrared cameras, which are arranged in a circular array around the operating space and, after calibration, form a unified three-dimensional measurement field. A total of fifteen highly reflective markers are affixed to the back of the target user's right hand, wrist, and key joints of the index, thumb, and middle fingers. The optical motion capture system continuously captures the three-dimensional coordinate data of these markers at a sampling frequency of two hundred frames per second, thereby generating temporal posture data that accurately describes more than twenty degrees of freedom of hand movement. A flexible pressure sensor matrix is ​​integrated into the inside of an elastic fabric data glove, which fits snugly against the user's hand. The matrix contains 128 independent pressure sensing units distributed in a matrix across the palm, fingertips, and thumb area. It outputs pressure distribution information across the entire hand contact surface—spatial pressure data—at a frequency of 100 times per second. A nine-axis inertial measurement unit (IMU) is securely attached to the back of the hand with straps. This IMU measures the hand's angular velocity, linear acceleration, and geomagnetic field strength along three axes in real time, outputting this dynamic motion data at a frequency of 400 times per second. The three sensor systems are hardware synchronized via a central synchronizer, ensuring all data streams have consistent and accurate timestamps.

[0022] In practice, fusing and preprocessing the collected raw hand manipulation data to generate a standardized hand data set is the foundation for subsequent analysis. The data alignment operation first strictly aligns the temporal posture data generated by the optical motion capture system, the spatial pressure data generated by the flexible pressure sensor matrix, and the dynamic motion data generated by the inertial measurement unit, based on the timestamps allocated by the central synchronizer. Since the original sampling frequencies of the optical motion capture system, the flexible pressure sensor matrix, and the inertial measurement unit differ, a cubic spline interpolation algorithm is used to resample all data streams to a unified time series of 200 frames per second, forming a complete and synchronized time-aligned data stream. Next, a Kalman filter is applied to the time-aligned data stream for noise filtering. For the temporal posture data, the state variables of the Kalman filter are set to the position and velocity of each marker point; for the spatial pressure data, the state variables are set to the pressure values ​​and their rates of change of each pressure sensor unit; and for the dynamic motion data, the state variables are set to angular velocity and linear acceleration. Based on the system's motion model and observation model, the Kalman filter performs optimal estimation of the data at each moment, effectively suppressing random white noise and short-term disturbances during the measurement process, resulting in a smooth data stream. Then, the smoothed data stream is subjected to min-max normalization. For joint angle values ​​in time-series pose data, the minimum and maximum angle values ​​in the entire data sequence are identified; for pressure values ​​in spatial pressure data, the minimum and maximum pressure values ​​in the entire data sequence are identified; and for acceleration values ​​in dynamic motion data, the minimum and maximum acceleration values ​​in the entire data sequence are identified. Each data point is linearly mapped to the interval between zero and one according to a formula, ensuring that data with different physical meanings and dimensions have the same scale, ultimately generating a standardized hand data set that can be used for feature extraction.

[0023] In some embodiments, the optical motion capture system can be configured with a different number of cameras, such as twelve or sixteen, as long as they can cover the entire range of hand movements of the target user without obstruction. The sampling frequency of the high-speed infrared camera can also be adjusted, for example, using a frequency of one hundred frames per second or three hundred frames per second, the key being the ability to capture subtle changes in hand movements. The number and placement of markers can be optimized based on a hand anatomical model, but must cover key motion nodes such as the wrist joint, metacarpophalangeal joints, and interphalangeal joints.

[0024] In some embodiments, the number and distribution density of sensing units in the flexible pressure sensor matrix can be varied, for example, using sixty-four or two hundred and fifty-six sensing units. The sampling frequency of the flexible pressure sensor matrix can also be set to fifty times per second or two hundred times per second; the core requirement is the ability to distinguish dynamic changes in pressure distribution. The material of the data gloves needs to have good breathability and elasticity to ensure the comfort of the target user during long-term data acquisition, avoiding the impact of discomfort on the accuracy of the operational data.

