Resistance positioning system based on chessboard and chess pieces and positioning method thereof

Abstract

The invention discloses a resistance positioning system based on a chessboard and chess pieces and a positioning method thereof. A 9 * 10 flexible resistor array is laid on the chessboard, a unique'fingerprint 'microstructure or weight difference is designed at the bottom of each chess piece, and a characteristic resistance value / resistance map is generated during chess piece falling; zPM zero potential scanning is adopted to suppress row and column crosstalk, 16-bit ADC high-speed sampling is carried out, digital domain imbalance gain calibration is carried out, and the stability of an atlas is ensured. The CNN / SVM performs modeling on the space-time sequence, finger operation and chess piece events are accurately distinguished, an event driving state machine analyzes grapes, falling and eating in real time and verifies rules, and a standard FEN character string is output. The system is low in cost and resistant to illumination and shielding, the positioning error is smaller than 1 mm, 20000 cycles are verified without attenuation, the system can be seamlessly connected to an AI engine and a mechanical arm, and real touch and digital intelligent seamless bridging are achieved.

Description

Technical Field

[0001] This invention relates to the field of resistance positioning and sensor linkage, specifically a resistance positioning system and positioning method based on a chessboard and chess pieces. Background Technology

[0002] In today's world, the application of artificial intelligence (AI) in the field of game playing has achieved remarkable success. Since Google's AlphaGo defeated Lee Sedol, the top human professional Go player, in a human-machine Go match in 2016, the powerful capabilities of AI in board games have been widely recognized. This breakthrough has not only promoted the popularization of AI technology but also spurred market demand for physical interactive systems that can combine traditional chess games with digital intelligence. Traditional human-computer games mostly rely on pure software simulation or remote camera capture, which prevents players from experiencing the tactile feel and immersive experience of real chess pieces. Therefore, developing an intelligent system that can seamlessly bridge the physical chessboard and the digital world has significant social and commercial value for users accustomed to traditional game playing methods.

[0003] However, existing smart chessboard solutions generally suffer from technical limitations. For example, while machine vision-based systems can recognize ordinary chessboards and pieces, their performance is affected by environmental factors such as lighting, shadows, and camera angles. Furthermore, player hands can obstruct the real-time recognition of the game, leading to delays and inaccuracies in data processing. Other technologies based on capacitive, NFC, or electromagnetic induction face challenges such as high costs, complex structures, or the need for customized pieces. These issues collectively constitute insurmountable obstacles for existing technologies, severely hindering the widespread adoption of smart chessboard products and the improvement of user experience.

[0004] Therefore, this invention proposes a resistance positioning system and method based on a chessboard and chess pieces. This invention converts the movement behavior of chess pieces in the physical world into precisely quantifiable digital signals. Through precise signal acquisition, multimodal data fusion, and behavior pattern recognition, this system overcomes the limitations of existing technologies, providing a low-cost, robust, and high-precision solution for intelligent chessboards. Summary of the Invention

[0005] The purpose of this invention is to provide a resistance positioning system and positioning method based on a chessboard and chess pieces, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a resistance positioning system based on a chessboard and chess pieces, comprising: An array of variable resistor sensors positioned below the chessboard is used to sense the distribution of voltage changes caused by changes in resistance values ​​at the chess pieces' positions. A differentiated physical structure is set on the bottom of each piece to form a unique resistance spectrum on the sensor array, serving as the piece's "fingerprint". The microcontroller is used to acquire the analog signals output by the sensor array and obtain digitized distribution data of voltage changes caused by changes in resistance value through analog-to-digital conversion. A behavior pattern recognition model deployed on the microcontroller is used to distinguish between human hand operations and chess piece placement events based on the spatial and temporal characteristics of the distribution data of voltage changes caused by changes in resistor value. The event-driven state machine module is used to update the game state in real time based on the identified behavioral events, including capturing, placing, taking pieces, and illegal moves; The chess piece recognition module is used to match the current resistance value spectrum with the pre-stored chess piece "fingerprint" template after confirming the chess piece placement event to identify the chess piece type; The coordinate mapping and FEN generation module is used to convert the position and identity of chess pieces into standard Chinese chess FEN strings for use by AI engines or chess notation systems.

