Analysis system based on intelligent Pick racket data acquisition

Abstract

The invention relates to the technical field of Pick balls, in particular to an analysis system based on intelligent Pick racket data acquisition, which is characterized in that racket swing deformation signals and ball hitting pressure data are synchronously acquired, and after quantitative indexes are extracted through frame dynamic deformation threshold filtering, coordinates of a sensing unit are corrected by adopting linear interpolation; performing weighted calculation on initial ball hitting coordinates through sweet area high-frequency data, optimizing an edge non-uniform pressure field in combination with K-means clustering, and generating a high-precision pressure distribution diagram; implementing a dormancy control simplified array according to the edge unit activation frequency and the information entropy, and iteratively correcting the coordinate deviation to output a stable hitting position; finally, deformation parameters are dynamically corrected through a sliding window averaging method, the sensing network is globally re-calibrated, lightweight entropy coding is started when nodes are reduced, and low-cost and high-precision positioning is achieved. According to the invention, positioning errors and power consumption are effectively reduced, multi-scene training requirements are met, accurate technical feedback is provided for players, and the method is suitable for industrial application of intelligent Pick rackets.

Description

Technical Field

[0001] This invention relates to the field of pickle technology, and more specifically to an analysis system based on data acquisition from a smart pickle racket. Background Technology

[0002] Against the backdrop of the rapid development of intelligent sports equipment, sports data acquisition and analysis systems are playing an increasingly prominent role as important tools for improving competitive performance and the scientific level of training. The intelligent Peak racket, through integrated sensors, acquires multi-dimensional data in real time, including swing motion, swing speed, hitting pressure, and hitting position, providing players with personalized technical feedback. This has become a key means of optimizing training strategies and improving competitive performance. This data-driven analysis approach helps identify deviations in the details of movement, supporting targeted technical adjustments.

[0003] Currently, most smart racket solutions use a fixed-layout sensor array to collect mechanical signals during the impact process and calculate the impact position using a preset coordinate system. However, in real-world high-intensity usage scenarios, the racket frame undergoes slight deformation due to repeated stress, causing a shift in the relative positions between sensors. This affects the accuracy of position calculations based on pressure differences and trigger timing. Furthermore, existing methods have relatively simplified compensation mechanisms for dynamic frame deformation, often failing to fully consider the nonlinear relationship between the degree of deformation and coordinate shift, leading to systematic biases in impact position determination. Since the analysis of pressure distribution highly depends on the accurate positioning of the contact point, position errors further propagate to the pressure analysis results, reducing the overall data reliability.

[0004] High-precision acquisition of the hitting position faces two interconnected challenges. Firstly, different areas of the racket face are used at significantly different frequencies. The sweet spot, as a high-frequency hitting area, requires higher resolution for pressure details; while the edge areas are used less frequently. A uniform distribution strategy in the sweet spot may result in insufficient information density, affecting the precision of feature extraction. Secondly, increasing the overall sensor density to improve sweet spot accuracy leads to increased costs and circuit complexity. More importantly, the structural deformation of the racket after continuous high-intensity hits alters the spatial relationships of the sensors, causing the calculation model based on the initial calibration coordinates to deviate from the actual state. For example, in intense competition, a hit that was originally located in the center of the sweet spot may be misjudged as being near the edge due to frame deformation, leading to incorrect technical feedback.

[0005] The aforementioned factors collectively constitute the technical challenge of achieving reliable ball positioning while balancing cost control and measurement accuracy. Therefore, under the dual constraints of racket dynamic deformation and uneven area usage, constructing an analysis system capable of adaptively adjusting sensor coordinates, optimizing data fusion strategies, and dynamically simplifying hardware configuration has become a key issue in improving the accuracy and practicality of data acquisition for the smart Peak racket. Summary of the Invention

[0006] The purpose of this invention is to address the aforementioned shortcomings in the prior art by providing an analysis system based on data acquisition from a smart Peak racket.

[0007] The objective of this invention is achieved through the following technical solution: an analysis system based on data acquisition from a smart Peak racket, comprising the following steps: S1. By using the strain sensor array and racket face pressure sensing layer embedded inside the racket frame, the frame deformation signal and pressure distribution data at the moment of impact are collected synchronously during the swing. The original deformation signal is filtered using a preset frame dynamic deformation threshold to extract the frame dynamic deformation and relative displacement change rate, and to generate a frame dynamic deformation index. S2. Based on the frame dynamic deformation index, the initial coordinates of each sensing unit of the racket pressure sensing layer are dynamically mapped. When the frame dynamic deformation index exceeds the preset frame dynamic deformation threshold, a linear interpolation algorithm is used to reconstruct the spatial position of the sensing unit of the racket pressure sensing layer between adjacent fixed reference points, and an adjusted set of racket pressure sensing unit coordinates is generated. S3. Based on the adjusted set of racket face pressure sensing unit coordinates, extract the sensing unit data of the racket face pressure sensing layer that is frequently triggered within the sweet spot area. Combine the pressure amplitude and trigger timestamp of each sensing unit in the racket face pressure sensing layer, and use a weighted average fusion strategy to calculate the pressure center position and generate the initial hitting coordinates. S4. Based on the initial hitting coordinates, identify the sensing units of the racket face pressure sensing layer that are triggered by low frequency in the adjacent edge region. If the pressure distribution in this region exhibits non-uniform gradient characteristics, then use the K-means clustering algorithm to group the pressure readings, reconstruct the local pressure field shape, and generate a pressure distribution map optimized based on racket face pressure sensing data. S5. Based on the pressure distribution map optimized by the surface pressure sensing data, the activation frequency and information contribution of the edge sensing units of the surface pressure sensing layer are statistically analyzed, adaptive sampling rules are set, periodic sleep control is implemented for low contribution units, and a simplified surface pressure sensing unit array configuration is generated. S6. Based on the simplified racket face pressure sensing unit array configuration, re-execute the ball hitting coordinate calculation process, compare the consistency of the trigger timing and the distribution of pressure difference between the old and new coordinates. If the deviation exceeds the allowable error bandwidth, return to step S2 to iteratively update the deformation compensation parameters until stable coordinates are output and the final ball hitting position coordinates are generated. S7. Based on the offset between the final hitting position coordinates and the initial calibration coordinates, combined with the sweet spot usage frequency statistics and edge detail retention index, the sliding window mean method is used to dynamically correct the frame dynamic deformation threshold and interpolation weight coefficient, and generate the deformation correction parameters for the next cycle. S8. Based on the deformation correction parameters of the next cycle, the coordinate mapping relationship of the entire sensor network is globally recalibrated. If the number of circuit nodes in the simplified array is reduced, lightweight data compression coding is enabled to entropy code the pressure distribution map to generate low-cost, high-precision positioning results.

[0008] The present invention is further configured such that the dynamic deformation index of the framework includes deformation amplitude detection, relative displacement rate calculation, and threshold exceedance determination; the adjusted racket face pressure sensing unit coordinate set includes reference point anchoring, linear interpolation weight allocation, and coordinate remapping; the preliminary hit coordinates include sweet spot data screening, pressure-time weighting factor setting, and centroid coordinate solution; the pressure distribution map optimized based on racket face pressure sensing data includes non-uniform pattern recognition, cluster center initialization, and local field reconstruction; the simplified racket face pressure sensing unit array configuration includes activation frequency statistics, information entropy evaluation, and sleep scheduling instruction generation; the final hit position coordinates include temporal consistency verification, pressure gradient matching, and iterative convergence determination; the deformation correction parameters for the next cycle include threshold sliding update, interpolation coefficient correction, and sweet spot weight enhancement; and the low-cost, high-precision positioning result includes node number compression ratio calculation, Huffman coding application, and positioning error backtracking verification.