[0025] It is understandable that the model and precision of the inertial measurement unit (IMU) can be selected, for example, a commercial-grade six-axis IMU or a high-precision industrial-grade nine-axis IMU. The fixing method of the IMU must ensure no relative slippage between it and the hand to avoid introducing additional motion noise. The synchronization accuracy of the central synchronizer needs to reach the microsecond level to ensure that the time alignment error between different sensor data is negligible. The interpolation algorithm used in the data alignment process is not limited to cubic spline interpolation; linear interpolation algorithms can also be used when the data changes relatively smoothly. The parameters of the Kalman filter, including the process noise covariance and the measurement noise covariance, need to be pre-calibrated according to the sensor characteristics before preprocessing. The minimum and maximum values ​​used for min-max normalization are calculated based on a single complete data acquisition session; the normalization references between different sessions are independent.

[0026] Optionally, after generating a standardized hand dataset, a quality check can be performed. For example, the signal-to-noise ratio (SNR) of each sensor channel can be calculated. If the SNR of a channel is lower than a preset threshold, the data for that channel is marked as invalid and excluded from subsequent feature extraction. This quality check mechanism can further improve the reliability of the input data. The fusion preprocessing can be performed in real-time or offline after data acquisition. Real-time processing requires sufficient throughput from the computing platform, while offline processing allows for more complex calculations and multiple iterative optimizations. The choice between real-time and offline modes depends on the specific application scenario and computing resource limitations.

[0027] Example 2: See Figure 3In practice, extracting multi-dimensional features from a standardized hand dataset is a systematic computational process aimed at transforming raw data into quantitative indicators with clear ergonomic significance. For time-series posture data within the standardized hand dataset, extracting posture stability indices is a core step. These indices quantify the hand's ability to maintain a specific posture during operation. The computation begins with applying a fixed-length sliding window, 500 milliseconds wide, sliding across the time-series posture data in 100-millisecond increments. Within each sliding window, for twenty key hand joint angles, the standard deviation of each joint angle across all sampling points within the window is calculated. The arithmetic mean of the twenty standard deviations is the initial posture fluctuation value for that window. Then, the resulting sequence of initial posture fluctuation values ​​is exponentially smoothed using a single exponential smoothing model with a smoothing coefficient of 0.3. This exponential smoothing eliminates random fluctuations, resulting in a smoothed posture stability index sequence. The posture stability index is a dimensionless value; a lower value indicates greater posture stability. When a value in the posture stability index sequence is lower than the preset stability threshold of 0.05, the system automatically marks the corresponding time period as a stable operation segment. The stable operation segment is an important data basis for identifying typical static grip postures.

[0028] In practical implementation, extracting pressure distribution patterns from spatial pressure data in a standardized hand dataset requires analyzing the statistical characteristics of pressure on the contact surface. Pressure distribution patterns are described by two main features: the distribution centroid and the degree of dispersion. The centroid is calculated by treating the flexible pressure sensor matrix as a two-dimensional plane. The coordinates of each pressure sensing unit are known, and the pressure value of each unit is used as a weight to calculate the weighted average center coordinates of the entire pressure distribution map. The centroid coordinates reflect the location of the main pressure-bearing areas. The degree of dispersion is calculated using the information entropy of the pressure distribution. First, the pressure value of each sensing unit is normalized to its proportion of the total pressure. Then, the uncertainty of the entire distribution is calculated using the information entropy formula. A higher entropy value indicates a more uniform and dispersed pressure distribution, while a lower entropy value indicates that the pressure is concentrated in a small area. Combining the location coordinates of the centroid and the entropy value of the dispersion, several typical pressure distribution patterns can be defined. For example, a pattern with the centroid located in the palm region and a lower entropy value is defined as "palm-concentrated," while a pattern with the centroid located in the finger root region and a higher entropy value is defined as "finger root dispersed."

[0029] In practical implementation, extracting the motion smoothness coefficient from dynamic motion data in a standardized hand dataset requires analyzing the dynamic characteristics of the motion. The motion smoothness coefficient is obtained by calculating the variances of acceleration and jerk. First, the hand motion velocity signal in the dynamic motion data is numerically differentiated, and the instantaneous acceleration is calculated using the central difference method. Then, the acceleration signal is numerically differentiated again to obtain the jerk signal. Jerk is the rate of change of acceleration, directly reflecting whether there are abrupt impacts during the motion. Within a sliding window with the same parameters as the posture stability index, the variances of acceleration and jerk are calculated separately for the three axes. Then, the variances of the three axes are summed to obtain the combined total variances of acceleration and jerk. The motion smoothness coefficient is ultimately defined as the weighted harmonic mean of the total variances of acceleration and jerk, with weighting coefficients of 0.7 and 0.3, respectively. The motion smoothness coefficient is also a dimensionless value; the lower the value, the smoother the motion.