[0007] The positioning method of the above-described resistor positioning system includes the following steps: S1 uses a variable resistor sensor array deployed under the chessboard to collect in real time the distributed analog signal of voltage change caused by resistance change at each chess position, and obtains a digital distributed data matrix of voltage change caused by resistance change through analog-to-digital conversion. S2 uses a zero-potential scanning method to scan the sensor array in rows and columns. When measuring the target sensing unit, the potential of the non-target row and non-target column is driven to 0 V to physically cut off the crosstalk current path, thereby suppressing the crosstalk error of the resistor array. S3 inputs the distribution data of voltage change caused by resistance change into a pre-trained lightweight machine learning model deployed on a local microcontroller, extracts spatial distribution features and temporal dynamic features to distinguish between brief finger contact and real chess piece placement events, and outputs corresponding behavioral event labels. S4 drives a finite state machine based on the behavioral event tags to sequentially complete the event recognition and game state update of capturing, placing, taking or illegal moves of pieces; After confirming the move, S5 performs template matching between the distribution map of voltage change caused by the change in resistance value at the current position and the pre-stored 14 types of chess piece "fingerprint" templates to identify the chess piece type and generate a standard Chinese chess FEN string by combining the row and column coordinates of the position. S6 sends the FEN string to the host computer or AI engine in real time via USB or Bluetooth to enable human-computer chess playing or game record recording.

[0008] Furthermore, the zero-potential scanning method employs a single shared operational amplifier and multiplexer structure to scan all rows and columns sequentially, thereby reducing hardware cost and power consumption.

[0009] Furthermore, the lightweight machine learning model is an 8-bit quantized convolutional neural network or support vector machine, with a model size ≤50 kB, inference time ≤3 ms, and runs directly on the microcontroller.

[0010] Furthermore, the chess piece "fingerprint" template is obtained through at least one of the following methods: Each chess piece has a unique microstructure on its bottom, forming a spatially unique resistance spectrum on the sensor array; Each piece has a calibrated, minute weight difference, resulting in a distinguishable resistance amplitude characteristic.

[0011] Furthermore, the template matching employs an Euclidean distance or correlation coefficient algorithm, and the matching threshold is dynamically adjusted.

[0012] Furthermore, the section after S1 and before S3 also includes: Digital domain offset and gain calibration steps are used to eliminate errors caused by sensor non-uniformity and temperature drift.

[0013] Furthermore, the FEN string conforms to the Chinese Chess Cloud Library specification and undergoes CRC verification after generation to ensure correct transmission.

[0014] Furthermore, in S4, if multiple people touch the game simultaneously or the pieces are overlapped, the state machine enters an error message mode and suspends FEN output until the error is resolved. Compared with the prior art, the beneficial effects of this invention are: 1. Integrated Identification of Piece Identity and Position Based on Variable Resistance: This invention employs a customized variable resistance sensor array as the core data acquisition layer. By designing unique microstructures or conductive patterns on the bottom of the piece, when the piece is placed, each sensor unit can not only sense the resistance value, but also uniquely identify the type of piece based on the distribution spectrum of voltage changes caused by the change in resistance value or the change in resistance value, rather than simply determining whether a piece is present.

[0015] 2. Behavioral Pattern Recognition Based on Multimodal Data Fusion: This invention creatively applies machine learning algorithms to the analysis of resistance sensor data streams. The algorithm can extract key features from multidimensional data (including resistance magnitude, spatial distribution, and temporal dynamic changes) and train a model to reliably distinguish between two distinct physical behaviors: player hand operations (such as grasping or brief contact) and piece placement.

[0016] 3. Event-Driven Chess State Machine Model: This invention constructs a streamlined and efficient finite state machine. This state machine is driven by discrete events identified in the sensor data stream (such as "sudden drop in resistance" and "sudden increase in resistance"), capturing key chess actions such as "capturing," "placing," and "taking pieces" in real time, and verifying them according to chess rules. This event-based logic processing mechanism ensures real-time and accurate updates to the chess state. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a three-round human-computer chess game according to the present invention; Figure 2 This is the logic diagram for the human-computer interaction in this invention; Figure 3 This is the actual robotic arm of the present invention. Detailed Implementation

[0018] 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.

