Intelligent visual sitting posture correction table system based on STM32F103ZET6 development board

Abstract

The invention discloses an intelligent visual sitting posture correction table system based on an STM32F103ZET6 development board. The intelligent visual sitting posture correction table system comprises a visual perception module, a motion control module, a voice interaction module and a main control module, the visual perception module is used for collecting a user posture image through a camera and identifying key points by adopting an improved YOLOv11-MSCA model; the motion control module is used for generating a desktop adjustment instruction based on a fuzzy control algorithm and driving a stepping motor to execute multi-axis adjustment; the voice interaction module is used for receiving instructions through non-specific person voice recognition and outputting graded voice prompts; and the main control module is used for coordinating the operation of each module and realizing real-time task scheduling by adopting a dual-processor architecture. Through multi-sensor fusion (vision + voice + motion control) and a fuzzy PID composite algorithm, breakthrough improvement of key indexes such as response time, control precision and interactive experience is achieved while 2.4 kg lightweight design is kept.

Description

Technical Field

[0001] The present invention belongs to the field of health technology, and in particular relates to an intelligent visual posture correction table system based on an STM32F103ZET6 development board. Background Art

[0002] Currently available intelligent visual posture correction desk systems on the market suffer from numerous shortcomings, severely limiting their practical application and user experience. Regarding health risks, long-term use of existing posture correction technology can easily lead to health issues such as dependency, muscle fatigue, and impaired blood circulation. Regarding technical limitations, existing systems are mostly passive and unable to adapt to the user's real-time posture. They lack personalized adaptation features, making it difficult to accommodate different body types and usage habits. Posture recognition and feedback are not always real-time, preventing timely correction of poor posture. Adjustment accuracy is low, impacting the effectiveness of posture correction. Regarding user experience, the systems offer poor comfort, making prolonged use uncomfortable. Human-computer interaction is weak, the operation is complex, and users struggle to easily set up and adjust them. Application scenarios are limited, failing to meet diverse needs. The price-performance ratio is mismatched, preventing users from receiving a quality experience commensurate with the price. Regarding design flaws, the systems lack ergonomic design, failing to provide a more comfortable user experience. They lack continuous monitoring capabilities, failing to track changes in posture in real time. Their single function set fails to meet user demands for multi-functionality. Furthermore, they lack child-specific features, failing to meet the needs of children. These defects not only affect the practicality and market acceptance of the product, but also highlight the urgency of improving the existing technology. It is urgent to propose an intelligent visual posture correction table system based on the STM32F103ZET6 development board. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes an intelligent visual sitting posture correction table system based on the STM32F103ZET6 development board, which can accurately identify the left and right and front and back tilt states of the user's sitting posture.

[0004] To achieve the above objectives, the present invention provides an intelligent visual posture correction table system based on the STM32F103ZET6 development board, comprising:

[0005] Visual perception module, motion control module, voice interaction module and main control module;

[0006] The visual perception module is used to collect user posture images through a camera and identify key points using an improved YOLOv11-MSCA model;

[0007] The motion control module is used to generate desktop adjustment instructions based on the fuzzy control algorithm and drive the stepper motor to perform multi-axis adjustment;

[0008] The voice interaction module is used to receive instructions through non-specific speaker voice recognition and output graded voice prompts;

[0009] The main control module is used to coordinate the operation of each module and adopts a dual-processor architecture to achieve real-time task scheduling.

[0010] Optionally, the visual perception module includes: an image acquisition unit, a key point extraction unit and a classification decision unit;

[0011] The image acquisition unit is used to capture a resolution image of a preset size using an OV5647 camera;

[0012] The key point extraction unit is used to extract a number of posture key points through the MediaPipe algorithm;

[0013] The classification decision unit is used to determine the sitting tilt state by using the YOLOv11 model with an added MSCA layer.

[0014] Optionally, the motion control module includes: a fuzzy control unit, a motor drive unit and a safety protection unit;

[0015] The fuzzy control unit is used to construct a 7-level quantization system with error, error change rate and historical adjustment amount as input;

[0016] The motor drive unit is used to implement 1 / 16 micro-step control of the 42 stepper motor through the A4988 chip;

[0017] The safety protection unit is used to integrate overheat protection and undervoltage lockout functions.