[0030] In some embodiments, the width of the sliding window can be adjusted according to the actual application scenario. For example, in scenarios requiring the capture of faster posture changes, the window width can be set to 300 milliseconds; in scenarios emphasizing long-term stability assessment, the window width can be set to 800 milliseconds. The sliding step size can also be set independently; a smaller step size, such as 50 milliseconds, can improve temporal resolution but increase computational load. The threshold of 0.05 for the posture stability index is an empirical value and can be fine-tuned according to the operating characteristics of different user groups. The analysis of pressure distribution patterns can introduce more feature dimensions, such as the symmetry index of the pressure distribution map, the size and shape factor of the main pressure areas, etc. The calculation of the distribution centroid can adopt different weighting strategies, such as considering only sensing units with pressure values ​​higher than the average. The measure of dispersion can also use concentration indicators from economics, such as the Gini coefficient or the Herfindahl-Hirschman index, instead of information entropy.

[0031] It is understandable that in the calculation of motion smoothness coefficients, the differential algorithm can choose forward differencing or backward differencing, but the central difference method has second-order accuracy and smaller error. The weighting method for the total variance of acceleration and the total variance of jerk can be adjusted, for example, by using an arithmetic mean or a geometric mean. The weighting coefficients of 0.7 and 0.3 for the motion smoothness coefficient reflect the relative importance attached to acceleration stability and acceleration smoothness; this ratio can be adjusted according to the requirements of motion stability for specific operational tasks. The extraction order of multi-dimensional features can be parallelized, that is, the calculation of posture stability index, pressure distribution pattern, and motion smoothness coefficient can be executed simultaneously in different computing threads or processor cores to improve overall processing efficiency. The output of each feature extraction module is time-series data, which needs to be strictly synchronized with the original data for subsequent joint pattern analysis.

[0032] Optionally, an outlier detection mechanism can be introduced during feature extraction. For example, if the variance of data within a sliding window suddenly increases sharply, exceeding the normal range by more than three standard deviations, the window is marked as an outlier, and its feature values ​​are excluded from subsequent analysis. The extracted multi-dimensional features can be visualized. For instance, posture stability indices, pressure distribution pattern classification results, and motion smoothness coefficients can be simultaneously displayed on a single chart along a time axis, allowing designers to intuitively understand user operation characteristics. The visualization interface can also allow interactive selection of time intervals for detailed analysis.

[0033] Example 3: In specific implementation, the core task of pattern mining for multi-dimensional features is to use a self-organizing map neural network (SAMR) to reduce the dimensionality and cluster high-dimensional feature data. The SAMR structure includes an input layer and a two-dimensional competition layer. The number of neurons in the input layer is strictly consistent with the number of dimensions of the multi-dimensional features. For example, if the input feature vector contains five dimensions: pose stability index, pressure distribution centroid x-coordinate, pressure distribution centroid y-coordinate, pressure distribution information entropy, and motion smoothness coefficient, then the input layer has five neurons. The two-dimensional competition layer is usually set as a rectangular grid, for example, 10 x 10, totaling one hundred neurons. Each neuron is fully connected to the input layer through a weight vector, which also has a five-dimensional dimension. The training process of the SAMR adopts an unsupervised competitive learning mechanism. The standardized multi-dimensional feature vector sequence is input into the SAMR one by one. For each input vector, the SAMR calculates the Euclidean distance between it and the weight vectors of all neurons in the competition layer, and the neuron with the smallest distance is determined as the winning neuron. Then, based on the position of the winning neuron, the weight vectors of the winning neuron and its topological neighbors are updated to make them closer to the current input vector. The neighborhood function is usually a Gaussian function, and the neighborhood radius gradually decreases with the increase of training iterations. After a sufficient number of iterations, the weight vectors of the neurons in the competitive layer will adaptively adjust so that adjacent neurons in the topological structure represent similar feature patterns in the input space, ultimately forming a feature map on a two-dimensional grid.