[0009] The present invention is further configured such that the step of simultaneously collecting frame deformation signals and pressure distribution data at the moment of impact during the swing by means of a strain sensor array and a racket face pressure sensing layer embedded inside the racket frame, filtering the original deformation signals using a preset frame dynamic deformation threshold, extracting the frame dynamic deformation and relative displacement change rate, and generating a frame dynamic deformation quantitative index is specifically as follows: Four strain gauges are embedded at the connection between the racket handle and the frame, at the top of the frame and at symmetrical positions on both sides as fixed reference points. The remaining strain sensors are arranged at equal intervals along the inner side of the frame. The racket face pressure sensing layer is made of a flexible piezoresistive film, covering the entire racket face and divided into 64 sensing units of the racket face pressure sensing layer. The sensing unit of each racket face pressure sensing layer is connected to the central processing module through an independent wire. The acquired raw deformation signal is filtered by a second-order Butterworth low-pass filter to remove high-frequency vibration noise, retaining the structural response signal in the 0.1-10 Hz frequency band of the field of puck ball technology. The length change rate of each strain sensor within adjacent sampling periods is calculated, and the two-dimensional displacement vector of each sensing unit relative to the initial installation position is calculated by combining the distance constraints between reference points. The Euclidean norm of the displacement vector is compared with the preset frame dynamic deformation threshold. If it exceeds the threshold, it is marked as a valid deformation event, and a quantitative index containing deformation amplitude, direction and duration is generated.

[0010] The present invention is further configured such that, based on the frame dynamic deformation index, the initial coordinates of each sensing unit of the racket face pressure sensing layer are dynamically mapped, and when the frame dynamic deformation index exceeds a preset frame dynamic deformation threshold, a linear interpolation algorithm is used to reconstruct the spatial position of the sensing units of the racket face pressure sensing layer between adjacent fixed reference points, generating an adjusted set of racket face pressure sensing unit coordinates. Establish the sensor unit coordinate matrix during initial calibration, where the sensor unit coordinates of each pressure sensing layer on the striking surface are calibrated by a laser positioning instrument and stored in non-volatile memory; Based on the displacement vector marked in the deformation event, determine the reference point interval to which the affected sensing unit belongs; A one-dimensional linear interpolation function is constructed between every two adjacent fixed reference points, with the actual displacement of the reference points as the endpoint value, and the displacement of the intermediate sensing unit is allocated according to the original ratio. The displacement is superimposed on the initial coordinates to generate a dynamically updated set of coordinates, which is then synchronously written into the coordinate mapping table for subsequent calculations.

[0011] The present invention is further configured such that, based on the adjusted set of racket face pressure sensing unit coordinates, the sensor unit data of the racket face pressure sensing layer triggered at high frequencies within the sweet spot area is extracted, and the pressure amplitude and trigger timestamp of each sensing unit in the racket face pressure sensing layer are combined, and a weighted average fusion strategy is used to calculate the pressure center position to generate preliminary hitting coordinates, the specific steps are as follows: The sweet spot is defined as the 30% area of ​​the center of the racket face, corresponding to sensor unit numbers 18-47; Filter the set of cells in the ball-hitting event where the pressure value is greater than 5N and the trigger time difference is less than 2ms; Each effective unit is assigned a weight, which is determined by the pressure amplitude and the reciprocal of the time delay, and the formula is w_i=P_i / (1+Δt_i); Calculate the weighted center coordinates x=Σ(w_i·x_i) / Σw_i, y=Σ(w_i·y_i) / Σw_i, and output the initial hitting coordinates.

[0012] The present invention is further configured such that, based on the preliminary hit coordinates, the sensing unit of the racket face pressure sensing layer that identifies the low-frequency triggered racket face pressure sensing layer in the vicinity of the edge region, if the pressure distribution in that region exhibits non-uniform gradient characteristics, uses the K-means clustering algorithm to group the pressure readings, reconstruct the local pressure field morphology, and generate a pressure distribution map optimized based on the racket face pressure sensing data, specifically as follows: A neighborhood with a radius of 15mm is defined centered on the initial ball-hitting coordinates, and the pressure values ​​of all sensing units within this area are extracted. Calculate the pressure gradient vector. If the angle between the maximum gradient direction and the radial direction is greater than 30 degrees, it is determined to be a non-uniform mode. Initialize the number of clusters k=2, and iteratively update the cluster centers with the pressure value as the feature dimension until convergence; By treating units within the same cluster as continuous pressure regions and using bilinear interpolation to fill the gaps, a high-resolution pressure distribution map is generated.

[0013] The present invention is further configured such that, based on the pressure distribution map optimized from the racket face pressure sensing data, the activation frequency and information contribution of the edge sensing units of the racket face pressure sensing layer are statistically analyzed, adaptive sampling rules are set, and periodic sleep control is implemented for low contribution units to generate a simplified racket face pressure sensing unit array configuration. Record the number of times the edge sensing unit of each racket face pressure sensing layer is triggered in the past 100 shots, and calculate the activation frequency f_i; Calculate the information entropy H_i=-Σp_jlog2p_j of the sensing unit in the pressure distribution reconstruction of each pressure sensing layer, where p_j is the probability of the unit's contribution to the clustering result; Set the sleep conditions: f_i<0.05 and H_i<0.3. If the conditions are met, the sensing unit of the pressure sensing layer will turn off the power supply in the next sampling cycle. Generate configuration instructions containing active and sleep states, and control multiple analog switches to switch sensing channels via GPIO interface.

[0014] The present invention is further configured such that, based on the simplified racket face pressure sensing unit array configuration, the hit coordinate calculation process is re-executed, and the consistency of the trigger timing and the pressure difference distribution between the old and new coordinates are compared. If the deviation exceeds the allowable error bandwidth, the process returns to step S2 to iteratively update the deformation compensation parameters until stable coordinates are output, and the final hit position coordinates are generated. The specific steps are as follows: Repeat step S3 under the simplified configuration to obtain the new hitting coordinates (x', y'); Calculate the time synchronization error between the original coordinates (x, y) and the new coordinates Δt = |t - t'|. If Δt > 1ms or the Euclidean distance d = √[(x - x')² + (y - y')²] > 3mm, then it is determined to be a mismatch. If there is a mismatch, the deformation threshold is lowered by 5%, and steps S2 to S5 are re-executed, for a maximum of 3 iterations. If the process still fails to converge after the third iteration, the last result is retained as the final shot position coordinates.

[0015] The present invention is further configured such that, based on the offset between the final hitting position coordinates and the initial calibration coordinates, combined with the sweet spot usage frequency statistics and edge detail retention index, the step of dynamically correcting the frame dynamic deformation threshold and interpolation weight coefficients using the sliding window mean method to generate the deformation correction parameters for the next cycle is specifically as follows: The sweet spot hit rate r_sweet in the last 50 hits is calculated. If r_sweet > 0.8, the sweet spot interpolation weight coefficient α = α × 1.1 is increased. Calculate the pressure detail retention rate of the edge region d_edge = actual number of clusters / theoretical maximum number of clusters. If d_edge < 0.6, then expand the non-uniform mode judgment angle to 40 degrees. A moving average filter with a window length of 20 is applied to the deformation threshold. The new threshold = 0.9 × old threshold + 0.1 × current average deformation. The updated α, judgment angle, and threshold are written into the parameter register as the basis for the next cycle of correction.