[0018] Optionally, the voice interaction module includes: a command recognition unit, a voice synthesis unit and an interaction logic unit;

[0019] The command recognition unit is used to support 50 voice command recognitions using the LD3320 chip;

[0020] The speech synthesis unit is used to realize text-to-speech output through the SYN6288 module;

[0021] The interactive logic unit is used to configure a three-level response strategy including reminder, warning and forced reset.

[0022] Optionally, the main control module includes: a visual computing unit, a real-time control unit and a communication relay unit;

[0023] The visual computing unit is used to run the gesture recognition algorithm using a Raspberry Pi 5 processor;

[0024] The real-time control unit is used to implement microsecond-level task switching based on STM32F103ZET6;

[0025] The communication relay unit is used to establish a serial port protocol to realize data transmission between modules.

[0026] Optionally, the key point extraction unit further includes: an anti-shake compensation subunit and a background filtering subunit;

[0027] The anti-shake compensation subunit is used to eliminate ±5° angular deviation through the dual-axis optical anti-shake module;

[0028] The background filtering subunit is used to separate the human body feature area using the HSV color space.

[0029] Optionally, the fuzzy control unit further includes: a rule base management subunit and a parameter self-learning subunit;

[0030] The rule base management subunit is used to store 25 control rules to implement Mamdani reasoning;

[0031] The parameter self-learning subunit is used to dynamically adjust the membership function according to user historical data.

[0032] Optionally, the speech synthesis unit further includes: a multi-language support subunit and a parameter adjustment subunit;

[0033] The multi-language support sub-unit is used to be compatible with Chinese, English and Japanese voice output;

[0034] The parameter adjustment subunit is used to independently control volume, speech speed and pitch parameters.

[0035] Technical effect of the present invention: The present invention discloses an intelligent visual posture correction table system based on the STM32F103ZET6 development board, which realizes adaptive adjustment of the desktop by precisely controlling the 42 stepper motors through the A4988 driver module, and realizes voice interaction function by combining the LD3320 voice recognition module and the SYN6288 voice synthesis module. The fuzzy control algorithm is innovatively adopted, and a 7-level quantitative control system is constructed through the three input variables of visual recognition error, error change rate and historical adjustment amount to achieve precise adjustment with a response time of <1 second and an overshoot of <5%. The system has functions such as intelligent sitting posture recognition, automatic adjustment of desktop height / angle, graded voice reminders and user habit learning. Through multi-sensor fusion and fuzzy PID control, it achieves a natural and smooth human-computer interaction experience while ensuring real-time performance (task switching at the microsecond level). BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0037] Figure 1 This is a schematic diagram of a visual recognition module according to an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of a Raspberry Pi 5 according to an embodiment of the present invention;

[0039] Figure 3 This is the onboard optical image stabilization module of an embodiment of the present invention;

[0040] Figure 4 A4988 driver chip according to an embodiment of the present invention;

[0041] Figure 5 This is a typical application diagram of A4988 according to an embodiment of the present invention;

[0042] Figure 6 This is a typical application diagram of HR4988 according to an embodiment of the present invention;

[0043] Figure 7 The LD3320 speech recognition chip according to an embodiment of the present invention;

[0044] Figure 8 This is a circuit diagram of the LD3320 speech recognition module according to an embodiment of the present invention;

[0045] Figure 9 This is a schematic diagram of the SYN6288 speech synthesis module according to an embodiment of the present invention;

[0046] Figure 10 This is a circuit diagram of the FREERTOS real-time operating system according to an embodiment of the present invention;

[0047] Figure 11 This is a structural diagram of an intelligent visual posture correction table system based on the STM32F103ZET6 development board according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0049] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0050] like Figure 11 As shown, this embodiment provides an intelligent visual posture correction table system based on the STM32F103ZET6 development board, including:

[0051] Visual perception module, motion control module, voice interaction module and main control module;

[0052] The visual perception module is used to collect user posture images through a camera and identify key points using an improved YOLOv11-MSCA model;

[0053] The motion control module is used to generate desktop adjustment instructions based on the fuzzy control algorithm and drive the stepper motor to perform multi-axis adjustment;

[0054] The voice interaction module is used to receive instructions through non-specific speaker voice recognition and output graded voice prompts;

[0055] The main control module is used to coordinate the operation of each module and adopts a dual-processor architecture to achieve real-time task scheduling.

