Sitting posture detection and interaction system and method based on piezoresistive array cushion and table lamp

Abstract

The invention relates to the technical field of sitting posture recognition, in particular to a sitting posture detection and interaction system and method based on a piezoresistive array cushion and a table lamp, and the system comprises an intelligent table lamp and the piezoresistive array cushion used for obtaining sitting posture information; the piezoresistive array cushion comprises a sensing array and a first controller electrically connected with the sensing array; the intelligent table lamp comprises a data processing platform used for receiving signals of the first controller, a second controller used for receiving signals of the data processing platform, and an illuminating lamp, a brightness indicating lamp and a gravity center indicating lamp ring which are electrically connected with the second controller. Sitting posture detection and information interaction are completed through the piezoresistive array cushion and the intelligent table lamp, the intelligent cushion system is achieved, the system can monitor various sitting postures of a user at high precision through the piezoresistive array cushion, and the user is reminded to adjust the sitting postures in real time through a visual feedback mechanism of the table lamp end; in addition, the system also has a sedentariness reminding function, so that the sitting posture health of the user is effectively managed.

Description

Technical Field

[0001] The present invention relates to the technical field of sitting posture recognition, and in particular to a sitting posture detection and interaction system and method based on a piezoresistive array cushion and a desk lamp. Background Art

[0002] Most existing smart cushions use rigid pressure sensors. Although these sensors can provide certain pressure detection functions, their rigidity will affect the comfort of the cushions, and long-term use will cause users to feel a foreign body. In addition, although wearable devices can obtain movement information of specific parts of the user, they require multiple sensors to be worn and fixed, which is cumbersome to use, and long-term wearing will cause discomfort, resulting in a poor user experience.

[0003] Common external devices are usually desktop devices, such as ultrasonic sensors or cameras. These devices have functional disadvantages such as limited monitoring range, occupying available desktop space, and unfriendly user interface. Secondly, image recognition through cameras also has the risk of infringing user privacy. Users will feel uncomfortable and resistant to being monitored during use, which also limits the application scenarios of the device.

[0004] In addition, existing sitting posture monitoring devices often lack real-time feedback functions and cannot promptly remind users to adjust bad sitting postures. In addition, existing feedback methods such as buzzer alarms and vibrations are too obvious and easily interfere with users' work or study, affecting their efficiency and concentration.

[0005] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0006] In view of the above-mentioned deficiencies in the prior art, the object of the present invention is to provide a sitting posture detection and interaction system and method based on a piezoresistive array cushion and a desk lamp, aiming to solve the problems that the existing sitting posture detection system cannot achieve non-intrusive and effective real-time feedback, and has poor comfort.

[0007] The technical solution of the present invention is as follows: A sitting posture detection and interaction system based on a piezoresistive array cushion and a desk lamp, comprising an intelligent desk lamp and a piezoresistive array cushion for obtaining sitting posture information; The piezoresistive array cushion includes a sensor array and a first controller electrically connected to the sensor array; The intelligent desk lamp comprises: a data processing platform for receiving signals from the first controller and a second controller for receiving signals from the data processing platform, as well as a lighting lamp, a brightness indicator light and a center of gravity indicator light ring electrically connected to the second controller respectively.

[0008] The sitting posture detection and interaction system based on the piezoresistive array cushion and the desk lamp, wherein the piezoresistive array cushion also includes an upper protective film of the pressure pad and a lower protective film of the pressure pad sandwiching the sensor array in the middle; the sensor array includes a plurality of sensor units arranged in a matrix, and the plurality of the sensor units are connected in series through conductive wires.

[0009] The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp, wherein the sensing unit is a pressure-sensitive conductive sheet Velostat.

[0010] The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp, wherein the smart desk lamp also includes a base, a display box arranged on the base, and a lampshade box connected to the display box; the brightness indicator light and the center of gravity indicator light ring are arranged on the display box, and the lighting lamp is arranged in the lampshade box.

[0011] The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp, wherein the base includes a base lower cover and a base upper cover matching the base lower cover; a lighting switch is provided on the base upper cover, and the lighting switch is connected to the second controller for controlling the brightness of the lighting.

[0012] The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp, wherein the display box includes a display box back cover and a display box front cover matching the display box back cover; the brightness indicator light and the center of gravity indicator light ring are arranged on the display box front cover.