[0019] See Figures 1 to 3 This invention provides a resistance positioning system and method based on a chessboard and chess pieces. The entire system is implemented through the collaborative work of an embedded resistance sensor matrix (chessboard) and a microcontroller (MCU) with host computer software (PC or mobile device). Its core function is to convert physical contact events on the chessboard into high-precision, interference-resistant digital information and update the chessboard status in real time.

[0020] 1 System Architecture and Data Flow Overview Overall positioning process: a. Physical input (chess piece / hand contact): Players perform capture or placement operations on the flexible variable resistance chessboard.

[0021] b. Signal Acquisition and Optimization: The microcontroller drives the signal conditioning circuit (ZPM circuit) to scan the chessboard sensor array in parallel, acquire analog signals and perform analog-to-digital conversion (ADC).

[0022] c. Event Detection (State Machine): Embedded software monitors data stream changes in real time, identifies discrete events such as "sudden drop in resistance" or "sudden increase in resistance", and triggers the finite state machine (FSM) to transition from states such as Idle and Piece_Lifted.

[0023] d. Data optimization and behavior recognition (AI algorithm): Within the time window of the event, the cached resistance data stream is fed into a pre-trained machine learning model. By analyzing multimodal features such as resistance value, spatial distribution and temporal dynamics, the model distinguishes between human hand operations (interference) and piece placement (effective movement).

[0024] e. Piece Identification and Positioning: Once a valid move is confirmed, the system performs pattern matching based on the distribution pattern (fingerprint) of the voltage change caused by the change in resistance value at that piece position, accurately identifies the type and identity of the piece, and maps its row and column coordinates to standard Chinese chess coordinates.

[0025] f. Chess game status output: The system generates a FEN string that conforms to the Chinese Chess Cloud Library specification and sends it to the host computer for human-computer play or chess game display via communication protocol.

[0026] 2. Sensing and Interaction Mechanism between Chess Pieces and the Chessboard a. Piece "Fingerprint" Design: To achieve the identification of piece types (such as "General", "Chariot", "Cannon", etc.), this invention proposes an innovative piece-to-board sensing interaction mechanism. The core idea of ​​this mechanism is that each piece has a unique "fingerprint" on its bottom corresponding to its type. When the piece is placed on the board, this fingerprint is physically transmitted to a resistive sensor array beneath the board.

[0027] Weight-based differentiation: Each piece possesses a calibrated, minute weight difference that matches its type. When a piece is placed on its position, sensors precisely measure the specific resistance value it generates. By establishing a mapping between this resistance value and the piece's type within a machine learning model, the system can identify the piece's identity.

[0028] The design is based on a distribution pattern of voltage changes caused by changes in resistance (preferred): the bottom of the chess piece is not a smooth plane, but rather a conductive material or microstructure with a unique pattern. When the chess piece is placed on the chessboard, these patterns generate a specific resistance value or resistance distribution pattern on the sensor array below, as unique as a human fingerprint. Even if different types of chess pieces have similar weights, the resulting resistance patterns will be quite different, providing a reliable basis for accurate identification.

[0029] b. Chessboard sensor array: The chessboard itself consists of a 9x10 variable resistance sensor matrix, which is made of a flexible, highly sensitive variable resistance film to ensure accurate capture of the tiny resistance values ​​of the chess pieces.

[0030] 3. Signal Acquisition and Optimization Processing (Hardware) This section ensures the accuracy of the conversion from physical contact to digital signals and is the first step in optimizing resistor positioning data.

[0031] a. Crosstalk Suppression: Application and Optimization of Zero-Potential Scanning Method The inherent "crosstalk effect" of resistive sensor arrays is a major technical challenge affecting positioning accuracy. This invention employs the zero-potential scanning (ZPM) method and optimizes it based on the traditional ZPM method.

[0032] ZPM Principle: When measuring a specific sensing unit, a driving voltage is applied to the target row via a driving circuit, while the potentials of all other non-target rows and all columns are clamped to 0V. Due to the "virtual ground" characteristic of the operational amplifier, the voltage difference across all unmeasured sensing units is 0V, thus physically eliminating the generation of crosstalk current and significantly improving measurement accuracy.