[0056] Furthermore, the visual perception module includes: an image acquisition unit, a key point extraction unit and a classification decision unit;

[0057] The image acquisition unit is used to capture a resolution image of a preset size using an OV5647 camera;

[0058] The key point extraction unit is used to extract a number of posture key points through the MediaPipe algorithm;

[0059] The classification decision unit is used to determine the sitting tilt state by using the YOLOv11 model with an added MSCA layer.

[0060] Furthermore, the motion control module includes: a fuzzy control unit, a motor drive unit and a safety protection unit;

[0061] The fuzzy control unit is used to construct a 7-level quantization system with error, error change rate and historical adjustment amount as input;

[0062] The motor drive unit is used to implement 1 / 16 micro-step control of the 42 stepper motor through the A4988 chip;

[0063] The safety protection unit is used to integrate overheat protection and undervoltage lockout functions.

[0064] Furthermore, the voice interaction module includes: a command recognition unit, a voice synthesis unit and an interaction logic unit;

[0065] The command recognition unit is used to support 50 voice command recognitions using the LD3320 chip;

[0066] The speech synthesis unit is used to realize text-to-speech output through the SYN6288 module;

[0067] The interactive logic unit is used to configure a three-level response strategy including reminder, warning and forced reset.

[0068] Furthermore, the main control module includes: a visual computing unit, a real-time control unit and a communication relay unit;

[0069] The visual computing unit is used to run the gesture recognition algorithm using a Raspberry Pi 5 processor;

[0070] The real-time control unit is used to implement microsecond-level task switching based on STM32F103ZET6;

[0071] The communication relay unit is used to establish a serial port protocol to realize data transmission between modules.

[0072] Furthermore, the key point extraction unit further includes: an anti-shake compensation subunit and a background filtering subunit;

[0073] The anti-shake compensation subunit is used to eliminate ±5° angular deviation through the dual-axis optical anti-shake module;

[0074] The background filtering subunit is used to separate the human body feature area using the HSV color space.

[0075] Furthermore, the fuzzy control unit further comprises: a rule base management subunit and a parameter self-learning subunit;

[0076] The rule base management subunit is used to store 25 control rules to implement Mamdani reasoning;

[0077] The parameter self-learning subunit is used to dynamically adjust the membership function according to user historical data.

[0078] Furthermore, the speech synthesis unit further includes: a multi-language support subunit and a parameter adjustment subunit;

[0079] The multi-language support sub-unit is used to be compatible with Chinese, English and Japanese voice output;

[0080] The parameter adjustment subunit is used to independently control volume, speech speed and pitch parameters.

[0081] like Figure 1As shown, the FreeRTOS real-time operating system is used for multi-task collaborative control. The system uses the A4988 driver module to precisely control 42-stepper motors for adaptive desktop adjustment. It integrates the LD3320 speech recognition module and the SYN6288 speech synthesis module to implement voice interaction. The system innovatively employs a fuzzy control algorithm, using three input variables: visual recognition error, error change rate, and historical adjustment value. This constructs a seven-level quantitative control system, achieving precise adjustment with a response time of less than 1 second and an overshoot of less than 5%. The system features intelligent posture recognition, automatic desktop height / angle adjustment, graded voice reminders, and user habit learning. Through multi-sensor fusion and fuzzy PID control, it ensures real-time performance while delivering a natural and smooth human-computer interaction experience (task switching at the microsecond level).

[0082] like Figure 2 As shown, the visual recognition module uses a Raspberry Pi 5 as the core computing platform, paired with an OV5647 camera for high-precision image acquisition. Combined with an improved YOLOv11-MSCA model and MediaPipe pose estimation technology, it can accurately identify the user's left-right and forward-backward tilt while sitting. The system utilizes a multi-threaded design for efficient image processing and serial communication, transmitting classification results to the main control unit in real time, providing data support for subsequent intelligent intervention. This module features low power consumption and high real-time performance, making it suitable for health management needs in long-term sitting scenarios such as office and education settings. During the development phase, we selected the Raspberry Pi 5 as the visual computing module after comprehensively considering factors such as development speed, cost, and platform computing power. Powered by a 64-bit quad-core Arm Cortex-A76 processor running at 2.4GHz, the Raspberry Pi 5 offers powerful computing power and is capable of handling general image signal processing tasks.

[0083] like Figure 3 As shown, the camera uses the OV5647, a MIPI-CSI camera with 5 megapixels and a 70° field of view. It also has an onboard optical image stabilization module that supports dual-axis image stabilization and a ±5° compensation angle.