[0016] The present invention is further configured such that, based on the deformation correction parameters of the next cycle, the coordinate mapping relationship of the entire sensor network is globally recalibrated; if the number of circuit nodes in the simplified array is reduced, lightweight data compression coding is enabled to perform entropy coding processing on the pressure distribution map to generate low-cost, high-precision positioning results. The specific steps are as follows: Reconstruct the dynamic coordinate mapping table of all sensing units based on the new parameters, overwriting the original stored content; The total number of active units N_active in the current pressure sensing layer is detected. If N_active ≤ 48, the Huffman coding module is activated. Using the pressure value quantification level as the symbol source, a frequency statistics table is constructed to generate a variable-length coding dictionary; The compressed pressure data and the final ball-hitting coordinates are packaged and transmitted to an external terminal to achieve low-cost, high-precision positioning results.

[0017] The beneficial effects of this invention are: 1. By collecting frame deformation signals through a strain sensor array and dynamically correcting the coordinates of the sensing units using linear interpolation, the coordinate offset caused by frame deformation is eliminated at the source. For the non-uniform pressure field at the edge, K-means clustering is used to group the pressure data and fill the gaps to optimize the resolution of the pressure distribution map. Then, by triggering dual verification of temporal consistency and spatial deviation, the deformation parameters are iteratively updated, and the final positioning error is stably controlled within 3mm, effectively solving the problem of misjudging the sweet spot as the edge and greatly improving the reliability of the data.

[0018] 2. The activation frequency and information contribution of the edge sensing units are statistically analyzed. Units with low contribution are put into periodic dormancy, which significantly reduces the number of activated units. When the number of activated units decreases, Huffman coding is used to compress the pressure distribution map data to reduce the amount of data transmission. The combined effect of these two measures greatly extends the racket's battery life on a single charge, meeting the long-term use requirements of high-intensity training scenarios.

[0019] 3. The system employs a sliding window averaging method to dynamically correct core parameters, adjusting the sweet spot interpolation weight based on the sweet spot hit rate. It also expands the non-uniform mode judgment angle to address insufficient edge pressure detail retention and dynamically updates the deformation threshold based on usage. This eliminates the need for manual parameter adjustments, adapting to various scenarios such as leisure and entertainment, professional training, and competitive sports, offering superior versatility compared to traditional systems with fixed parameters.

[0020] 4. High-precision positioning can be achieved with existing sensor configurations without relying on high-density sensor arrays, reducing hardware material costs; the simplified sensor array simplifies circuit design and improves production yield; overall cost control significantly reduces the final selling price, making it more acceptable to the mass market and supporting the large-scale production and promotion of smart Peak rackets.

[0021] 5. It outputs precise data such as the final hitting position, optimized pressure distribution, and sweet spot hit rate, accurately displaying the distance and direction of the hitting point from the center of the sweet spot, presenting the trajectory of the pressure center of gravity movement at the moment of impact, and can also accumulate data to generate a personal technical profile. It provides data support for players to adjust their movements and coaches to develop targeted training programs, promoting a shift from experience-driven to data-driven training and significantly improving training efficiency. Attached Figure Description

[0022] The invention will be further illustrated with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention. For those skilled in the art, other drawings can be obtained based on the following drawings without any creative effort.

[0023] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0024] The present invention will be further described in conjunction with the following embodiments.

[0025] Depend on Figure 1 As can be seen, the analysis system based on data acquisition from a smart Peak racket described in this embodiment includes the following steps: S1. By using the strain sensor array and racket face pressure sensing layer embedded inside the racket frame, the frame deformation signal and pressure distribution data at the moment of impact are collected synchronously during the swing. The original deformation signal is filtered using a preset frame dynamic deformation threshold to extract the frame dynamic deformation and relative displacement change rate, and to generate a frame dynamic deformation index. S2. Based on the frame dynamic deformation index, the initial coordinates of each sensing unit of the racket pressure sensing layer are dynamically mapped. When the frame dynamic deformation index exceeds the preset frame dynamic deformation threshold, a linear interpolation algorithm is used to reconstruct the spatial position of the sensing unit of the racket pressure sensing layer between adjacent fixed reference points, and an adjusted set of racket pressure sensing unit coordinates is generated. S3. Based on the adjusted set of racket face pressure sensing unit coordinates, extract the sensing unit data of the racket face pressure sensing layer that is frequently triggered within the sweet spot area. Combine the pressure amplitude and trigger timestamp of each sensing unit in the racket face pressure sensing layer, and use a weighted average fusion strategy to calculate the pressure center position and generate the initial hitting coordinates. S4. Based on the initial hitting coordinates, identify the sensing units of the racket face pressure sensing layer that are triggered by low frequency in the adjacent edge region. If the pressure distribution in this region exhibits non-uniform gradient characteristics, then use the K-means clustering algorithm to group the pressure readings, reconstruct the local pressure field shape, and generate a pressure distribution map optimized based on racket face pressure sensing data. S5. Based on the pressure distribution map optimized by the surface pressure sensing data, the activation frequency and information contribution of the edge sensing units of the surface pressure sensing layer are statistically analyzed, adaptive sampling rules are set, periodic sleep control is implemented for low contribution units, and a simplified surface pressure sensing unit array configuration is generated. S6. Based on the simplified racket face pressure sensing unit array configuration, re-execute the ball hitting coordinate calculation process, compare the consistency of the trigger timing and the distribution of pressure difference between the old and new coordinates. If the deviation exceeds the allowable error bandwidth, return to step S2 to iteratively update the deformation compensation parameters until stable coordinates are output and the final ball hitting position coordinates are generated. S7. Based on the offset between the final hitting position coordinates and the initial calibration coordinates, combined with the sweet spot usage frequency statistics and edge detail retention index, the sliding window mean method is used to dynamically correct the frame dynamic deformation threshold and interpolation weight coefficient, and generate the deformation correction parameters for the next cycle. S8. Based on the deformation correction parameters of the next cycle, the coordinate mapping relationship of the entire sensor network is globally recalibrated. If the number of circuit nodes in the simplified array is reduced, lightweight data compression coding is enabled to entropy code the pressure distribution map to generate low-cost, high-precision positioning results.

[0026] This embodiment describes an analysis system based on data acquisition from a smart Peak racket. The framework's dynamic deformation metrics include deformation amplitude detection, relative displacement rate calculation, and threshold exceedance determination. The adjusted racket face pressure sensor unit coordinate set includes reference point anchoring, linear interpolation weight allocation, and coordinate remapping. The initial hit coordinates include sweet spot data filtering, pressure-time weighting factor setting, and centroid coordinate solving. The optimized pressure distribution map based on racket face pressure sensor data includes non-uniform pattern recognition, cluster center initialization, and local field reconstruction. The simplified racket face pressure sensor unit array configuration includes activation frequency statistics, information entropy evaluation, and sleep scheduling instruction generation. The final hit position coordinates include temporal consistency verification, pressure gradient matching, and iterative convergence determination. The deformation correction parameters for the next cycle include threshold sliding update, interpolation coefficient correction, and sweet spot weight enhancement. The low-cost, high-precision positioning results include node number compression ratio calculation, Huffman coding application, and positioning error backtracking verification.