[0013] The sitting posture detection and interaction system based on the piezoresistive array seat cushion and desk lamp, wherein the lampshade box includes a lampshade upper cover and a lampshade lower cover matching the lampshade upper cover; the lampshade lower cover is provided with a lighting window, and a transparent lampshade is provided at the lighting window; the lampshade box is connected to the display box through a hollow hose.

[0014] A sitting posture detection and interaction method based on a piezoresistive array cushion and a desk lamp comprises the following steps: The original pressure data is collected by using a piezoresistive array cushion and normalized to obtain multiple sets of original pressure data sets; Using the original pressure data set to train a neural network model to obtain a sitting posture recognition model; Using the piezoresistive array cushion to collect real-time pressure data, the data processing platform is sent to the data processing platform for storage, and the data processing platform simultaneously sends the real-time pressure data to the sitting posture recognition model and outputs the sitting posture type; The sitting posture type is fed back through the lighting effect of the smart desk lamp.

[0015] The sitting posture detection and interaction method based on the piezoresistive array cushion and the desk lamp, wherein the step of collecting raw pressure data using the piezoresistive array cushion and performing normalization processing to obtain multiple groups of raw pressure data sets includes: The sensor array composed of 9×9 sensor units collects the original pressure data and performs normalization processing to obtain 9×9 matrix data; Performing a pooling operation on the 9×9 matrix data, including 2×2 pooling and 3×3 pooling; The 2×2 pooling and 3×3 pooling are respectively operated to obtain a 4×4 feature map and a 3×3 feature map; The 4×4 feature map and the 3×3 feature map are flattened and concatenated into a comprehensive feature vector, that is, a set of original pressure data sets.

[0016] The method for sitting posture detection and interaction based on a piezoresistive array cushion and a desk lamp, wherein the step of using the original pressure data set to train a neural network model to obtain a sitting posture recognition model comprises: Inputting the comprehensive feature vector into a fully connected neural network model for training to obtain a sitting posture recognition model; Among them, the fully connected neural network model includes a first fully connected layer, a second fully connected layer and a third fully connected layer; the first fully connected layer has 128 neurons, and the activation function is ReLU; the second fully connected layer has 64 neurons, and the activation function is ReLU; the third fully connected layer has 32 neurons, and the activation function is ReLU; the output layer of the fully connected neural network model has 7 neurons, and the activation function is Softmax.

[0017] Beneficial effects: The present invention provides a sitting posture detection and interaction system and method based on a piezoresistive array cushion and a desk lamp. The sitting posture detection and interaction system based on a piezoresistive array cushion and a desk lamp includes an intelligent desk lamp and a piezoresistive array cushion for obtaining sitting posture information; the piezoresistive array cushion includes a sensor array and a first controller electrically connected to the sensor array; the intelligent desk lamp includes: a data processing platform for receiving signals from the first controller and a second controller for receiving signals from the data processing platform, as well as a lighting lamp, a brightness indicator light and a center of gravity indicator light ring electrically connected to the second controller, respectively. The present invention utilizes a piezoresistive array cushion and an intelligent desk lamp to complete sitting posture detection and information interaction, realizing an intelligent seat cushion system, which can monitor various sitting postures of users with high precision through the piezoresistive array cushion, and can remind users to adjust their sitting postures in real time through a visual feedback mechanism on the desk lamp end; and the system also has a sedentary reminder function to effectively manage the user's sitting posture health. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1It is a schematic diagram of the structure of the piezoresistive array cushion in the present invention; Figure 2 It is a structural schematic diagram of the intelligent desk lamp in the present invention; Figure 3 This is a schematic diagram of the exploded structure of the smart desk lamp in the present invention; Figure 4 It is a schematic diagram of the working principle of the sitting posture detection and interaction system based on the piezoresistive array cushion and the desk lamp of the present invention; Figure 5 A schematic diagram of the flow of the sitting posture detection and interaction process performed by the sitting posture detection and interaction system based on the piezoresistive array cushion and the desk lamp of the present invention; Figure 6 This is a schematic diagram of the principle of the multi-scale splicing multi-layer perceptron method; Figure 7 This is a data chart of real-time detection accuracy of 7 sitting postures; Explanation of the reference numerals: sensor array 10, pressure pad upper protective film 11, pressure pad lower protective film 12, sensor unit 13, conductive wire 14, first controller 20, lighting lamp 30, brightness indicator light 40, center of gravity indicator light ring 50, base 60, base lower cover 61, base upper cover 62, lighting lamp switch 63, display box 70, display box back cover 71, display box front cover 72, lampshade box 80, lampshade upper cover 81, lampshade lower cover 82, transparent lampshade 83, hollow hose 90, tooth tube 100, type-c charging module 200. DETAILED DESCRIPTION

[0019] The present invention provides a sitting posture detection and interaction system and method based on a piezoresistive array cushion and a desk lamp. In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail below. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "back" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the said features.