[0084] A4988 drives 42 stepper motor:

[0085] The A4988 is a highly integrated microstepping motor driver chip with built-in stepper control logic and current regulator. It can easily drive bipolar stepper motors and supports full-step, half-step, 1 / 4, 1 / 8, and 1 / 16 microstepping modes. With a drive capability of 35V / ±1A, it is suitable for a variety of low- and medium-power stepper motor control scenarios.

[0086] During the design process, we chose to use a 42mm stepper motor paired with the A4988 driver module to control the tabletop's movement, allowing it to adapt to different postures. Everyone has different sitting habits, and adaptive adjustment can accommodate different people, making it easier for people to use the table in their daily work.

[0087] like Figure 4 As shown in the figure, the A4988 driver chip, through its intelligent micro-step control technology, can achieve precise driving in five micro-step modes (from full step to 1 / 16 step) with only a simple pulse signal, significantly improving the smoothness and positioning accuracy of the motor operation. The chip adopts innovative adaptive current regulation technology, integrates a fixed off-time regulator and a mixed decay mode, and optimizes energy efficiency while reducing operating noise and vibration. Its efficient power management system integrates synchronous rectification PWM control and multiple protection mechanisms (including overheating protection, undervoltage lockout, etc.), and is equipped with an 8-35V wide voltage input design. While ensuring system safety and stability, it provides a highly simplified drive solution for embedded applications, greatly reducing development complexity and hardware resource requirements. As shown in the figure, the A4988 driver chip can achieve precise driving in five micro-step modes (from full step to 1 / 16 step) with only a simple pulse signal, significantly improving the smoothness and positioning accuracy of the motor operation. The chip adopts innovative adaptive current regulation technology, integrates a fixed off-time regulator and a mixed decay mode, and optimizes energy efficiency while reducing operating noise and vibration. Its efficient power management system integrates synchronous rectification PWM control and multiple protection mechanisms (including overheating protection, undervoltage lockout, etc.), and combines it with an 8-35V wide voltage input design. While ensuring system safety and stability, it provides a highly simplified drive solution for embedded applications, greatly reducing development complexity and hardware resource requirements. Figure 5-Figure 6 They are the typical application diagrams of A4988 and HR4988 respectively.

[0088] like Figure 7-Figure 8 As shown, the LD3320 speech recognition module:

[0089] Developed by ICRoute, the LD3320 is an intelligent voice processing chip based on speaker-independent speech recognition technology. This chip utilizes advanced speech recognition algorithms to achieve highly accurate voice interaction without requiring pre-training. Its innovative design integrates multiple functional modules, providing a complete solution for embedded voice applications.

[0090] This module achieves up to 95% recognition accuracy without pre-training. The chip integrates a 16-bit AD / DA converter and audio amplifier circuitry, supports dynamic vocabulary management, and supports real-time recognition of up to 50 commands, each containing 10 Chinese characters. Through its flexible parallel / serial interface and 3.3V low-power design, the LD3320 can directly connect to a variety of audio devices. Its built-in MP3 decoding function provides a plug-and-play voice interaction solution for smart homes, consumer electronics, and other applications. Its innovative architecture simplifies system design while achieving excellent voice recognition performance.

[0091] During the design process, the voice recognition module is used to perform some voice recognition control functions on the desktop. For example, voice control is used to move the desktop, mainly in six directions: forward, backward, left, right, up, and down. Voice commands are also used to control the on and off of LED lights.

[0092] like Figure 9-10 As shown, the SYN6288 speech synthesis module:

[0093] The SYN6288 is an intelligent speech synthesis module based on a high-performance DSP architecture. It utilizes advanced digital signal processing technology to convert text information into natural and fluent speech output in real time. This module excels in speech clarity, adjustability, and system integration, making it suitable for a variety of voice interaction scenarios.

[0094] The SYN6288 speech synthesis module, with its superior multi-language support (including Chinese, English, and Japanese) and highly adjustable voice parameters (volume, speaking rate, and pitch), provides users with a high-quality speech output experience that is close to real-person pronunciation. The module uses a high-fidelity digital synthesis algorithm and seamlessly integrates with mainstream MCU platforms via a standard UART interface. Its low-power architecture and miniaturized package design are particularly suitable for embedded system integration. While ensuring clear and natural speech, it also takes into account system power consumption and space efficiency, making it an ideal voice solution for enhancing the human-computer interaction experience of smart devices.