[0027] The analysis system based on intelligent Peak racket data acquisition described in this embodiment includes the following steps: Firstly, by simultaneously acquiring frame deformation signals and pressure distribution data at the moment of impact during the swing using a strain sensor array and racket face pressure sensing layer embedded within the racket frame. Secondly, filtering the original deformation signals using a preset frame dynamic deformation threshold to extract the frame dynamic deformation and relative displacement change rate, and generating dynamic deformation indices for the frame, the system proceeds as follows: Four strain gauges are embedded at the connection between the racket handle and the frame, at the top of the frame and at symmetrical positions on both sides as fixed reference points. The remaining strain sensors are arranged at equal intervals along the inner side of the frame. The racket face pressure sensing layer is made of a flexible piezoresistive film, covering the entire racket face and divided into 64 sensing units of the racket face pressure sensing layer. The sensing unit of each racket face pressure sensing layer is connected to the central processing module through an independent wire. The acquired raw deformation signal is filtered by a second-order Butterworth low-pass filter to remove high-frequency vibration noise, retaining the structural response signal in the 0.1-10 Hz frequency band of the field of puck ball technology. The length change rate of each strain sensor within adjacent sampling periods is calculated, and the two-dimensional displacement vector of each sensing unit relative to the initial installation position is calculated by combining the distance constraints between reference points. The Euclidean norm of the displacement vector is compared with the preset frame dynamic deformation threshold. If it exceeds the threshold, it is marked as a valid deformation event, and a quantitative index containing deformation amplitude, direction and duration is generated.

[0028] The analysis system based on data acquisition from a smart Peak racket described in this embodiment involves the following steps: First, based on the frame dynamic deformation index, the initial coordinates of each sensing unit in the racket face pressure sensing layer are dynamically mapped. Second, when the frame dynamic deformation index exceeds a preset frame dynamic deformation threshold, a linear interpolation algorithm is used to reconstruct the spatial positions of the sensing units in the racket face pressure sensing layer between adjacent fixed reference points, generating an adjusted set of racket face pressure sensing unit coordinates. Establish the sensor unit coordinate matrix during initial calibration, where the sensor unit coordinates of each pressure sensing layer on the striking surface are calibrated by a laser positioning instrument and stored in non-volatile memory; Based on the displacement vector marked in the deformation event, determine the reference point interval to which the affected sensing unit belongs; A one-dimensional linear interpolation function is constructed between every two adjacent fixed reference points, with the actual displacement of the reference points as the endpoint value, and the displacement of the intermediate sensing unit is allocated according to the original ratio. The displacement is superimposed on the initial coordinates to generate a dynamically updated set of coordinates, which is then synchronously written into the coordinate mapping table for subsequent calculations.

[0029] The analysis system based on data acquisition from a smart Peak racket described in this embodiment involves the following steps: Based on the adjusted set of racket face pressure sensing unit coordinates, extracting sensor unit data from the racket face pressure sensing layer that is frequently triggered within the sweet spot; combining the pressure amplitude and trigger timestamp of each sensing unit in the racket face pressure sensing layer; and using a weighted average fusion strategy to calculate the pressure center position and generate preliminary hitting coordinates. The sweet spot is defined as the 30% area of ​​the center of the racket face, corresponding to sensor unit numbers 18-47; Filter the set of cells in the ball-hitting event where the pressure value is greater than 5N and the trigger time difference is less than 2ms; Each effective unit is assigned a weight, which is determined by the pressure amplitude and the reciprocal of the time delay, and the formula is w_i=P_i / (1+Δt_i); Calculate the weighted center coordinates x=Σ(w_i·x_i) / Σw_i, y=Σ(w_i·y_i) / Σw_i, and output the initial hitting coordinates.

[0030] The analysis system based on data acquisition from a smart Peak racket described in this embodiment includes the following steps: Based on the initial hit coordinates, identifying the sensing units of the racket face pressure sensing layer in the vicinity of the edge region, which are low-frequency triggered; if the pressure distribution in this region exhibits a non-uniform gradient characteristic, then using the K-means clustering algorithm to group the pressure readings, reconstructing the local pressure field morphology, and generating a pressure distribution map optimized based on the racket face pressure sensing data. A neighborhood with a radius of 15mm is defined centered on the initial ball-hitting coordinates, and the pressure values ​​of all sensing units within this area are extracted. Calculate the pressure gradient vector. If the angle between the maximum gradient direction and the radial direction is greater than 30 degrees, it is determined to be a non-uniform mode. Initialize the number of clusters k=2, and iteratively update the cluster centers with the pressure value as the feature dimension until convergence; By treating units within the same cluster as continuous pressure regions and using bilinear interpolation to fill the gaps, a high-resolution pressure distribution map is generated.

[0031] The analysis system based on data acquisition from a smart Peak racket described in this embodiment includes the following steps: Based on the pressure distribution map optimized from racket face pressure sensing data, the system statistically analyzes the activation frequency and information contribution of the edge sensing units in the racket face pressure sensing layer, sets adaptive sampling rules, implements periodic sleep control for low-contribution units, and generates a simplified racket face pressure sensing unit array configuration. Record the number of times the edge sensing unit of each racket face pressure sensing layer is triggered in the past 100 shots, and calculate the activation frequency f_i; Calculate the information entropy H_i=-Σp_jlog2p_j of the sensing unit in the pressure distribution reconstruction of each pressure sensing layer, where p_j is the probability of the unit's contribution to the clustering result; Set the sleep conditions: f_i<0.05 and H_i<0.3. If the conditions are met, the sensing unit of the pressure sensing layer will turn off the power supply in the next sampling cycle. Generate configuration instructions containing active and sleep states, and control multiple analog switches to switch sensing channels via GPIO interface.

[0032] The analysis system based on data acquisition from a smart Peak racket described in this embodiment, which re-executes the shot coordinate calculation process based on the simplified racket face pressure sensor array configuration, compares the consistency of trigger timing and pressure difference distribution between the old and new coordinates, and if the deviation exceeds the allowable error bandwidth, returns to step S2 to iteratively update the deformation compensation parameters until stable coordinates are output, and generates the final shot position coordinates, specifically includes the following steps: Repeat step S3 under the simplified configuration to obtain the new hitting coordinates (x', y'); Calculate the time synchronization error between the original coordinates (x, y) and the new coordinates Δt = |t - t'|. If Δt > 1ms or the Euclidean distance d = √[(x - x')² + (y - y')²] > 3mm, then it is determined to be a mismatch. If there is a mismatch, the deformation threshold is lowered by 5%, and steps S2 to S5 are re-executed, for a maximum of 3 iterations. If the process still fails to converge after the third iteration, the last result is retained as the final shot position coordinates.