[0034] Identifying typical operating patterns on a feature map requires analyzing the activation frequencies of neurons. After training, all multi-dimensional feature vectors are re-input into the trained self-organizing map neural network, recording the number of times each competing layer neuron is selected as the winning neuron, i.e., the activation frequency. The activation frequency values ​​are normalized and visualized on a two-dimensional grid, forming an activation frequency heatmap. On the activation frequency heatmap, high-density regions are represented by darker colors and continuous regions with significantly higher activation frequencies than surrounding neurons; each continuous high-density region corresponds to a dense cluster in the feature space. The set of multi-dimensional feature vectors responding to all neurons within each high-density region is defined as a typical operating pattern. To verify the statistical significance of the identified typical operating patterns, contour analysis is performed to calculate the contour coefficient of each feature vector. The contour coefficient measures the similarity of samples within the same cluster and the difference from samples in the nearest neighbor cluster. The formula for calculating the contour coefficient is:

[0035] in: Indicates the first The silhouette coefficients of each eigenvector. Indicates the first The average distance from a vector to all other vectors in its cluster. Indicates the first The average distance from each vector to all vectors in its nearest neighbor cluster. The average of the silhouette coefficients of all feature vectors is used as a measure of the overall separation of this typical operating mode. A silhouette coefficient average greater than 0.7 is generally considered to indicate that the mode has good separation.

[0036] In practice, retrieving the pre-defined handle design knowledge base is a key step in obtaining design parameters based on identified typical operating modes. The handle design knowledge base is constructed using an attribute graph model. Nodes in the graph represent different operating mode prototypes or design parameter constraints, and edges represent similarity relationships between modes or association rules between parameters. Each operating mode node stores a representative multi-dimensional feature vector and the range of handle design parameters for associated successful design cases, such as grip diameter range, button pressure range, and material friction coefficient range. The retrieval process uses a graph traversal algorithm, specifically a breadth-first search algorithm based on a priority queue, using the feature vector of the typical operating mode of the current target user as the query point. The algorithm calculates the Euclidean distance between the query point and the feature vector stored in each operating mode node in the knowledge base, and uses the distance value as the priority of that node; the smaller the distance, the higher the priority. The algorithm traverses from the node with the highest priority, checks its associated design parameter constraint nodes, and recursively visits other operating mode nodes connected to it in the graph by "highly similar" edges, forming a candidate node set. From the final set of candidate nodes, extract the value ranges of all associated design parameters. For example, the grip diameter might range from 30 mm to 35 mm, and the thumb rest angle from 15 degrees to 20 degrees. For each design parameter, randomly sample values ​​within its range according to a uniform distribution to generate a specific set of basic design parameters.

[0037] In some embodiments, the grid shape of the competitive layer of the self-organizing map neural network can be selected as a hexagonal grid, in which the neighborhood of each neuron is more symmetrically defined. The size of the competitive layer can be adjusted according to the amount of data and computational resources, for example, using a 15x15 grid to obtain finer pattern division, or a 5x5 grid to obtain a more macroscopic pattern overview. The learning rate and neighborhood radius decay strategy during training can adopt linear decay or exponential decay. The threshold for contour analysis can be adjusted; for less stringent applications, a contour coefficient average greater than 0.5 is considered acceptable for pattern separation. When identifying high-density regions, clustering algorithms such as DBSCAN can be combined to cluster the neurons in the competitive layer, identifying each cluster as a high-density region. This method has better recognition ability for irregularly shaped clusters. The graph model of the handle design knowledge base can introduce more complex edge types, such as "complementary pattern" edges, connecting operationally complementary pattern nodes for comprehensive reference during retrieval. The graph traversal algorithm can also adopt the A* algorithm, using the semantic similarity between nodes as a heuristic function to improve retrieval efficiency and accuracy.

[0038] It's understandable that the number of training epochs for a self-organizing map neural network (SAMR) needs to be sufficiently large to ensure network convergence, typically requiring thousands to tens of thousands of iterations. The quality of the feature map formation is closely related to the standardization of the input feature vectors; ensuring that each feature dimension has a similar numerical range is a crucial prerequisite. Contour analysis results can serve as feedback; if the contour coefficient of a identified pattern is too low, adjustments to the SAMR parameters or re-engineering of features can be considered. Building a controller design knowledge base is a continuous process of accumulation. As more user data and successful design cases are added, the nodes and edges in the knowledge base will continuously enrich, and the reliability of the retrieval results will gradually improve. The range of design parameters extracted from the knowledge base is a guiding interval based on historical experience, rather than an absolute constraint, leaving room for subsequent optimization and correction. After completing pattern mining and knowledge base retrieval, the feature vectors of typical operating patterns of the target user can be compared and analyzed with the feature vectors of the most matching prototype patterns in the knowledge base, generating a difference report that details the user's main characteristics in terms of posture stability, pressure distribution, and motion smoothness. This report can provide qualitative references for personalized adjustments to the controller design.