[0033] Optimization Solution: Traditional ZPM requires an operational amplifier for each column. This invention achieves polling scanning of all columns by employing a high-precision multiplexer and a shared operational amplifier. This design minimizes the number and complexity of required components while ensuring measurement accuracy.

[0034] b. Analog-to-digital conversion and signal calibration: After crosstalk is successfully suppressed, the acquired analog voltage signal needs to be converted into a digital signal for processing by the microcontroller.

[0035] Data Acquisition: This invention employs a high-resolution, multi-channel analog-to-digital converter (ADC) to acquire signals from the sensor array in parallel or at high speed. High resolution is crucial for capturing subtle resistance changes in the "fingerprints" of the chess pieces, while the multi-channel design supports parallel scanning, improving data acquisition speed.

[0036] Signal calibration involves offset and gain calibration in the digital domain after ADC conversion to compensate for electrical differences and background noise between sensors. This includes eliminating background noise (offset) by acquiring reference data in a pieceless state and compensating for nonlinear differences between sensors using software algorithms.

[0037] 4. Localization and Behavior Pattern Recognition Algorithm (AI Core) This is the core innovation of the invention, which solves the problem of interference caused by manual operation and achieves highly robust positioning.

[0038] a. Event-based state machine detection model This algorithm first introduces an event-driven finite state machine (FSM) to efficiently track changes in the chess game by continuously monitoring the data flow in the sensor array.

[0039] State transitions: The state machine defines core states such as Idle, Piece_Lifted, Piece_Placed, and Capture. When the system detects a piece marked as "with a piece" whose resistance value drops sharply to zero within a short period of time, the Piece_Lifted (capture) event is triggered.

[0040] Data caching: This event-driven architecture activates data caching within the time window of the capture and placement events, storing time-series slices of sensor data streams as input for the next stage of AI model analysis.

[0041] b. The resistance data stream cached by multimodal data fusion and behavior pattern recognition is fed into the machine learning model to distinguish between "human hand operation" and "chess piece placement", which is the key to solving hand interference.

[0042] Distinguishing between resistance patterns: The resistance "fingerprints" left by the player's hand and the chess pieces on the chessboard differ significantly in space and time.

[0043] Spatial distribution characteristics: The resistance value spectrum of the chess pieces is highly concentrated, affecting only a single chess position; while the voltage change caused by the change in resistance value of the human hand is unevenly distributed and may span multiple adjacent chess positions.

[0044] Time dynamic characteristics: The resistance signal of manual operation ("grabbing-lifting-moving") shows a dynamic curve and is unstable; the placement process is a rapid "placement-stabilization" process, and the signal quickly tends to be stable.

[0045] Resistance amplitude characteristics: The force applied by a human hand is usually much greater than the weight of the chess piece itself.

[0046] Model Training and Implementation: Supervised learning methods (such as Convolutional Neural Networks (CNNs) or Support Vector Machines (SVMs) are used to train the model on collected sensor data samples with clear labels (human hand operations and chess piece placements). The trained lightweight model is deployed on a microcontroller (local deployment / on-chip intelligence) to achieve low-latency real-time recognition and classification.

[0047] c. Piece Location and Identity Recognition Once the behavior pattern recognition algorithm successfully determines that a piece has been placed or captured, the system immediately enters the final location and recognition stage.

[0048] Fingerprint-based type recognition: The system compares the resistance spectrum (or electrical resistance spectrum) collected by the piece placement sensor with 14 pre-stored "fingerprint" templates for different piece types. Using algorithms such as Euclidean distance and correlation coefficient analysis, the system finds the best-matching template, thus accurately identifying the piece type.

[0049] Coordinate mapping and FEN value generation: The sensor row and column coordinates of the recognition results are mapped to standard Chinese chess board coordinates. Finally, the system generates a FEN string conforming to the Chinese Chess Cloud Database specification based on the layout information of all pieces on the board.

[0050] 5. Application in human-computer chess games a. Embedded System Operation: The microcontroller (MCU) serves as the core of the embedded system, responsible for executing all low-level data optimization processes described in Sections .3 and .4, including signal acquisition, ZPM crosstalk suppression, state machine event detection, AI behavior pattern recognition, and chess piece fingerprint matching. After processing, the MCU sends the generated FEN string and key event data to the host computer in real time via a communication protocol (such as USB-CDC or Bluetooth).