[0033] The analysis system based on data acquisition from a smart Peak racket described in this embodiment, specifically the step of dynamically correcting the frame's dynamic deformation threshold and interpolation weight coefficients using a sliding window mean method based on the offset between the final hitting position coordinates and the initial calibration coordinates, combined with statistics on the frequency of sweet spot usage and the edge detail retention index, to generate deformation correction parameters for the next cycle, is as follows: The sweet spot hit rate r_sweet in the last 50 hits is calculated. If r_sweet > 0.8, the sweet spot interpolation weight coefficient α = α × 1.1 is increased. Calculate the pressure detail retention rate of the edge region d_edge = actual number of clusters / theoretical maximum number of clusters. If d_edge < 0.6, then expand the non-uniform mode judgment angle to 40 degrees. A moving average filter with a window length of 20 is applied to the deformation threshold. The new threshold = 0.9 × old threshold + 0.1 × current average deformation. The updated α, judgment angle, and threshold are written into the parameter register as the basis for the next cycle of correction.

[0034] The analysis system based on smart Peak racket data acquisition described in this embodiment involves the following steps: First, based on the deformation correction parameters of the next cycle, the coordinate mapping relationship of the entire sensor network is globally recalibrated. If the number of circuit nodes in the simplified array is reduced, lightweight data compression encoding is used to entropy encode the pressure distribution map, generating low-cost, high-precision positioning results. Reconstruct the dynamic coordinate mapping table of all sensing units based on the new parameters, overwriting the original stored content; The total number of active units N_active in the current pressure sensing layer is detected. If N_active ≤ 48, the Huffman coding module is activated. Using the pressure value quantification level as the symbol source, a frequency statistics table is constructed to generate a variable-length coding dictionary; The compressed pressure data and the final ball-hitting coordinates are packaged and transmitted to an external terminal to achieve low-cost, high-precision positioning results.

[0035] Specifically, the analysis system based on data acquisition from a smart Peak racket described in this embodiment features a racket frame integrally molded from carbon fiber composite material, with microgrooves inside for embedding a strain sensor array. The strain sensor array consists of multiple miniature resistance strain gauges arranged along the inner contour of the racket frame and electrically connected to the central processing module via a flexible printed circuit. The racket face pressure sensing layer covers the entire hitting surface of the racket and is made of a flexible piezoresistive film, divided into 64 independent sensing units. Each sensing unit is led out through an independent wire and connected to the analog input port of the central processing module. The central processing module is fixedly installed in the internal cavity of the grip handle, and integrates an analog-to-digital converter, a microcontroller, a non-volatile memory, and a wireless communication unit. The grip handle is a hollow cylindrical structure, covered with an anti-slip rubber layer, and has an internal mounting bracket for accommodating the central processing module and a battery compartment.

[0036] In terms of hardware layout, the strain sensor array includes four fixed reference points and several intermediate strain sensors. The four fixed reference points are located on both sides of the connection between the handle and the frame (i.e., symmetrical positions), the center point of the top of the frame, and the center point of the bottom of the frame, totaling four points, serving as the reference anchor points for coordinate mapping. The remaining strain sensors are evenly distributed along the inner circumference of the racket frame at equal intervals, with an adjacent sensor spacing of approximately 8 mm. The total number is determined based on the circumference of the frame, typically 12-16. All strain sensors are connected to the differential signal input interface of the central processing module via twisted-pair shielded cables to suppress electromagnetic interference. The racket face pressure sensing layer uses a 0.2 mm thick flexible piezoresistive film, with its surface divided into 64 sensing units in an 8×8 grid pattern. Each sensing unit has an area of ​​10 mm × 10 mm, and insulating isolation strips are formed between the units through laser etching. The output of each unit is connected to the 12-bit ADC channel of the central processing module via a multiplexer. The central processing module uses an ARM Cortex-M7 architecture microcontroller with a main frequency of 480MHz, 2MB Flash and 512KB SRAM, and runs a real-time operating system to schedule various processing tasks.

[0037] During system operation, the initialization phase involves self-testing and calibration of each component. The central processing module reads the pre-stored initial sensor unit coordinate matrix from the non-volatile memory. This matrix is ​​obtained by a high-precision laser positioning instrument performing three-dimensional spatial coordinate measurements on each pressure sensor unit before shipment and is stored in the form of (x_i, y_i), where i = 1 to 64. Simultaneously, the physical positions of four fixed reference points are also recorded as baseline coordinates. The system then enters the data acquisition phase: when the user swings the racket to hit the ball, the racket frame undergoes elastic deformation due to the force, and the strain sensor array synchronously outputs a voltage signal reflecting the local strain state. At the same time, the racket face pressure sensing layer generates a pressure response at the moment of ball contact, and each triggered unit outputs an analog voltage value proportional to the applied pressure. Both signals are pre-amplified and sent to the central processing module for synchronous sampling at a sampling frequency of 1kHz to ensure timestamp alignment.

[0038] The raw deformation signal is first processed by a second-order Butterworth low-pass filter with a cutoff frequency of 10Hz to filter out noise components caused by air resistance or high-frequency vibration, retaining the structural dynamic response signal in the range of 0.1-10Hz. The central processing module calculates the length change rate of each strain sensor in adjacent sampling periods, using the formula ΔL / Δt, where ΔL is derived from the strain-displacement conversion coefficient. Combining the geometric constraints between the four fixed reference points (i.e., the distance between two points is a known constant when there is no external force), the two-dimensional displacement vector (u_i, v_i) of each intermediate strain sensor relative to its initial installation position is calculated using the least squares method. Subsequently, the Euclidean norm √(u_i²+v_i²) of this displacement vector is calculated and compared with the preset frame dynamic deformation threshold of 0.15mm. If it exceeds this threshold, the moment is marked as a valid deformation event, and the deformation amplitude, direction angle, and duration are recorded to generate a dynamic deformation index for the frame.

[0039] Upon detecting a valid deformation event, the system enters the dynamic coordinate mapping stage. The central processing module determines the reference point interval to which the affected pressure sensing unit belongs based on the displacement vector marked in the deformation event. For example, if a sensing unit is located on the arc between the left handle connection point and the top reference point, a one-dimensional linear interpolation function is constructed within this interval using the actual displacements of its two reference points as boundary conditions. Let the displacement of the left reference point be (u_A, v_A), and the right reference point be (u_B, v_B). The ratio of the sensing unit's distance from point A is λ (0 < λ < 1), then its displacement is (u_A + λ(u_B - u_A), v_A + λ(v_B - v_A)). This displacement is superimposed on the unit's initial coordinates (x_i, y_i) to obtain the adjusted dynamic coordinates (x'_i, y'_i). After updating the coordinates of all 64 units, a new coordinate set is formed and written into a coordinate mapping table in RAM for subsequent steps.

[0040] After obtaining the adjusted set of racket face pressure sensor unit coordinates, the system performs sweet spot data filtering. The sweet spot is defined as the 30% area of ​​the racket face center, corresponding to sensor units numbered 18 to 47 (numbered in row priority). The central processing module iterates through the pressure readings of these 30 units, filtering out the subset of data with pressure values ​​greater than 5N and trigger times less than 2ms from other valid units. For each valid unit i, the weight w_i = P_i / (1 + Δt_i) is calculated, where P_i is the pressure amplitude (in N), and Δt_i is the time difference between the trigger time of this unit and the earliest triggering unit (in ms). Subsequently, the weighted center coordinates are calculated: x = Σ(w_i·x'_i) / Σw_i, y = Σ(w_i·y'_i) / Σw_i, and the initial hitting coordinates (x, y) are output.