[0039] See Figure 4 A heatmap was used to present the activation frequency distribution of neurons in the competitive layer of the neural network. The color gradient from light yellow to dark red represents the change in neuron activation frequency from low to high, with dark red areas indicating that neurons at that location frequently become the best matching units, corresponding to areas with dense data points in the feature space. The position of each neuron on the two-dimensional grid reflects its topological relationship in the feature space, with adjacent neurons representing similar operating patterns. By analyzing the high-density areas on the heatmap, typical user operating patterns can be identified. These patterns reflect the user's behavioral habits in terms of stable posture, pressure distribution characteristics, and movement smoothness when holding the controller. The continuous blocks formed by high activation frequency areas correspond to dense clusters in the feature space, with each cluster representing a statistically significant operating pattern, providing data support for subsequent personalized controller design.

[0040] Example 4: In specific implementation, the core of iteratively correcting the set of basic design parameters is to construct an objective function that can quantitatively evaluate the merits of the handle design, and to use a simulated annealing algorithm for optimization search. Constructing the objective function first requires defining a comfort index. This comfort index is calculated based on a parameterized ergonomic model, which matches the digital 3D model of the handle with the anatomical structure of the target user's hand. The matching analysis process includes calculating the fit between the surface curvature of the handle and the arc of the user's palm arch, evaluating the coordination between the contour of the handle's grip area and the direction of the flexor tendons in the user's fingers, and analyzing the overlap between the button layout of the handle and the natural landing point of the user's fingertips. The fit, coordination, and overlap—these three sub-indicators—are weighted together to form the core of the comfort index, with weight coefficients determined based on biomechanical studies of the sensitivity of different areas of the hand. The objective function also needs to incorporate constraints, primarily covering material strength and weight limitations. The material strength constraint ensures that the maximum stress value of the handle under simulated normal use load is lower than the yield strength of the selected material, while the weight limitation constraint requires that the total mass of the handle does not exceed the upper limit of the ergonomically recommended long-term grip comfort weight. The objective function is ultimately a weighted combination of comfort index and constraint violation degree. Specifically, the comfort index is treated as a positive benefit term, and the material strength constraint violation amount and weight constraint violation amount are treated as penalty terms. A comprehensive evaluation value is formed by weighted summation.

[0041] In the specific implementation, referring to Table 1, the simulated annealing algorithm is used to iteratively correct the basic design parameter set. Simulated annealing is a probabilistic global optimization algorithm that simulates the solid annealing process. The optimization process uses the basic design parameter set retrieved from the handle design knowledge base as the initial solution, and sets a relatively high initial temperature parameter, typically several times the initial value of the objective function. In each iteration, the simulated annealing algorithm applies a random perturbation to the current design parameter set to generate a new candidate design parameter set. The random perturbation is performed by applying a Gaussian random shift to each design parameter within its value range, with the standard deviation of the shift proportional to the current temperature parameter. The objective function value corresponding to the newly generated candidate design parameter set is calculated and compared with the objective function value of the current design parameter set. If the objective function value of the candidate design parameter set is better, it is unconditionally accepted as the new current solution. If the objective function value of the candidate design parameter set is worse, the inferior solution is accepted with a probability according to the Metropolis criterion. The acceptance probability is calculated as an exponential function, where the exponent is the negative difference in objective function values ​​divided by the current temperature parameter. This mechanism of accepting inferior solutions with a certain probability allows the simulated annealing algorithm to escape local optima. As iterations proceed, the temperature parameter is gradually reduced according to a predetermined cooling schedule, for example, using an exponential cooling strategy where the temperature is multiplied by a cooling factor less than one after a certain number of iterations. The decreasing temperature gradually reduces the probability of accepting inferior solutions, and the simulated annealing algorithm's search behavior gradually shifts from extensive initial exploration to refined local search in later stages. When the temperature parameter drops below the termination temperature, or when multiple iterations fail to improve the optimal solution, the algorithm terminates and outputs the set of design parameters with the best objective function values ​​encountered throughout the process as the optimized design parameter set.