[0021] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0022] There are many existing patents that use smart cushions to monitor sitting posture, but the pressure sensors used are mostly rigid sensors. Although the cushion is covered with soft materials, it will also make the user feel a foreign body. In addition, most smart cushions lack feedback, and there is no sitting posture reminder after sitting posture monitoring. Patented non-invasive sitting posture monitoring smart cushion (CN204743303U), the cushion body is embedded with multiple piezoelectric sensors, after signal processing, the microprocessor analyzes the data to determine the user's sitting posture. Patented array-type smart cushion for sitting posture monitoring (CN215838064U), the upper part of the cushion body is provided with multiple inlay holes, the force-bearing member is inlaid in the inlay holes, and the upper surface of the force-bearing member protrudes from the upper surface of the cushion body. Both patents use stiffness sensors, which may potentially cause discomfort to users during trial use; at the same time, the two focus on the realization of the sitting posture monitoring function, without considering the data delivery after monitoring and the interaction function with other devices. Due to the lack of feedback to users, users are unable to adjust bad sitting postures in time. Sitting for long periods of time also has negative effects on human health, and most of the existing patents with sitting posture monitoring functions also lack reminder functions for sitting for long periods of time.

[0023] In addition, in terms of sitting posture reminders, existing sitting posture reminders (such as CN102289917A, CN219758932U) use buzzer alarms, warning lights, vibrations and other methods to remind users of bad sitting postures. In fact, this method can only provide feedback to users on bad sitting postures, including hunchbacks, lowered heads and other small variable sitting posture classifications. It is unable to intuitively inform users of more specific bad sitting postures and correction suggestions, causing users to not know how to adjust to the correct sitting posture. More importantly, because reminders such as buzzer alarms, warning lights, and vibrations are too obvious, they can easily cause additional interference to the user's current behavior or the current activities of the scene, greatly reducing the efficiency and concentration of users in performing focused activities in a certain scene.

[0024] Based on this, Figure 1-Figure 3 As shown, the present invention provides a sitting posture detection and interaction system based on a piezoresistive array cushion and a desk lamp, comprising an intelligent desk lamp and a piezoresistive array cushion for obtaining sitting posture information; The piezoresistive array cushion includes a sensor array 10 and a first controller 20 electrically connected to the sensor array 10; The smart desk lamp includes: a data processing platform for receiving the first controller signal 20 and a second controller for receiving the data processing platform signal, and a lighting lamp 30, a brightness indicator light 40 and a center of gravity indicator light ring 50 respectively electrically connected to the second controller.

[0025] In this embodiment, a piezoresistive array cushion and an intelligent desk lamp are used to complete sitting posture detection and information interaction, thereby realizing an intelligent cushion system. The system can monitor the user's various sitting postures with high precision through the piezoresistive array cushion, and remind the user to adjust the sitting posture in real time through the visual feedback mechanism of the desk lamp. In addition, the system also has a sedentary reminder function to effectively manage the user's sitting posture health.

[0026] Specifically, the sensor array 10 is electrically connected to the first controller 20. When a user sits on the piezoresistive array cushion, the sensor array can detect the user's pressure distribution in real time and generate corresponding pressure data by collecting the voltage change of each unit; that is, the first controller is responsible for scanning the sensor array line by line, reading the voltage value of each unit, and converting it into pressure data.

[0027] In some embodiments, the piezoresistive array cushion further includes a pressure pad upper protective film 11 and a pressure pad lower protective film 12 sandwiching the sensor array 10 in the middle; the sensor array 10 includes a plurality of sensor units 13 arranged in a matrix, and the plurality of sensor units are connected in series through conductive wires 14. The pressure data of different parts of the user on the piezoresistive array cushion can be obtained by using the plurality of sensor units, and the plurality of sensor units are connected in series through conductive wires, which facilitates the acquisition of pressure data and reduces wiring; in addition, the pressure pad upper protective film and the pressure pad lower protective film are arranged on both sides of the sensor array, which can protect the sensor array and extend the service life of the cushion.