[0051] b. Host computer data flow and decision-making: The host computer software is responsible for receiving the FEN string sent by the embedded terminal and parsing it.

[0052] UI Synchronization: The analyzed game information is synchronized to the UI interface for visual display.

[0053] AI Engine Interaction: Send the FEN value of the chess game to the Chinese Chess Cloud Library or a local AI engine (such as Stockfish) to receive the AI's decision.

[0054] Robotic arm control (if applicable): If the system includes a robotic arm, the host computer will convert the AI ​​decision (such as "Cannon 2 to 5") into mechanical coordinate instructions, and drive the robotic arm to perform physical actions such as picking up pieces, moving pieces, and placing pieces through a serial interface.

[0055] The entire process forms an intelligent closed loop of "player movement -> resistance positioning data optimization -> FEN generation -> AI decision -> (robotic arm execution)".

[0056] 6 Hardware Module Selection and Integration The data processing method of this invention relies on a carefully designed hardware architecture.

[0057] Microcontroller (MCU): A low-power microcontroller with high computing performance and rich I / O interfaces (such as STMicroelectronics' STM32 series) is selected. This chip is responsible for managing the scanning of the sensor array, data preprocessing, machine learning model inference, and communication with the host computer.

[0058] Sensor array and signal conditioning circuitry: A customized flexible variable resistor matrix is ​​integrated into the chessboard body. A signal conditioning circuit based on a zero-potential scanning method is deployed between the microcontroller and the sensor array to ensure the purity and accuracy of the acquired data.

[0059] Data conversion unit: integrates a multi-channel, high-resolution ADC chip. The ADC resolution (16-bit) is crucial for capturing subtle resistance changes in the "fingerprint" of the chess pieces, while the multi-channel design supports parallel scanning and improves data acquisition speed.

[0060] As described above, despite the numerous advantages of resistance sensing, its large-scale array application faces a major technical challenge: crosstalk. In a resistor array of M rows and N columns, when the system attempts to measure the resistance of a specific cell (e.g., the i-th row and j-th column), current may create unexpected current paths through adjacent, unmeasured sensing cells, leading to measurement distortion and errors. This error is particularly significant when the sensor resistance is low.

[0061] To address this issue, various solutions have been proposed in the industry. Among them, the Zero-Potential Scanning Method (ZPM) is widely used. The core idea of ​​ZPM is to drive the potential of all non-target rows and columns to 0V by using operational amplifiers or digital buffers, thereby physically cutting off all potential crosstalk current paths. This ensures that the current flows only through the target sensing unit being measured, significantly improving measurement accuracy. Based on this, this invention achieves low-cost integration of ZPM with large-scale arrays through optimized design.

[0062] The above technical solution also has the following characteristics: Distinguishing between hand and piece resistance patterns The resistance "fingerprints" left by human hands and chess pieces on the chessboard differ significantly in both space and time. This invention uses a machine learning model to extract and analyze these differences from sensor data, using them as key features to distinguish between the two.

[0063] Spatial distribution characteristics: When a hand touches the chessboard, the voltage change caused by the change in resistance is usually uneven and may span multiple adjacent positions. In contrast, the resistance pattern of a chess piece, due to its fixed base, is highly concentrated and predictable, typically affecting only the sensor at a single position. Machine learning models can easily distinguish between these two patterns by analyzing the resistance "heatmap" in the sensor matrix.

[0064] Temporal dynamics: When a hand manipulates a chess piece, the applied resistance signal exhibits a dynamic "grab-lift-move-release" curve. The placement process is a rapid "placement-stabilization" process, with the resistance signal quickly stabilizing after reaching a peak. The model can be trained to recognize these unique temporal patterns, thereby determining whether it is a brief hand contact or a complete placement action.

[0065] Resistance amplitude characteristics: The force applied by a human hand when grasping the piece is usually much greater than the weight of the piece itself, which provides another quantifiable basis for distinguishing between the two.

[0066] Machine learning model training and implementation This invention proposes using a supervised learning method to train a behavior pattern recognition model.

[0067] Data collection: First, a comprehensive training dataset was constructed by collecting sensor data samples with clear labels through a large number of experiments on the chessboard involving "human operation" and "piece placement".