[0042] Table 1: Design Parameter Optimization Constraints

[0043] It is understandable that the weight coefficients of each term in the objective function have a decisive influence on the optimization direction. The setting of the weight coefficients needs to be combined with the design priority. For example, if the weight of comfort is set much higher than the penalty weight for constraint violation, the optimization process will tend to seek the ultimate comfort and may slightly violate some constraints. If the constraint penalty weight is set high, the optimization result will strictly satisfy the constraint conditions, but the comfort may not be optimal. The parameter settings of the simulated annealing algorithm, including the initial temperature, cooling rate, Markov chain length, etc., need to be adjusted according to the complexity of the specific problem and the computational resources. Too high an initial temperature will lead to slow convergence, while too low an initial temperature may cause premature entrapment in local optima.

[0044] In some embodiments, the constraints can be further extended, for example, by adding manufacturability constraints requiring that all cavities of the handle be accessible to the CNC milling cutter; or by adding cost constraints to limit the total volume of the handle to control material costs. The objective function can also be more complex, for example, by introducing a fatigue life estimation model, incorporating the predicted long-term fatigue as a negative indicator into the objective function.

[0045] Optionally, during the simulated annealing algorithm's operation, the objective function value and the trajectory of key design parameters for each generation of optimal solutions can be recorded, forming optimization history curves. These curves can be used to visually monitor the optimization process, determine whether the algorithm has converged normally, and provide a basis for necessary manual intervention. To accelerate the computation process, the calculation of comfort indices and constraint violations in the objective function can be replaced by surrogate models instead of complex finite element analysis or sophisticated ergonomic simulations. Surrogate models, such as Kriging models or radial basis function networks, can quickly predict the response values ​​corresponding to new design parameters at a lower computational cost by training on pre-generated sample points, thereby greatly improving the iteration speed of the simulated annealing algorithm, especially suitable for complex optimization problems with large parameter spaces.

[0046] See Figure 5 The blue curve shows the trend of the objective function value of the current solution as the number of iterations increases, while the red curve records the cost changes of the historical best solution. The objective function comprehensively considers the comfort index of the handle design and the degree to which various constraints are met, including multiple design parameters such as grip diameter, button pressure, material friction coefficient, handle length, and weight. In the early stages of optimization, the algorithm accepts more inferior solutions to expand the search range, which is reflected in the large fluctuations of the curve, helping to escape local optima. As the temperature parameter gradually decreases, the algorithm gradually converges to the vicinity of the global optimum, and the curve tends to stabilize. The optimization process considers the balance between ergonomic comfort and engineering constraints, and the final combination of design parameters maximizes user comfort while meeting hard constraints such as material strength and weight limitations. The entire optimization process demonstrates the effectiveness and robustness of the simulated annealing algorithm in complex multi-objective optimization problems.

[0047] Example 5: In specific implementation, generating a digital model of a virtual shooting handle using a 3D modeling tool based on an optimized design parameter set is a process of converting data parameters into geometric entities. The optimized design parameter set includes key values ​​such as grip diameter, length, button tilt angle, and surface continuity constraints. The implementation process begins with the development of a parameter parsing and command mapping interface. This interface reads each parameter in the optimized design parameter set and converts it into a sequence of script commands that can be understood by the application programming interface of the specific 3D modeling software. For example, if one parameter in the optimized design parameter set, "grip body diameter," is 35 mm, the interface program will convert it into a command to create a sketch circle, with the circle's diameter parameter value being 35. Another parameter, "thumb rest area tilt angle," is 15 degrees, which will be mapped to an operation instruction with a draft angle when rotating the sketch profile or executing an extrusion command. This mapping relationship is predefined in a configuration file, ensuring that each design parameter can accurately drive specific steps in the 3D modeling process. The generation of script commands follows a strict sequential logic, typically first constructing the core body outline, then adding local features, and finally performing Boolean operations and filleting.