[0028] In a preferred embodiment, the sensor array is composed of 9×9 sensor units, each of which can independently detect pressure changes to form a high-resolution pressure image; at the same time, the sensor array has the expandability of flexibly adjusting the number of rows and columns. In addition, the pressure-sensitive sensor array of the present invention adopts a modular design, has good scalability and maintainability, supports remote firmware upgrades, and facilitates function updates and optimizations.

[0029] In some embodiments, the pressure pad upper protective film 11 and the pressure pad lower protective film 12 are plastic films.

[0030] In some embodiments, the sensing unit is a pressure-sensitive conductive sheet Velostat, and a pressure-sensitive sensing array is constructed using Velostat material, which improves the comfort and acceptance of the seat cushion. Specifically, based on the physical and electrical properties of Velostat, a low-cost, highly expandable signal acquisition circuit, namely, a sensing array, is built based on a single-chip microcomputer.

[0031] In some embodiments, the data transmission between the first controller, the second controller and the data processing platform adopts WiFi and MQTT protocols; in addition, the data transmission between the first controller, the second controller and the data processing platform also includes but is not limited to Bluetooth, Zigbee, 2.4GHz, wired transmission, etc.

[0032] In some embodiments, ESP32 is used as the first controller and the second controller, and the first controller and the second controller may also include but are not limited to Arduino series, Raspberry Pi series, Nordic nRF52 series, and STM32 series.

[0033] In some embodiments, the data processing platform is but not limited to the Alibaba Cloud platform. The Alibaba Cloud platform serves as a data transfer and storage center, receives data from the piezoresistive array cushion, i.e., the signal sent by the first controller; and transfers it to the second controller of the smart desk lamp, thereby realizing real-time data transmission and interaction.

[0034] Specifically, the Alibaba Cloud platform is configured with an MQTT server to receive data sent from the first controller on the cushion side and store it in the data processing platform database for subsequent analysis and processing. Furthermore, the ESP32 controller on the desk lamp side is connected to the local area network via WiFi to ensure that the device is connected to the network; subsequently, the ESP32 controller on the desk lamp side and the ESP32 controller on the cushion side both establish a connection with the MQTT server, and the ESP32 controller on the desk lamp side receives the real-time sitting posture data transmitted from the cushion side via the MQTT protocol, calculates the pressure center position of the cushion through data processing, and determines whether the lighting mode needs to be switched according to the duration of the user's sitting posture, such as entering the breathing light mode or turning off the light to remind the user that the sitting posture needs to be changed.

[0035] In some embodiments, the smart desk lamp further includes a base 60, a display box 70 disposed on the base 60, and a lampshade box 80 connected to the display box 70; the brightness indicator light 40 and the center of gravity indicator light ring 50 are disposed on the display box 70, and the lighting lamp 30 is disposed in the lampshade box 80. The base 60 can play a role in fixing the desk lamp and accommodate the data processing platform and the second controller; the brightness indicator light 40 and the center of gravity indicator light ring 50 are disposed on the display box 70, so that the user can intuitively observe the changes of the brightness indicator light 40 and the center of gravity indicator light ring 50, which plays a reminder role; the lampshade box 70 can protect the lighting lamp 30, and the reflector in the lampshade box 80 can reflect the light of the lighting lamp well, thereby improving the utilization rate of the light source.

[0036] In some embodiments, the base 60 includes a base lower cover 61 and a base upper cover 62 matched with the base lower cover 61; the base upper cover 62 is provided with a lighting switch 63, and the lighting switch is connected to the second controller to control the brightness of the lighting. The base is assembled by matching the base lower cover and the base upper cover, which is convenient for maintenance of components such as the circuit, data processing platform and second controller arranged inside the base, and is easy to install. Preferably, the lighting switch is a touch switch, and the brightness of the lighting can be adjusted by touching.

[0037] In some embodiments, the display box 70 includes a display box rear cover 71 and a display box front cover 72 matching the display box rear cover 71; the brightness indicator light 40 and the center of gravity indicator light ring 50 are disposed on the display box front cover 72. The display box is formed by using the display box rear cover and the display box front cover, which facilitates the installation of the brightness indicator light and the center of gravity indicator light ring on the display box front cover and facilitates the subsequent maintenance of the brightness indicator light and the center of gravity indicator light ring.

[0038] In some embodiments, the brightness indicator lights include but are not limited to three, and the gear position of the lighting lamp is indicated by displaying different numbers of brightness indicator lights.