[0095] The SYN6288 speech synthesis module is used to broadcast the real-time status, such as sedentary reminders, bad sitting posture reminders, etc. In the initialization state, it will have the function of broadcasting prompts for the opening and closing of voice control recognition.

[0096] FreeRTOS is an open-source real-time operating system kernel designed specifically for embedded systems. It features lightweight, high reliability, and scalability. This system uses FreeRTOS as the underlying task scheduling core to ensure real-time response and stable operation of each functional module.

[0097] FreeRTOS, the real-time operating system kernel for this system, provides a comprehensive multitasking management mechanism, supporting preemptive scheduling and multiple priority levels. It also implements task coordination through a rich set of inter-process communication methods, such as queues, semaphores, and mutexes. The system boasts excellent real-time performance, with task switching times down to microseconds. It also provides configurable clock cycles and priority inheritance mechanisms to optimize system responsiveness. Regarding resource management, FreeRTOS supports both static and dynamic memory allocation, with the smallest kernel requiring only 6-12KB of ROM and 1KB of RAM. It also incorporates security features such as stack overflow detection. Its exceptional hardware compatibility covers over 40 processor architectures, and it is specifically optimized for the ARM Cortex-M series processors. It provides standardized driver interface templates for embedded development, ensuring efficient and stable system operation even in resource-constrained environments.

[0098] In the Vdesk intelligent visual posture correction table system, FreeRTOS modularizes core functions into five priority tasks through a detailed task partitioning, with motor control being given the highest priority to ensure real-time response. The system uses message queues to efficiently transmit voice commands to motor actions, protects key resources through mutex locks, and optimizes event processing using software timers and task notification mechanisms. Through optimization measures such as static memory allocation and stack statistics, the system improves operational stability while ensuring real-time performance. The introduction of FreeRTOS not only enables smooth coordination between voice interaction and motor control, but also, through its lightweight nature (requiring only 6KB of ROM at the minimum), enables the system to maintain excellent performance in resource-constrained embedded environments. This provides a solid foundation for product function expansion and maintenance upgrades, fully demonstrating its unique advantages in intelligent hardware development.

[0099] Fuzzy control algorithm:

[0100] Fuzzy control algorithms are intelligent control methods based on fuzzy logic, particularly suitable for controlling systems with nonlinear and time-varying characteristics. In this project, fuzzy control algorithms are used to implement intelligent adjustments for posture correction. By simulating the human decision-making process, precise control of the desktop height and tilt angle is achieved.

[0101] This system employs an intelligent fuzzy control algorithm. Using visual recognition of three key input variables—error (e), error change rate (Δe), and historical adjustment value (u_prev)—and combining triangular and trapezoidal membership functions, it constructs a seven-level quantitative control system consisting of five fuzzy sets (NB to PB). Based on the Mamdani inference method, the system implements a multi-conditional control rule base, including "IF eis NB AND Δeis NB THEN Δuis PB," and intelligent decision-making through MAX-MIN synthesis. At the output stage, the center of gravity method is used for defuzzification, converting the fuzzy value into a precise number of motor adjustment pulses (Δu) and voice prompt level. Intelligent features such as adaptive parameter adjustment, user habit learning, and environmental compensation ensure a more natural and smooth desktop adjustment process. The entire control process comprises three core steps: fuzzification processing, rule inference, and defuzzification. Output limiting ensures stable system operation.

[0102] The intelligent visual posture correction system demonstrates excellent overall performance: In terms of control performance, the system response time is less than 1 second, overshoot is controlled within 5%, and steady-state error is less than 2%, ensuring fast and accurate posture correction. In terms of adaptability, the system intelligently identifies different user body types, dynamically optimizes control parameters, and effectively resists environmental interference. In terms of user experience, the system achieves smooth, jitter-free adjustment, coupled with real-time, accurate voice prompts and personalized memory functions, providing users with a natural and comfortable human-computer interaction experience. These advantages collectively ensure the product's reliability and user satisfaction in real-world scenarios.

[0103] Through the fuzzy control algorithm, the system can: intelligently identify bad sitting postures (such as hunchback, forward leaning, etc.); automatically adjust the desktop to the optimal height and angle; provide graded voice reminders based on usage; learn user habits and optimize control parameters; this algorithm significantly improves the product's intelligence level and user experience, making the correction process more natural and comfortable, and avoiding the mechanical feel and discomfort that may be brought about by traditional PID control.

