Artificial intelligence-based adaptive adjustment table lamp for sitting posture and light

Abstract

This invention discloses an artificial intelligence-based posture and light adaptive desk lamp, belonging to the cross-disciplinary field of intelligent health lighting. It includes a visual perception module, a physiological perception module, an ambient light perception module, a state assessment module, a light control execution module, and a generative interaction module. The visual perception module collects image data of the user's upper body and sends the image data to the state assessment module. The physiological perception module acquires the user's heart rate and respiratory rate signals via millimeter-wave radar and sends these signals to the state assessment module. The ambient light perception module detects ambient illuminance and color temperature values ​​and sends these values ​​to the state assessment module. The state assessment module calculates the spatial position of the user's key skeletal points based on the image data, generating a posture state vector including cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance.

Description

Technical Field

[0001] This invention relates to the field of intelligent health lighting technology, and in particular to a desk lamp that adapts to sitting posture and light based on artificial intelligence. Background Technology

[0002] Intelligent health lighting cross-technology refers to a technology system that integrates multiple disciplines such as artificial intelligence, biosensing, human factors engineering, and lighting science. It dynamically adjusts light source parameters to promote visual comfort, circadian rhythm health, and physical and mental well-being by sensing users' physiological state, behavior, posture, and ambient light conditions in real time. This technology not only focuses on seeing clearly but also emphasizes proper lighting. That is, it actively optimizes the light environment according to individual differences and usage scenarios to prevent myopia, relieve eye fatigue, improve concentration, or aid sleep. It is widely used in education, office, home, and medical scenarios and is a typical example of the deep integration of smart hardware and proactive health concepts.

[0003] Traditional smart desk lamps mostly use a single infrared or ultrasonic ranging method to determine head-down behavior. They can only measure the vertical distance from the head to the table and cannot distinguish between normal reading and poor posture such as forward head posture. They are also easily affected by the user's height and chair height, and lack quantitative assessment of the overall posture of the torso. Furthermore, the existing automatic dimming function usually only adjusts the brightness based on the ambient illuminance value without considering the user's current physiological state and eye load. This results in the output of high color temperature and strong light when the user is tired or stressed, which will exacerbate visual and psychological stress. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an AI-based posture and light adaptive desk lamp that solves the problem that traditional smart desk lamps mostly use a single infrared or ultrasonic ranging method to judge head-down behavior. They can only measure the vertical distance from the head to the table and cannot distinguish between normal reading and poor posture such as forward head posture. In addition, they are easily affected by the user's height and seat height, and lack quantitative assessment of the overall posture of the torso. Furthermore, existing automatic dimming functions usually only adjust the brightness based on the ambient illuminance value without considering the user's current physiological state and eye load. This results in the output of high color temperature and strong light when the user is tired or stressed, which exacerbates the problem of visual and psychological stress.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a posture and light adaptive desk lamp based on artificial intelligence, comprising: The system includes a visual perception module, a physiological perception module, an ambient light perception module, a state assessment module, a light control execution module, and a generative interaction module. The visual perception module is used to collect image data of the user's upper body and send the image data to the status assessment module; The physiological sensing module is used to acquire the user's heart rate and respiratory rate signals through millimeter-wave radar, and send the heart rate and respiratory rate signals to the status assessment module. The ambient light sensing module is used to detect the ambient illuminance value and color temperature value, and send the illuminance value and color temperature value to the status assessment module; The status assessment module is used to calculate the spatial position of the user's key skeletal points based on image data, generate a sitting posture state vector including cervical spine forward tilt angle, trunk tilt angle and eye-to-book distance, fuse the sitting posture state vector, heart rate signal, respiratory rate signal, illuminance value and color temperature value, and combine it with the locally stored personal health digital twin model to output dynamic light guidance instructions, light environment collaborative optimization instructions and personalized health suggestion text. The light control execution module is used to project a guide light spot within the user's field of vision according to the dynamic light guidance command, and to adjust the brightness and color temperature of the light source according to the light environment collaborative optimization command. The generative interaction module is used to convert personalized health advice text into voice or graphic prompts, and to write the user's feedback on the voice or graphic prompts into the personal health digital twin model.