[0039] In some embodiments, the lampshade box 80 includes a lampshade upper cover 81 and a lampshade lower cover 82 matched with the lampshade upper cover 81; the lampshade lower cover 82 is provided with a lighting window, and a transparent lampshade 83 is provided at the lighting window; the lampshade box 80 is connected to the display box 70 through a hollow hose 90. The hollow hose acts as a universal hose, and the lampshade box is connected to the display box by using it, and the angle and height of the lampshade box can be adjusted.

[0040] In some embodiments, a tooth tube 100 is further provided in the lampshade box, which mainly plays a role of fixing and connecting during the assembly process.

[0041] In some embodiments, the smart desk lamp is further provided with a type-c charging module 200 for powering the smart desk lamp; preferably, the type-c charging module includes an ESP32 controller and a type-c charging interface.

[0042] In some embodiments, the first controller is provided with an analog input pin and a digital output pin; wherein the analog input pin is connected to the row electrode of the sensor array for reading the voltage; and the digital output pin is connected to the column electrode of the sensor array for controlling the voltage level. A voltage source (Vg) provides a driving voltage, and the voltage is applied to each column in turn through the column electrode, and a matrix scanning mechanism is used to reduce the pin requirement from 81 to 18.

[0043] For columns, each time a column is activated, the column electrode is set to a low level (0V), and the other column electrodes are set to a floating input state (high impedance state) to avoid mutual interference and prevent interference with the current column measurement; for row reading, the row electrode scans and reads the voltage on each row. At this time, the read value reflects the sensor resistance value at the current intersection. The voltage value read through the analog input pin ranges from 0V to 5V (corresponding to the 10-bit ADC of Arduino, the digital range is 0 to 1023).

[0044] During the process, when the pressure of the sensor array increases, the resistance value decreases, so the voltage drop decreases, and the read voltage value increases; when the pressure decreases, the resistance value increases, so the voltage drop increases, and the read voltage value decreases. Assume that the power supply voltage is Vg , the resistance of the sensor array is R p , reference resistance R = 20kΩ pull-up resistor built into each IO port of ATmega328P; V out is the actual measured voltage value, ; The analog value to be read Convert to actual voltage value, where the reference value of Arduino output voltage is =5V, ; Calculate the resistance of the sensor array from the actual voltage value R p , ; Finally, the resistance value can be converted into the corresponding pressure value through the characteristic curve of the sensor array.

[0045] In some embodiments, the lighting lamp, the brightness indicator light, and the center of gravity indicator light ring may adopt different lighting technologies and designs, including but not limited to LED light bars or light belts, OLED display screens, smart light bulbs, etc.

[0046] Specifically, the working principle of the sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp is as follows: Figure 4 As shown, including the following: 1. Within 1 hour after the user sits down, it is determined that the user has not engaged in sedentary behavior: 1) For the lighting, the user switches the brightness level of the lighting by touching the touch switch. There are 4 levels in total: 0, 1, 2, and 3. At level 0, the lighting is off; the brightness of the lighting gradually increases from 1 to 3. At level 3, touch the touch switch again to return to level 0. At levels 1 to 3, if a user is detected to have an unhealthy sitting posture, the brightness of the LED lighting will be reduced to 50% of the current brightness to prompt the user to adjust their sitting posture.

[0047] 2) For the brightness indicator light, by touching the touch switch, the brightness indicator light corresponding to the brightness level of the lighting lamp is switched. There are 4 levels, 0, 1, 2 and 3. At 0, all the brightness indicators are off; at 1, the bottom brightness indicator light is on; at 2, the bottom and middle brightness indicators are on; at 3, all three brightness indicators are on.

[0048] 3) If the user leaves the seat, the lamp automatically returns to the 0 gear. When the user sits down, the sitting time starts. If the user is away from the seat for more than 1 minute, the sitting time is reset. If the user is away from the seat for less than 1 minute, the sitting time continues. When the user sits down again, the lamp returns to the original setting gear.

[0049] 4) For the center of gravity indicator ring, when the center of gravity of the user's sitting posture is detected to be tilted, the desk lamp will map the calculated user's pressure center point to the corresponding position of the center of gravity indicator ring. When the pressure center deviates from the central area, the corresponding position of the center of gravity indicator ring will light up in red, and other positions will not light up, to prompt the user to adjust the sitting posture. When the pressure center is in the middle, all the lamp beads of the center of gravity indicator ring light up, indicating that the sitting posture is normal. The center of gravity indicator ring lights up when the user sits down and goes out when leaving, and is not controlled by the gear position.