[0041] Next, the system defines a circular neighborhood with a radius of 15mm centered on the initial hit coordinates and extracts the pressure values ​​of all sensing units within this region. The central processing module calculates the pressure gradient vector between each unit by performing a finite difference approximation on the pressure field within the neighborhood to obtain the direction of the maximum gradient. If the angle between this direction and the radial vector pointing from the initial hit coordinates to the unit position is greater than 30 degrees, the pressure distribution in this region is determined to exhibit non-uniform gradient characteristics. At this point, the K-means clustering algorithm is initiated, initializing the number of clusters k=2. Using the pressure value of each unit as a one-dimensional feature input, the cluster centers are iteratively updated until the center offset is less than 0.1N between two consecutive iterations. After clustering is completed, units within the same category are considered as continuous pressure regions, and bilinear interpolation is used to fill the pressure values ​​in the gaps they enclose, generating an optimized pressure distribution map with a resolution of 2mm×2mm.

[0042] Based on the optimized pressure distribution map derived from racket face pressure sensing data, the system statistically analyzes the activation frequency f_i of each unit in the edge region (unit numbers 1-17 and 48-64) over the past 100 shots, which is calculated as the number of triggers divided by 100. Simultaneously, the information entropy H_i is calculated: first, the probability p_j (j=1,2) of the unit belonging to a certain category in the clustering results is calculated, and then substituted into the formula H_i=-Σp_jlog2p_j. If a unit satisfies f_i<0.05 and H_i<0.3, it is determined to be a low-contribution unit. The central processing module generates a sleep scheduling instruction, controlling a multiplexed analog switch via GPIO pins to cut off the power supply and signal path of the unit, putting it into sleep mode in the next sampling cycle. The remaining units remain active, forming a simplified racket face pressure sensing unit array configuration.

[0043] In the simplified configuration, the system re-executes step S3 to calculate the ball-hitting coordinates, obtaining the new ball-hitting coordinates (x', y') and their trigger time t'. Then, it compares the time synchronization error Δt = |t - t'| and the spatial deviation d = √[(x - x')² + (y - y')²] between the original coordinates (x, y, t) and the new coordinates. If Δt > 1ms or d > 3mm, a mismatch is determined, triggering an iterative correction mechanism. At this point, the central processing module lowers the deformation threshold by 5% (e.g., from 0.15mm to 0.1425mm) and returns to step S2, using the new threshold to re-perform deformation event detection, coordinate mapping, and subsequent calculations. This iterative process is executed a maximum of 3 times. If the third iteration fails to converge, the result of the last calculation is retained as the final ball-hitting position coordinates.

[0044] After obtaining the final shot position coordinates, the system enters the parameter adaptive update phase. The central processing module accumulates the number of sweet spot hits in the last 50 shots and calculates the sweet spot usage frequency r_sweet. If r_sweet > 0.8, the sweet spot interpolation weight coefficient α is multiplied by 1.1 (initial α = 1.0). Simultaneously, the edge detail retention d_edge is calculated: the ratio of the actual number of clusters (usually 1 or 2) to the theoretical maximum number of clusters (set to 3). If d_edge < 0.6, the non-uniform mode judgment angle is increased from 30 degrees to 40 degrees. Furthermore, a moving average filter with a window length of 20 is applied to the deformation threshold; the new threshold = 0.9 × old threshold + 0.1 × the average of the current 20 deformation values. The updated α value, judgment angle, and deformation threshold are written to the parameter register as deformation correction parameters for the next cycle.

[0045] Finally, the system performs a global recalibration of the entire sensor network based on the new calibration parameters. The central processing module reconstructs the dynamic coordinate mapping table of all 64 sensor units in the racket face pressure sensing layer according to the updated interpolation weights and reference point displacement relationships, overwriting the original stored content. Simultaneously, it detects the total number of active units N_active in the current racket face pressure sensing layer. If N_active ≤ 48 (i.e., at least 16 units are in a dormant state), the Huffman coding module is activated. This module uses the 8-bit quantization levels (0-255) of the pressure value as the symbol source, statistically analyzes the frequency of each level in the past 10 shots, constructs a Huffman tree, and generates a variable-length coding dictionary. The pressure distribution map data, optimized based on the racket face pressure sensing data, is encoded using this dictionary, packaged with the final shot coordinates, and transmitted to an external terminal device via Bluetooth 5.0, completing the output of low-cost, high-precision positioning results. The entire processing flow is completed within 200ms after a single shot event, ensuring real-time performance.

[0046] To enable those skilled in the art to fully understand and implement this invention, the following supplementary explanation of the operating principle and technical effect implementation mechanism of this invention is provided in conjunction with a specific application scenario.

[0047] During a high-intensity pickle training session, the athlete performs consecutive forehand smashes, with the ball repeatedly striking the area slightly above the sweet spot on the racket face. At this time, the racket frame experiences minute but cumulative elastic deformation due to repeated stress, causing dynamic changes in the local strain state sensed by the strain sensor array embedded in its internal micro-grooves. The central processing module simultaneously receives voltage signals from the strain sensor array and pressure response signals from the racket face pressure sensing layer. Because the physical positions of the four fixed reference points in the strain sensor array (located on the left and right sides of the connection between the handle and the frame, the top center of the frame, and the bottom center of the frame, respectively) are structurally highly rigid, their displacements can serve as a stable reference. The intermediate strain sensors, located in a flexible arc segment, experience relative displacement after being subjected to stress. The central processing module filters out high-frequency interference using a second-order Butterworth low-pass filter, then calculates the two-dimensional displacement vector (u_i, v_i) of each intermediate strain sensor relative to its initial installation position using the least squares method, and calculates its Euclidean norm. When the norm exceeds the preset frame dynamic deformation threshold of 0.15mm, the system determines that the current ball-hitting event is accompanied by significant structural deformation and needs to activate the coordinate dynamic mapping mechanism.

[0048] In this scenario, the initial coordinates of the pressure sensing unit (e.g., number 22) located on the upper left arc of the frame are (x... 22 ,y 22 Since the unit is located within the interval between the left handle connection point and the top reference point, the central processing module calculates its dynamic displacement using a linear interpolation function based on the measured displacements (u_A, v_A) and (u_B, v_B) of these two fixed reference points, combined with the geometric scale λ of the unit on the arc segment. This dynamic displacement is then superimposed onto the initial coordinates to obtain the updated coordinates (x'). 22 ,y' 22 This process is executed synchronously on all 64 pressure sensing units, ensuring that the spatial position on which the subsequent pressure center of gravity calculation depends reflects the current true geometric state of the racket, thereby solving the coordinate offset problem caused by frame deformation.

[0049] Subsequently, the system focuses on data filtering within the sweet spot (corresponding to units 18-947). Assuming the shot triggered units 20, 21, 22, 29, and 30, with pressure values ​​of 6.2N, 7.8N, 8.5N, 5.3N, and 6.0N respectively, and all trigger time differences less than 2ms, the central processing module assigns weights to each unit according to the formula w_i=P_i / (1+Δt_i), with the earliest triggered unit having the highest weight (Δt_i=0). After weighted center calculation, the initial shot coordinates (x,y) are output. These coordinates are based on a dynamically corrected coordinate set, avoiding the technical deviation of misjudging a sweet spot shot as an edge area due to frame deformation.