[0048] In practice, executing script commands to construct the initial surface model of the handle is the core step in automated modeling. The generated script command sequence is sent to the application programming interface (API) of the 3D modeling software for execution. The execution process begins with creating a reference plane and a sketch. Based on the "total handle length" and "grip section eccentricity" parameters in the optimized design parameter set, the script commands draw a two-dimensional sketch defining the main contour of the handle on an orthogonal reference plane in 3D space. The sketch consists of a series of straight lines and spline curves, and the coordinates of the control points of the spline curves are controlled by parameters such as "palm fit curvature" in the optimized design parameter set. After completing the sketch, the script commands call the extrude or revolve commands to generate the initial solid model of the handle, forming a basic 3D solid. Next, the script continues to execute to add other features. For example, based on the "top recorded button boss height" and "button array spacing" parameters, commands for extrude bosses and cut holes are used to create button mounting positions on the main model. The initial surface model is a three-dimensional solid with a complete parametric feature history. Every dimension and geometric constraint in the model is associated with a parameter name in the set of optimization design parameters, which enables the model to be subsequently parametrically adjusted.

[0049] In practice, parametric adjustment and detail sculpting of the initial surface model are key steps in improving its ergonomics and aesthetics. Parametric adjustment primarily utilizes the parametric design function of 3D modeling software, achieved by modifying and optimizing the model-driven dimensions associated with the set of design parameters. For example, when checking the manufacturability of the initial model, if the wall thickness in certain areas is found to be below requirements, the "internal cavity shrinkage offset" parameter in the parametric association table can be adjusted. The system will automatically update the relevant sketch outline, thereby increasing the wall thickness without remodeling. Detail sculpting focuses on complex surfaces that are difficult to describe with simple parameters. Operators use the freeform sculpting tools provided by the 3D modeling software, referring to hand anatomy data, to fine-tune local areas of the initial surface model. For example, in the palm contact area, slight pushing and pulling of the control points slightly raises the center of the curved surface to better support the palm arch; on the button surface touched by the index fingertip, tiny concave surfaces are engraved to increase tactile feedback and anti-slip properties; at all edge junctions, different degrees of rounding commands are applied according to the "functional fillet radius" parameter to eliminate sharp edges and improve grip. The final output digital model is a complete 3D solid file containing all geometric information and good watertightness. The format can be STEP, IGES, or the native format of the modeling software, and can be directly used for 3D printing rapid prototyping or imported into a virtual imaging engine for assembly simulation.

[0050] In some embodiments, the selected 3D modeling tools can be computer-aided design software with powerful parametric capabilities, which offers mature support for parametric design and correlated updates. Script command generation can be aided by the software's built-in macro recording function; that is, the modeling process is manually executed once, and the software automatically records the corresponding application programming interface commands. The developer then parametrically modifies the recorded code, replacing fixed values ​​with variables read from the optimized design parameter set. Detail sculpting can be performed using virtual reality devices. Designers wear VR headsets and handheld controllers, intuitively grasping, rotating, and sculpting the digital model in a 3D virtual space. This immersive environment helps to more accurately assess the interaction between the handle's shape and the hand. Each modification record during the sculpting process can be saved as a non-parametric historical feature, coexisting with the parametric feature history within the model.

[0051] Optionally, an automated model checking process can be integrated before outputting the final digital model. This process uses scripts to call the analysis functions of the 3D modeling software to perform draft angle analysis, wall thickness checks, and minimum radius of curvature verification on the digital model, ensuring that the model meets the process requirements of injection molding or 3D printing. A report is generated based on the check results; if any non-conformities are found, the system can automatically feed back to the parametric adjustment stage, triggering iterative correction of the relevant parameters. The generated digital model can be automatically assigned preliminary material properties. Script commands can specify corresponding visual material properties, such as diffuse color and roughness value, for different surface areas of the model in the 3D modeling software based on the "surface treatment type" parameter in the optimized design parameter set. This ensures that the output digital model not only has geometric shape but also a more realistic visual appearance, facilitating effect review. The entire model generation process can be packaged into a batch processing task. Users only need to submit the optimized design parameter set file, and the system can automatically start the 3D modeling software in the background, execute the script, and complete the entire process of modeling, adjustment, checking, and output without manual intervention. This is particularly useful for scenarios that require generating a large number of design variations for automated simulation.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for designing an ergonomic virtual shooting handle, characterized in that, The method includes: Hand operation data of the target user is collected through multi-source sensors. The hand operation data includes temporal posture data, spatial pressure data, and dynamic motion data. The hand manipulation data is fused and preprocessed to generate a standardized hand data set, wherein the fusion and preprocessing includes data alignment, noise filtering and scale normalization; Multi-dimensional features are extracted from the standardized hand dataset, which include posture stability index, pressure distribution pattern and motion smoothness coefficient; Pattern mining is performed on the multi-dimensional features to identify the typical operation patterns of the target user; Based on the typical operation mode, a preset handle design knowledge base is retrieved to obtain a set of basic design parameters; The basic design parameter set is iteratively modified to generate an optimized design parameter set; Based on the optimized design parameter set, the 3D modeling tool is controlled to generate a digital model of the virtual shooting handle.