[0050] 2. When the user has been seated for 1 hour, it is determined that the user has been sedentary for a long time, and the working mode of the above-mentioned lighting, brightness indicator light and center of gravity indicator light ring is interrupted, and the following working mode is executed: 1) The light breathes twice (the brightness gradually decreases to 50% of the current brightness and then gradually returns to the original brightness, and repeats twice without interruption) and then returns to the original working mode (the working mode when no sedentary behavior occurs).

[0051] 2) After the lighting light breathes twice, the brightness indicator light and the center of gravity indicator light ring continue to flash, reminding the user that he has been sitting for a long time. After the user stands up, the LED lighting light goes out, and the brightness indicator light and the center of gravity indicator light ring continue to flash until the cushion detects that the user has left the seat for 1 minute. After that, the sitting time is reset, and the brightness indicator light and the center of gravity indicator light ring return to the original working mode.

[0052] In this embodiment, through the dual-end ESP32 cloud communication, the cushion-side monitoring data provides an intuitive sitting posture and sedentary system through the light of the desk lamp, thereby improving the benefit of persuasive technology to users; in addition, the system adopts modular and wireless communication design, supports remote firmware upgrades and function expansion, and has the possibility of using cloud data for various records.

[0053] Specifically, the process of performing sitting posture detection and interaction using the sitting posture detection and interaction system based on the piezoresistive array cushion and the desk lamp is as follows: Figure 5 As shown, a piezoresistive array cushion made of Velostat material is used as a flexible pressure sensor array for data collection and preprocessing (i.e., normalization processing), and the processed data is used to train a sitting posture model to obtain a sitting posture recognition model; the flexible pressure sensor array is reused for real-time data collection and preprocessing, and the data is transmitted to the Alibaba Cloud platform for data recording and cloud streaming to the desk lamp end ESP32 to parse the data and identify the sitting posture, and finally the desk lamp is controlled through instructions to interact.

[0054] In addition, the present invention also provides a sitting posture detection and interaction method based on a piezoresistive array cushion and a desk lamp, comprising the steps of: Step S10: using the piezoresistive array cushion to collect raw pressure data, and performing normalization processing to obtain multiple groups of raw pressure data sets; Step S20: using the original pressure data set to train the neural network model to obtain a sitting posture recognition model; Step S30: using the piezoresistive array cushion to collect real-time pressure data and sending it to the data processing platform for storage. The data processing platform also sends the real-time pressure data to the sitting posture recognition model and outputs the sitting posture type. Step S40: The sitting posture type is fed back through the lighting effect of the smart desk lamp.

[0055] In this embodiment, the original pressure data is analyzed by a deep learning algorithm and the neural network model is trained to obtain a sitting posture recognition model; then the model is used to analyze the real-time pressure data to achieve accurate judgment and classification of various sitting postures and intelligent interaction with the desk lamp, thereby promoting the formation of healthy sitting habits and effectively improving the health of users.

[0056] In some embodiments, in step S10, the step of collecting raw pressure data using a piezoresistive array cushion and performing normalization processing to obtain multiple sets of raw pressure data sets is completed using a multi-scale splicing multi-layer perceptron method, such as Figure 6 As shown, including: Step S11: The sensor array composed of 9×9 sensor units collects raw pressure data and performs normalization processing to obtain 9×9 matrix data; Step S12: performing a pooling operation on the 9×9 matrix data, including 2×2 pooling and 3×3 pooling; Step S13: performing operations on the 2×2 pooling and the 3×3 pooling respectively to obtain a 4×4 feature map and a 3×3 feature map; Step S14: Flatten the 4×4 feature map and the 3×3 feature map, and concatenate them into a comprehensive feature vector, that is, a set of original pressure data sets.

[0057] In some implementations, the length of the comprehensive feature vector is 25, of which 4×4 flattened is 16 and 3×3 flattened is 9.

[0058] Compared with the existing methods, the multi-scale feature processing in this embodiment extracts the multi-scale features of the sensor array data in the piezoresistive array cushion through a multi-scale pooling operation, and splices the features of different scales, comprehensively utilizes the information at each scale, enhances the feature representation capability, and thus improves the model's ability to recognize different sitting postures.