[0104] Raspberry Pi 5:

[0105] Dataset Preparation: We collected data from people sitting in front of a desk in various postures and divided it into two groups: one for training the model to recognize left-right tilt and the other for training the model to recognize front-back tilt. The dataset contains approximately 1,600 images, 15% of which are used as a test set.

[0106] For sitting posture recognition, we only need to identify the human face, shoulders, and the relationship between the two. It is not sensitive to information such as the user's appearance, clothing, and background. To improve the accuracy of the model, we use MediaPipe to pre-extract the human posture in the image to obtain the final training and test sets.

[0107] Model training: To balance the impact of small-scale key points (such as facial key points) and blurred large-scale structures (such as shoulders), and to adapt to application scenarios with variable target scales, we added the MSCA layer to the YOLOv11 model.

[0108] The Raspberry Pi 5 first extracts video frames through OpenCV, uses MediaPipe to extract pose key points from the video frames, and finally uses the pre-trained YOLOv11-MSCA model for classification and recognition. The classification results are then sent to the main control chip using the serial port.

[0109] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. The intelligent visual posture correction table system based on the STM32F103ZET6 development board is characterized by: include: Visual perception module, motion control module, voice interaction module and main control module; The visual perception module is used to collect user posture images through a camera and identify key points using an improved YOLOv11-MSCA model; The motion control module is used to generate desktop adjustment instructions based on the fuzzy control algorithm and drive the stepper motor to perform multi-axis adjustment; The voice interaction module is used to receive instructions through non-specific speaker voice recognition and output graded voice prompts; The main control module is used to coordinate the operation of each module and adopts a dual-processor architecture to achieve real-time task scheduling.

2. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 1, characterized in that, The visual perception module includes: an image acquisition unit, a key point extraction unit and a classification decision unit; The image acquisition unit is used to capture a resolution image of a preset size using an OV5647 camera; The key point extraction unit is used to extract a number of posture key points through the MediaPipe algorithm; The classification decision unit is used to determine the sitting tilt state by using the YOLOv11 model with an added MSCA layer.

3. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 1, characterized in that, The motion control module includes: a fuzzy control unit, a motor drive unit and a safety protection unit; The fuzzy control unit is used to construct a 7-level quantization system with error, error change rate and historical adjustment amount as input; The motor drive unit is used to implement 1 / 16 micro-step control of the 42 stepper motor through the A4988 chip; The safety protection unit is used to integrate overheat protection and undervoltage lockout functions.

4. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 1, characterized in that, The voice interaction module includes: a command recognition unit, a voice synthesis unit and an interaction logic unit; The command recognition unit is used to support 50 voice command recognitions using the LD3320 chip; The speech synthesis unit is used to realize text-to-speech output through the SYN6288 module; The interactive logic unit is used to configure a three-level response strategy including reminder, warning and forced reset.

5. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 1, characterized in that, The main control module includes: a visual computing unit, a real-time control unit and a communication relay unit; The visual computing unit is used to run the gesture recognition algorithm using a Raspberry Pi 5 processor; The real-time control unit is used to implement microsecond-level task switching based on STM32F103ZET6; The communication relay unit is used to establish a serial port protocol to realize data transmission between modules.

6. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 2, characterized in that, The key point extraction unit further includes: an anti-shake compensation subunit and a background filtering subunit; The anti-shake compensation subunit is used to eliminate ±5° angular deviation through the dual-axis optical anti-shake module; The background filtering subunit is used to separate the human body feature area using the HSV color space.

7. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 3, characterized in that, The fuzzy control unit further comprises: a rule base management subunit and a parameter self-learning subunit; The rule base management subunit is used to store 25 control rules to implement Mamdani reasoning; The parameter self-learning subunit is used to dynamically adjust the membership function according to user historical data.

8. The intelligent visual posture correction table system based on the STM32F103ZET6 development board as claimed in claim 4, characterized in that, The speech synthesis unit further includes: a multi-language support subunit and a parameter adjustment subunit; The multi-language support sub-unit is used to be compatible with Chinese, English and Japanese voice output; The parameter adjustment subunit is used to independently control volume, speech speed and pitch parameters.

Images