2. The ergonomic virtual shooting handle design method as described in claim 1, characterized in that, The fusion preprocessing of the hand operation data includes: The temporal pose data, spatial pressure data, and dynamic motion data are synchronized and aligned according to timestamps to form a time-aligned data stream; A Kalman filter is applied to filter noise from the time-aligned data stream to obtain a smooth data stream; The smoothed data stream is subjected to min-max normalization to map all data values ​​to the interval between zero and one, generating a standardized set of hand data.

3. The ergonomic virtual shooting handle design method as described in claim 2, characterized in that, Extracting multi-dimensional features from the standardized hand dataset includes: Calculate the magnitude of change in time-series pose data per unit time and derive pose stability indices; Analyze the centroid and dispersion of spatial pressure data to determine the pressure distribution pattern; Differentiate the dynamic motion data to obtain acceleration and jerk, and calculate their variance as the motion smoothness coefficient.

4. The ergonomic virtual shooting handle design method as described in claim 1, characterized in that, Pattern mining of the multi-dimensional features includes: A self-organizing map neural network is used to reduce the dimensionality and cluster multi-dimensional features to form a feature map. High-density regions are identified on the feature map, and the feature combinations corresponding to the high-density regions are defined as typical operation modes. Profile analysis was used to verify the separation of typical operating patterns and ensure the saliency of the patterns.

5. The ergonomic virtual shooting handle design method as described in claim 4, characterized in that, The preset handle design knowledge base includes: The handle design knowledge base stores the association rules between operating modes and design parameters; Use graph traversal algorithms to find rule nodes in association rules that match typical operation patterns; Extract the range of design parameter values ​​from the rule nodes and randomly sample to generate a set of basic design parameters.

6. The ergonomic virtual shooting handle design method as described in claim 5, characterized in that, Iterative correction of the aforementioned set of basic design parameters includes: Construct an objective function that measures the deviation between the set of design parameters and user comfort. Simulated annealing algorithm is used to randomly perturb and evaluate the set of basic design parameters; Based on the objective function value, the perturbation is accepted or rejected, and the system gradually converges to the optimal design parameter set.

7. The ergonomic virtual shooting handle design method as described in claim 6, characterized in that, The objective function is constructed by including: A comfort index is defined, which is based on an ergonomic model to calculate the degree of matching between the handle shape and the anatomical structure of the hand; Add constraints, including material strength and weight limits; The objective function is a weighted combination of comfort index and the degree of constraint violation.

8. The ergonomic virtual shooting handle design method as described in claim 1, characterized in that, Data collected from the target user's hand gestures using multi-source sensors includes: Use an optical motion capture system to collect temporal posture data of hand joints; Spatial pressure data at the hand contact point is collected using a flexible pressure sensor matrix; Dynamic motion data of hand movements are collected using an inertial measurement unit.

9. The ergonomic virtual shooting handle design method as described in claim 3, characterized in that, The calculation of posture stability indices includes: A sliding window is applied to the time-series pose data to calculate the standard deviation of the pose angles within the window; Exponential smoothing of the standard deviation yields the posture stability index; When the posture stability index is below the threshold, it is marked as a stable operating segment.

10. The ergonomic virtual shooting handle design method as described in claim 1, characterized in that, Controlling the generation of digital models using 3D modeling tools includes: Convert the set of optimized design parameters into script commands for 3D modeling software; Execute script commands to construct the initial surface model of the handle; The initial surface model is parametrically adjusted and detailed, and the final digital model is output.

Images