[0050] To further improve positioning accuracy, the system defines a 15mm neighborhood centered at (x,y) and extracts the pressure values ​​of all units within the neighborhood. If the pressure gradient analysis shows that the angle between the maximum gradient direction and the radial direction reaches 35°, exceeding the 30° threshold, K-means clustering is initiated. For example, the clustering results group high-pressure units (21, 22) into one class and low-pressure units (20, 29, 30) into another, indicating an asymmetric pressure distribution in the contact area. In this case, the system uses bilinear interpolation to fill the pressure values ​​in the gap area surrounded by the two classes of units (such as units 28, 37, etc., which were not triggered but are located within the neighborhood), generating an optimized pressure distribution map with a resolution of 2mm×2mm, thereby restoring a more realistic ball-racket contact pattern.

[0051] Based on this optimization graph, the system backtracks the historical usage data of the edge regions (units 1-17 and 48-64). If unit 5 has only been triggered 3 times in the past 100 shots (f_i=0.03), and it always belongs to the same category in the cluster (p1≈1, p2≈0), the information entropy H_i≈0, satisfying the sleep condition of f_i<0.05 and H_i<0.3. The central processing module then controls the multiplexer via GPIO pins to cut off the power supply and signal path of unit 5, putting it into sleep mode. This reduces system power consumption and data processing load without losing critical information.

[0052] With a simplified configuration, the system recalculates the shot coordinates to obtain (x', y'). If a spatial deviation of d=3.5mm is found during comparison, exceeding the 3mm tolerance, iterative correction is triggered: the deformation threshold is lowered from 0.15mm to 0.1425mm, and deformation detection and coordinate mapping are re-executed. After the second iteration, the deviation is reduced to 2.1mm, meeting the convergence condition, and the final shot coordinates are confirmed.

[0053] Subsequently, the system calculated that 45 out of the last 50 shots hit the sweet spot (r_sweet=0.9>0.8), so the sweet spot interpolation weight coefficient α was increased from 1.0 to 1.1 to enhance the influence of sweet spot data in subsequent center of gravity calculations. Meanwhile, because the actual number of clusters was mostly 1 (d_edge=1 / 3≈0.33<0.6), the non-uniform pattern judgment angle was expanded from 30° to 40° to accommodate more complex pressure distribution patterns. The deformation threshold was also updated using a moving average, incorporating the statistical mean of the latest 20 deformations to improve the threshold's adaptability.

[0054] Finally, the central processing module reconstructs the dynamic coordinate mapping table for all 64 units based on the updated parameters. Since the current number of active units is 46 (≤48), the system enables Huffman coding: an encoding dictionary is built based on the frequency of occurrence of pressure quantification levels in the past 10 shots, and the optimized pressure distribution map is compressed and transmitted to the mobile app along with the final shot coordinates via Bluetooth 5.0. The entire process is completed within 200ms, ensuring real-time feedback while achieving low-cost operation through hardware sleep and data compression.

[0055] The above process demonstrates that the present invention, through the collaborative sensing of the strain sensor array and the racket face pressure sensing layer, combined with the dynamic coordinate mapping, sweet spot weighted centering calculation, pressure field optimization reconstruction, sensor sleep scheduling, and parameter adaptive updating mechanisms executed by the central processing module, effectively overcomes the problem of misjudgment of the hitting position caused by structural deformation of the racket frame during high-intensity use. It also suppresses redundant edge sampling while ensuring high resolution of the sweet spot, ultimately achieving high-precision, low-power, and robust motion data acquisition and analysis.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. An analysis system based on data acquisition from a smart Peak racket, characterized in that: Includes the following steps: S1. By using the strain sensor array and racket face pressure sensing layer embedded inside the racket frame, the frame deformation signal and pressure distribution data at the moment of impact are collected synchronously during the swing. The original deformation signal is filtered using a preset frame dynamic deformation threshold to extract the frame dynamic deformation and relative displacement change rate, and to generate a frame dynamic deformation index. S2. Based on the frame dynamic deformation index, the initial coordinates of each sensing unit of the racket pressure sensing layer are dynamically mapped. When the frame dynamic deformation index exceeds the preset frame dynamic deformation threshold, a linear interpolation algorithm is used to reconstruct the spatial position of the sensing unit of the racket pressure sensing layer between adjacent fixed reference points, and an adjusted set of racket pressure sensing unit coordinates is generated. S3. Based on the adjusted set of racket face pressure sensing unit coordinates, extract the sensing unit data of the racket face pressure sensing layer that is frequently triggered within the sweet spot area. Combine the pressure amplitude and trigger timestamp of each sensing unit in the racket face pressure sensing layer, and use a weighted average fusion strategy to calculate the pressure center position and generate the initial hitting coordinates. S4. Based on the initial hitting coordinates, identify the sensing units of the racket face pressure sensing layer that are triggered by low frequency in the adjacent edge region. If the pressure distribution in this region exhibits non-uniform gradient characteristics, then use the K-means clustering algorithm to group the pressure readings, reconstruct the local pressure field morphology, and generate a pressure distribution map optimized based on racket face pressure sensing data. S5. Based on the pressure distribution map optimized by the surface pressure sensing data, the activation frequency and information contribution of the edge sensing units of the surface pressure sensing layer are statistically analyzed, adaptive sampling rules are set, periodic sleep control is implemented for low contribution units, and a simplified surface pressure sensing unit array configuration is generated. S6. Based on the simplified racket face pressure sensing unit array configuration, re-execute the ball hitting coordinate calculation process, compare the consistency of the trigger timing and the distribution of pressure difference between the old and new coordinates. If the deviation exceeds the allowable error bandwidth, return to step S2 to iteratively update the deformation compensation parameters until stable coordinates are output and the final ball hitting position coordinates are generated. S7. Based on the offset between the final hitting position coordinates and the initial calibration coordinates, combined with the sweet spot usage frequency statistics and edge detail retention index, the sliding window mean method is used to dynamically correct the frame dynamic deformation threshold and interpolation weight coefficient, and generate the deformation correction parameters for the next cycle. S8. Based on the deformation correction parameters of the next cycle, the coordinate mapping relationship of the entire sensor network is globally recalibrated. If the number of circuit nodes in the simplified array is reduced, lightweight data compression coding is enabled to entropy code the pressure distribution map to generate low-cost, high-precision positioning results.

2. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The framework's dynamic deformation metrics include deformation amplitude detection, relative displacement rate calculation, and threshold exceedance determination. The adjusted racket face pressure sensing unit coordinate set includes reference point anchoring, linear interpolation weight allocation, and coordinate remapping. The initial hit coordinates include sweet spot data filtering, pressure-time weighting factor setting, and centroid coordinate solving. The pressure distribution map optimized based on racket face pressure sensing data includes non-uniform pattern recognition, cluster center initialization, and local field reconstruction. The simplified racket face pressure sensing unit array configuration includes activation frequency statistics, information entropy evaluation, and sleep scheduling instruction generation. The final hit position coordinates include temporal consistency verification, pressure gradient matching, and iterative convergence determination. The deformation correction parameters for the next cycle include threshold sliding update, interpolation coefficient correction, and sweet spot weight enhancement. The low-cost, high-precision positioning results include node number compression ratio calculation, Huffman coding application, and positioning error backtracking verification.

3. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The steps of synchronously collecting frame deformation signals and pressure distribution data at the moment of impact during the swing and using a strain sensor array and racket face pressure sensing layer embedded inside the racket frame, filtering the original deformation signals using a preset frame dynamic deformation threshold, extracting the frame dynamic deformation and relative displacement change rate, and generating a dynamic deformation index of the frame are as follows: Four strain gauges are embedded at the connection between the racket handle and the frame, at the top of the frame and at symmetrical positions on both sides as fixed reference points. The remaining strain sensors are arranged at equal intervals along the inner side of the frame. The racket face pressure sensing layer is made of a flexible piezoresistive film, covering the entire racket face and divided into 64 sensing units of the racket face pressure sensing layer. The sensing unit of each racket face pressure sensing layer is connected to the central processing module through an independent wire. The acquired raw deformation signal is filtered by a second-order Butterworth low-pass filter to remove high-frequency vibration noise, retaining the structural response signal in the 0.1-10 Hz frequency band of the field of puck ball technology. The length change rate of each strain sensor within adjacent sampling periods is calculated, and the two-dimensional displacement vector of each sensing unit relative to the initial installation position is calculated by combining the distance constraints between reference points. The Euclidean norm of the displacement vector is compared with the preset frame dynamic deformation threshold. If it exceeds the threshold, it is marked as a valid deformation event, and a quantitative index containing deformation amplitude, direction and duration is generated.

4. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The steps of dynamically mapping the initial coordinates of each sensing unit in the racket face pressure sensing layer based on the frame dynamic deformation index, and reconstructing the spatial position of the sensing units in the racket face pressure sensing layer between adjacent fixed reference points using a linear interpolation algorithm to generate an adjusted set of racket face pressure sensing unit coordinates are as follows: Establish the sensor unit coordinate matrix during initial calibration, where the sensor unit coordinates of each pressure sensing layer on the striking surface are calibrated by a laser positioning instrument and stored in non-volatile memory; Based on the displacement vector marked in the deformation event, determine the reference point interval to which the affected sensing unit belongs; A one-dimensional linear interpolation function is constructed between every two adjacent fixed reference points, with the actual displacement of the reference points as the endpoint value, and the displacement of the intermediate sensing unit is allocated according to the original ratio. The displacement is superimposed on the initial coordinates to generate a dynamically updated set of coordinates, which is then synchronously written into the coordinate mapping table for subsequent calculations.

5. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The steps for extracting sensor unit data of the racket face pressure sensing layer that are frequently triggered within the sweet spot area based on the adjusted set of racket face pressure sensing unit coordinates, and calculating the pressure center position by combining the pressure amplitude and trigger timestamp of each sensing unit in the racket face pressure layer with a weighted average fusion strategy to generate preliminary hitting coordinates are as follows: The sweet spot is defined as the 30% area of ​​the center of the racket face, and the corresponding sensor unit number is 18-47; Filter the set of cells in the ball-hitting event where the pressure value is greater than 5N and the trigger time difference is less than 2ms; Each effective unit is assigned a weight, which is determined by the pressure amplitude and the reciprocal of the time delay, and the formula is w_i=P_i / (1+Δt_i); Calculate the weighted center coordinates x=Σ(w_i·x_i) / Σw_i, y=Σ(w_i·y_i) / Σw_i, and output the initial hitting coordinates.

6. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The specific steps for identifying the low-frequency triggered racket face pressure sensing layer in the vicinity of the initial hit coordinates, and for grouping the pressure readings using the K-means clustering algorithm to reconstruct the local pressure field morphology and generate a pressure distribution map optimized based on racket face pressure sensing data, are as follows: A neighborhood with a radius of 15mm is defined centered on the initial ball-hitting coordinates, and the pressure values ​​of all sensing units within this area are extracted. Calculate the pressure gradient vector. If the angle between the maximum gradient direction and the radial direction is greater than 30 degrees, it is determined to be a non-uniform mode. Initialize the number of clusters k=2, and iteratively update the cluster centers with the pressure value as the feature dimension until convergence; By treating units within the same cluster as continuous pressure regions and using bilinear interpolation to fill the gaps, a high-resolution pressure distribution map is generated.

7. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The steps for generating a simplified racket pressure sensing unit array configuration based on the optimized pressure distribution map derived from the racket pressure sensing data, statistically analyzing the activation frequency and information contribution of the edge sensing units in the racket pressure sensing layer, setting adaptive sampling rules, implementing periodic sleep control for low-contribution units, and so on are as follows: Record the number of times the edge sensing unit of each racket face pressure sensing layer is triggered in the past 100 shots, and calculate the activation frequency f_i; Calculate the information entropy H_i=-Σp_jlog2p_j of the sensing unit in the pressure distribution reconstruction of each pressure sensing layer, where p_j is the probability of the unit's contribution to the clustering result; Set the sleep conditions: f_i<0.05 and H_i<0.

3. If the conditions are met, the sensing unit of the pressure sensing layer will turn off the power supply in the next sampling cycle. Generate configuration instructions containing active and sleep states, and control multiple analog switches to switch sensing channels via GPIO interface.

8. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The steps for re-executing the ball-hitting coordinate calculation process based on the simplified racket face pressure sensing unit array configuration, comparing the consistency of trigger timing and pressure difference distribution between the old and new coordinates, and if the deviation exceeds the allowable error bandwidth, returning to step S2 to iteratively update the deformation compensation parameters until stable coordinates are output, and generating the final ball-hitting position coordinates are as follows: Repeat step S3 under the simplified configuration to obtain the new hitting coordinates (x', y'); Calculate the time synchronization error between the original coordinates (x, y) and the new coordinates Δt = |t - t'|. If Δt > 1ms or the Euclidean distance d = √[(x - x')² + (y - y')²] > 3mm, then it is determined to be a mismatch. If there is a mismatch, the deformation threshold is lowered by 5%, and steps S2 to S5 are re-executed, for a maximum of 3 iterations. If the process still fails to converge after the third iteration, the last result is retained as the final shot position coordinates.

9. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The specific steps for generating deformation correction parameters for the next cycle, based on the offset between the final hitting position coordinates and the initial calibration coordinates, combined with sweet spot usage frequency statistics and edge detail retention indicators, and using the sliding window mean method to dynamically correct the frame dynamic deformation threshold and interpolation weight coefficients, are as follows: The sweet spot hit rate r_sweet in the last 50 hits is calculated. If r_sweet > 0.8, the sweet spot interpolation weight coefficient α = α × 1.1 is increased. Calculate the pressure detail retention rate of the edge region d_edge = actual number of clusters / theoretical maximum number of clusters. If d_edge < 0.6, then expand the non-uniform mode judgment angle to 40 degrees. A moving average filter with a window length of 20 is applied to the deformation threshold. The new threshold = 0.9 × old threshold + 0.1 × current average deformation. The updated α, judgment angle, and threshold are written into the parameter register as the basis for the next cycle of correction.

10. The analysis system based on intelligent Peak racket data acquisition according to claim 1, characterized in that: The steps for globally recalibrating the coordinate mapping relationship of the entire sensor network based on the deformation correction parameters of the next cycle, and for using lightweight data compression coding to entropy-encode the pressure distribution map to generate low-cost, high-precision positioning results, are as follows: Reconstruct the dynamic coordinate mapping table of all sensing units based on the new parameters, overwriting the original stored content; The total number of active units N_active in the current pressure sensing layer is detected. If N_active ≤ 48, the Huffman coding module is activated. Using the pressure value quantification level as the symbol source, a frequency statistics table is constructed to generate a variable-length coding dictionary; The compressed pressure data and the final ball-hitting coordinates are packaged and transmitted to an external terminal to achieve low-cost, high-precision positioning results output.

Images