[0068] Model selection: Given that resistance data has two-dimensional spatial and one-dimensional temporal characteristics, convolutional neural networks (CNNs) are an ideal choice, as they can effectively extract spatial features from resistance maps. In addition, classic classification algorithms such as support vector machines (SVMs) are also capable of handling this task.

[0069] Localized deployment: To achieve low-latency, real-time recognition, the trained lightweight model will be deployed on the microcontroller. This "on-chip intelligence" approach eliminates the need to send data to the cloud, reducing reliance on network bandwidth and significantly lowering computational power consumption.

[0070] Piece Location and Identification Once the behavior pattern recognition algorithm successfully determines an event as "piece placement" or "capture", the system will enter the piece recognition stage and eventually generate the FEN value of the game.

[0071] Chess piece type identification based on the distribution of voltage changes caused by changes in resistance value. This invention utilizes the "fingerprint" mechanism of chess pieces described in the text to identify their type.

[0072] Pattern matching: The system performs pattern matching between the resistance spectrum (or resistance spectrum) collected by the move position sensor and the "fingerprint" templates of 14 types of chess pieces (such as the red "general", the black "general", the red "cannon" etc.) pre-stored in the microcontroller.

[0073] Algorithm: By using algorithms such as Euclidean distance and correlation coefficient analysis, the best-matching template is found, thereby accurately identifying the type of chess piece. Due to the uniqueness of the chess piece's fingerprint, this identification method has extremely high accuracy; even if the chess pieces look similar, they can be distinguished through physical contact.

[0074] Chessboard coordinate mapping and FEN value generation After completing behavior recognition and piece type recognition, the system obtains the position and identity information of all pieces.

[0075] Coordinate mapping: The sensor row and column coordinates of the chess piece placement are mapped to standard Chinese chess board coordinates.

[0076] FEN Value Generation: Finally, the system generates a FEN string conforming to the Chinese Chess Cloud Database standard based on information such as the final layout of all pieces on the board and which player's turn it is. This standardized data format allows game information to be seamlessly integrated with any compatible AI engine or chess notation database.

[0077] Positioning accuracy and system verification Positioning Error Analysis and Suppression The inherent crosstalk problem in resistive sensor arrays can lead to distorted voltage measurements, thereby introducing positioning errors. To address this issue, this invention employs a zero-potential scanning (ZPM) method, which physically eliminates crosstalk currents by driving the potential on non-target sensor paths to zero, significantly improving measurement accuracy.

[0078] Nevertheless, other factors can still affect accuracy, such as: Secondary crosstalk: Components such as internal resistors in multiplexers can generate secondary crosstalk, which can affect the accuracy of sensor readings. Some studies have proposed neutralizing this secondary crosstalk by adding calibration columns or using algorithms such as least squares, thereby obtaining the true sensor readings.

[0079] Sensor non-uniformity: Even sensors from the same batch may exhibit slight differences in their electrical characteristics. This invention addresses this issue by performing offset and gain calibration in the digital domain after analog-to-digital conversion (ADC). This method ensures the reliability of chess piece "fingerprint" recognition by acquiring reference data in a piece-free state to eliminate background noise and using software algorithms to compensate for non-linear differences between sensors.

[0080] System Validation and Robustness Evaluation System functionality and performance verification are critical steps in ensuring its reliability. This invention primarily employs the following methods for verification: Machine Learning Model Validation: The effectiveness and generalization ability of a model are its core values. To ensure that the model can accurately distinguish between human hands and chess pieces, this invention will rigorously validate the model by dividing the dataset into training and test sets. Furthermore, methods such as k-fold cross-validation can be used to evaluate the model's stability and robustness on different subsets of data.

[0081] Physical Performance Verification: In addition to verifying the software algorithm, the physical performance of the hardware system is equally crucial. This includes testing the durability and stability of the sensor array to ensure it can withstand repeated piece placement and movement. For example, verifying whether the sensor's sensitivity remains within acceptable limits after 20,000 cycles of use. Simultaneously, the system also needs to verify its sensitivity to low resistance values ​​to ensure it can reliably detect pieces of varying weights, even slight touches.