[0059] In some embodiments, the step of training the neural network model using the original pressure data set to obtain the sitting posture recognition model uses TensorFlowLite to achieve the effect of real-time classification and monitoring of 7 sitting postures, specifically including: inputting the comprehensive feature vector into the fully connected neural network model for training to obtain the sitting posture recognition model; wherein the fully connected neural network model includes a first fully connected layer, a second fully connected layer and a third fully connected layer; the first fully connected layer has 128 neurons, and the activation function is ReLU; the second fully connected layer has 64 neurons, and the activation function is ReLU; the third fully connected layer has 32 neurons, and the activation function is ReLU; the output layer of the fully connected neural network model has 7 neurons, and the activation function is Softmax. Compared with the traditional LSTM network, this embodiment uses a multi-layer fully connected network with low computational overhead, which is suitable for real-time model operation and processing deployed on the microprocessor side. After inspection, its parameter amount is only 12% of the LSTM network method.

[0060] Specifically, during the training process, this embodiment uses TensorFlow because it can be converted into a TfLite file using TensorFlow Lite for Microcontrollers; after training, the file is converted into the TensorFlow Lite format, and then the weights are converted into a C byte array using the xxd command; then the TensorFlow Lite for Microcontrollers library is used to burn the weight header file into the second controller for calculation.

[0061] In actual detection, this embodiment switches the 7 sitting postures in sequence for 3 rounds every 5 seconds. At a recognition frequency of 1.5 Hz, the switching or accurate recognition within 2 frames is recorded as a successful recognition. The real-time detection accuracy of the 7 sitting postures is as follows: Figure 7 shown.

[0062] In some embodiments, the sitting posture recognition model is stored in the second controller; preferably, the sitting posture recognition model is stored in the ESP32 controller.

[0063] In some embodiments, in the steps S30-S40, after the piezoresistive array cushion collects real-time pressure data, it also includes: performing data noise reduction processing using the Kalman filter algorithm through the ESP32 controller to improve the accuracy and stability of the data; the processed data is packaged into JSON format data and sent to the data processing platform through the MQTT protocol, and the data processing platform then sends the JSON format data to the ESP32 controller at the desk lamp end, and the ESP32 controller extracts the COP_X and COP_Y coordinates by parsing the JSON format data for calculating the pressure center position; and determines whether the light mode needs to be switched according to the duration of the user's sitting posture, such as entering the breathing light mode or turning off the light. Then, according to the calculated pressure center point (COP_X and COP_Y), the system maps it to the lamp bead index to control the corresponding light state. When the user's pressure center deviates from the central area, the lamp bead in the corresponding direction will light up red to prompt the user to adjust the sitting posture. The mapping process determines the corresponding lamp bead index by calculating the offset and azimuth of the pressure center relative to the cushion center.

[0064] Specifically, in step S30, the real-time pressure data collected by the piezoresistive array cushion is standardized. The standardization method uses the mean and standard deviation of the data to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. The formula is as follows: ; in, X For real-time pressure data, μ is the mean of real-time pressure data, σis the standard deviation of the real-time pressure data.

[0065] Furthermore, in order to better capture the dynamic changes of data, a sliding window method is used to process the data; wherein the sliding window contains a fixed number of sampling points, and continuous data segments are generated by continuously moving the window; and the window interval can be adjusted according to actual needs to balance the time resolution and computational complexity.

[0066] In some embodiments, during the model selection process, a lightweight neural network model, such as a multilayer perceptron (MLP), is sampled to adapt to resource-constrained embedded devices. The input feature is to expand the standardized 9×9 pressure matrix into an 81-dimensional vector as the input feature of the model, and the output label of the model is the sitting posture type number, indicating the sitting posture category corresponding to the current data.

[0067] In some embodiments, in addition to the light effect feedback, step S40 also adopts or adds other soft and low-intrusive user feedback methods to remind the user to adjust the sitting posture or avoid sitting for a long time, including but not limited to: sound prompts, vibration prompts, and mobile device notifications.

[0068] In summary, the present invention provides a sitting posture detection and interaction system and method based on a piezoresistive array cushion and a desk lamp. The sitting posture detection and interaction system based on a piezoresistive array cushion and a desk lamp includes an intelligent desk lamp and a piezoresistive array cushion for obtaining sitting posture information; the piezoresistive array cushion includes a sensor array and a first controller electrically connected to the sensor array; the intelligent desk lamp includes: a data processing platform for receiving signals from the first controller and a second controller for receiving signals from the data processing platform, as well as a lighting lamp, a brightness indicator light and a center of gravity indicator light ring electrically connected to the second controller respectively. The present invention utilizes a piezoresistive array cushion and an intelligent desk lamp to complete sitting posture detection and information interaction, realizing an intelligent seat cushion system, which can monitor various sitting postures of users with high precision through the piezoresistive array cushion, and can remind users to adjust their sitting postures in real time through a visual feedback mechanism on the desk lamp end; and the system also has a sedentary reminder function to effectively manage the user's sitting posture health.