[0082] Resistance scenario testing: By simulating real-world "resistance scenarios," such as players moving pieces rapidly, multiple players touching the board simultaneously, and irregularly placed pieces, the system's recognition accuracy under complex conditions is tested. These rigorous tests verify the system's robustness and reliability in the face of various uncertainties.

[0083] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0084] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A resistance positioning system based on a chessboard and chess pieces, characterized in that, include: An array of variable resistor sensors positioned below the chessboard is used to sense the distribution of voltage changes caused by changes in resistance values ​​at the chess pieces' positions. A differentiated physical structure is set on the bottom of each piece to form a unique resistance spectrum on the sensor array, serving as the piece's "fingerprint". The microcontroller is used to acquire the analog signals output by the sensor array and obtain digitized distribution data of voltage changes caused by changes in resistance value through analog-to-digital conversion. A behavior pattern recognition model deployed on the microcontroller is used to distinguish between human hand operations and chess piece placement events based on the spatial and temporal characteristics of the distribution data of voltage changes caused by changes in resistor value. The event-driven state machine module is used to update the game state in real time based on the identified behavioral events, including capturing, placing, taking pieces, and illegal moves; The chess piece recognition module is used to match the current resistance value spectrum with the pre-stored chess piece "fingerprint" template after confirming the chess piece placement event to identify the chess piece type; The coordinate mapping and FEN generation module is used to convert the position and identity of chess pieces into standard Chinese chess FEN strings for use by AI engines or chess notation systems.

2. The positioning method of the resistance positioning system according to claim 1, characterized in that, Includes the following steps: S1 uses a variable resistor sensor array deployed under the chessboard to collect in real time the distributed analog signal of voltage change caused by resistance change at each chess position, and obtains a digital distributed data matrix of voltage change caused by resistance change through analog-to-digital conversion. S2 uses a zero-potential scanning method to scan the sensor array in rows and columns. When measuring the target sensing unit, the potential of the non-target row and non-target column is driven to 0 V to physically cut off the crosstalk current path, thereby suppressing the crosstalk error of the resistor array. S3 inputs the distribution data of voltage change caused by resistance change into a pre-trained lightweight machine learning model deployed on a local microcontroller, extracts spatial distribution features and temporal dynamic features to distinguish between brief finger contact and real chess piece placement events, and outputs corresponding behavioral event labels. S4 drives a finite state machine based on the behavioral event tags to sequentially complete the event recognition and game state update of capturing, placing, taking or illegal moves of pieces; After confirming the move, S5 performs template matching between the distribution map of voltage change caused by the change in resistance value at the current position and the pre-stored 14 types of chess piece "fingerprint" templates to identify the chess piece type and generate a standard Chinese chess FEN string by combining the row and column coordinates of the position. S6 sends the FEN string to the host computer or AI engine in real time via USB or Bluetooth to enable human-computer chess playing or game record recording.

3. The positioning method according to claim 2, wherein the zero-potential scanning method adopts a single shared operational amplifier and multiplexer structure to scan all rows and columns sequentially, so as to reduce hardware cost and power consumption.

4. The localization method according to claim 2, wherein the lightweight machine learning model is an 8-bit quantized convolutional neural network or support vector machine, the model size is ≤50 kB, the inference time is ≤3 ms, and it runs directly on the microcontroller.

5. The positioning method according to claim 2, wherein the chess piece "fingerprint" template is obtained through at least one of the following methods: Each chess piece has a unique microstructure on its bottom, forming a spatially unique resistance spectrum on the sensor array; Each piece has a calibrated, minute weight difference, resulting in a distinguishable resistance amplitude characteristic.

6. The positioning method according to claim 2, wherein the template matching adopts an Euclidean distance or correlation coefficient algorithm, and the matching threshold is dynamically adjusted.

7. The positioning method according to claim 2, wherein after S1 and before S3, it further comprises: Digital domain offset and gain calibration steps are used to eliminate errors caused by sensor non-uniformity and temperature drift.

8. The positioning method according to claim 2, wherein the FEN string conforms to the Chinese Chess Cloud Database specification and is subjected to CRC verification after generation to ensure correct transmission.

9. The positioning method according to claim 2, wherein in S4, if multiple people are detected touching the game simultaneously or the pieces are placed overlapping, the state machine enters an error message mode and suspends FEN output until the error is resolved.

Images