[0069] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A sitting posture detection and interaction system based on a piezoresistive array cushion and a desk lamp, characterized in that: Includes a smart desk lamp and a piezoresistive array cushion for obtaining sitting posture information; The piezoresistive array cushion includes a sensor array and a first controller electrically connected to the sensor array; The intelligent desk lamp comprises: a data processing platform for receiving signals from the first controller and a second controller for receiving signals from the data processing platform, as well as a lighting lamp, a brightness indicator light and a center of gravity indicator light ring electrically connected to the second controller respectively.

2. The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp according to claim 1 is characterized in that: The piezoresistive array cushion also includes a pressure pad upper protective film and a pressure pad lower protective film sandwiching the sensor array in the middle; the sensor array includes a plurality of sensor units arranged in a matrix, and the plurality of sensor units are connected in series through conductive wires.

3. The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp according to claim 2 is characterized in that: The sensing unit is a pressure-sensitive conductive sheet Velostat.

4. The sitting posture detection and interaction system based on piezoresistive array cushion and desk lamp according to claim 1, characterized in that: The smart desk lamp also includes a base, a display box arranged on the base, and a lampshade box connected to the display box; the brightness indicator light and the center of gravity indicator light ring are arranged on the display box, and the lighting lamp is arranged in the lampshade box.

5. The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp according to claim 4 is characterized in that: The base comprises a base lower cover and a base upper cover matching the base lower cover; a lighting switch is arranged on the base upper cover, and the lighting switch is connected to the second controller for controlling the brightness of the lighting.

6. The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp according to claim 4 is characterized in that: The display box comprises a display box rear cover and a display box front cover matching the display box rear cover; the brightness indicator light and the center of gravity indicator light ring are arranged on the display box front cover.

7. The sitting posture detection and interaction system based on the piezoresistive array cushion and desk lamp according to claim 4 is characterized in that: The lampshade box comprises a lampshade upper cover and a lampshade lower cover matched with the lampshade upper cover; the lampshade lower cover is provided with a lighting window, and a transparent lampshade is provided at the lighting window; the lampshade box is connected to the display box through a hollow hose.

8. A sitting posture detection and interaction method based on a piezoresistive array cushion and a desk lamp, characterized in that: Includes steps: The original pressure data is collected by using a piezoresistive array cushion and normalized to obtain multiple sets of original pressure data sets; Using the original pressure data set to train a neural network model to obtain a sitting posture recognition model; Using the piezoresistive array cushion to collect real-time pressure data, the data processing platform is sent to the data processing platform for storage, and the data processing platform simultaneously sends the real-time pressure data to the sitting posture recognition model and outputs the sitting posture type; The sitting posture type is fed back through the lighting effect of the smart desk lamp.

9. The sitting posture detection and interaction method based on the piezoresistive array cushion and desk lamp according to claim 8, characterized in that: The step of collecting raw pressure data using the piezoresistive array cushion and performing normalization processing to obtain multiple groups of raw pressure data sets includes: The sensor array composed of 9×9 sensor units collects the original pressure data and performs normalization processing to obtain 9×9 matrix data; Performing a pooling operation on the 9×9 matrix data, including 2×2 pooling and 3×3 pooling; The 2×2 pooling and 3×3 pooling are respectively operated to obtain a 4×4 feature map and a 3×3 feature map; The 4×4 feature map and the 3×3 feature map are flattened and concatenated into a comprehensive feature vector, that is, a set of original pressure data sets.

10. The sitting posture detection and interaction method based on the piezoresistive array cushion and desk lamp according to claim 9, characterized in that: The steps of training a neural network model using the original pressure data set to obtain a sitting posture recognition model include: Inputting the comprehensive feature vector into a fully connected neural network model for training to obtain a sitting posture recognition model; Among them, the fully connected neural network model includes a first fully connected layer, a second fully connected layer and a third fully connected layer; the first fully connected layer has 128 neurons, and the activation function is ReLU; the second fully connected layer has 64 neurons, and the activation function is ReLU; the third fully connected layer has 32 neurons, and the activation function is ReLU; the output layer of the fully connected neural network model has 7 neurons, and the activation function is Softmax.

Images