[0007] As a preferred embodiment of the AI-based posture and light adaptive desk lamp of the present invention, wherein: the state evaluation module calculates the spatial position of the user's key skeletal points based on image data, and the specific steps are as follows: Perform human two-dimensional pose estimation processing on the upper body image output by the visual perception module to obtain the pixel positions of the head, left shoulder, right shoulder, midpoint of the thoracic vertebrae, upper point of the lumbar vertebrae, and center of the eyes in the image coordinate system; By combining the corresponding depth map obtained by the depth sensor, the pixel position is back-projected onto the table lamp body coordinate system using the pre-calibrated camera intrinsic parameter matrix to obtain the three-dimensional spatial coordinates of six key skeletal points. Using the midpoint of the thoracic vertebrae as the reference origin, a local trunk coordinate system is constructed, and the head and neck direction vector and the trunk principal axis direction vector are defined. The cervical lordosis angle is calculated as the angle between the head and neck direction vector and the vertically upward unit vector. The trunk tilt angle is calculated as the angle between the trunk principal axis direction vector and the vertically upward unit vector. The eye-to-book distance is calculated as the vertical distance from the center point of both eyes to the desktop working plane, which is determined by the installation height of the desk lamp base and the horizontal calibration parameters. The cervical spine forward tilt angle, trunk tilt angle and eye-to-book distance are combined to form a sitting posture state vector; Among them, cervical anteversion angle The calculation uses the following formula: ; In the formula, This represents the spatial vector pointing from the midpoint of the thoracic vertebrae to the head, which is obtained by subtracting the three-dimensional coordinates of the head from the three-dimensional coordinates of the midpoint of the thoracic vertebrae. This represents the unit vector along the direction of gravity in the coordinate system of the lamp body, with the direction vertically upward; Represents vector The Euclidean norm.

[0008] As a preferred embodiment of the AI-based posture and light adaptive desk lamp of the present invention, the state evaluation module integrates posture state vector, heart rate signal, respiratory rate signal, illuminance value and color temperature value, and the specific steps are as follows: The cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance in the sitting posture state vector, together with the heart rate signal and respiratory rate signal output by the physiological perception module, and the illuminance value and color temperature value output by the ambient light perception module, constitute a multimodal input vector. The multimodal input vectors are fed into a locally deployed hierarchical artificial intelligence processing engine, which includes a lower-level multimodal fusion network and an upper-level behavior reasoning module. The underlying multimodal fusion network adopts a lightweight convolution-attention hybrid structure to extract features and perform cross-modal alignment on data from different modalities, and outputs a structured comprehensive state vector; The upper-level behavioral reasoning module compares the structured integrated state vector with the historical state sequence stored in the personal health digital twin model in a time sequence to identify the degree of deviation of the current state from the health baseline. Based on the degree of deviation, user age tags, and historical correction response records, dynamic light guidance instructions, light environment collaborative optimization instructions, and personalized health suggestion texts are generated by matching from a preset behavior rule library.

[0009] As a preferred embodiment of the AI-based posture and light adaptive adjustment desk lamp of the present invention, wherein: the light control execution module projects a guide light spot within the user's field of vision according to a dynamic light guidance command, the specific steps of which are: Analyze the target guidance position and guidance trajectory type contained in the dynamic optical guidance command; When the forward tilt angle of the cervical spine exceeds the healthy threshold, the target guidance position is set in the space directly in front of the user and slightly above the current line of sight. Control the miniature MEMS mirror in the programmable optical component to reflect the LED light source to the direction corresponding to the target guidance position; The light spot moves along a vertical or horizontal path at a preset rate to form a continuous visual guidance path, guiding the user to naturally look up or tilt back. The brightness of the guiding light spot is dynamically adjusted according to the distance between the eyes and the book. When the distance between the eyes and the book is less than the safe threshold, the brightness of the light spot is increased to enhance the prompt intensity. The light spot's movement trajectory avoids the area where the user is reading, so as not to interfere with normal study or work.

[0010] As a preferred embodiment of the AI-based posture and light adaptive adjustment desk lamp of the present invention, the light control execution module adjusts the brightness and color temperature of the light source according to the light environment collaborative optimization command, and the specific steps are as follows: Extract target brightness and target color temperature from the light environment collaborative optimization command; Read the current ambient light level and the user's heart rate signal; If the heart rate is higher than the resting baseline and the ambient light is too low, set the target brightness to medium-high and the target color temperature to the low color temperature range to create a relaxing atmosphere. If the heart rate is below the resting baseline and the ambient light is too high, set the target brightness to low brightness and the target color temperature to the high color temperature range to maintain wakefulness. Drive the tunable spectrum LED array so that the output luminous flux and color coordinates approximate the target brightness and target color temperature; The adjustment process is smooth, avoiding visual discomfort caused by sudden changes in brightness or color temperature.

[0011] As a preferred embodiment of the AI-based posture and light adaptive desk lamp described in this invention, the generative interaction module converts personalized health advice text into voice or graphic prompts, specifically through the following steps: Receive personalized health advice text output by the status assessment module, including the type of sitting posture deviation and a description of specific corrective actions; Call the local miniaturized generative language model to map the text into an expression that matches the user's age and cognitive level; If the user is under 12 years old, a cartoon icon will be used in conjunction with a short voice broadcast. If the user is 12 years or older, then text scrolling prompts and natural speech synthesis will be used. The voice or text prompts clearly point to specific body parts and directions of movement.

[0012] As a preferred embodiment of the AI-based posture and light adaptive desk lamp described in this invention, the generative interaction module writes the user's feedback on voice or graphic prompts into a personal health digital twin model, specifically through the following steps: Within three seconds after the voice or text prompt is issued, continuously monitor the changes in the sitting posture vector, including changes in the cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance. If the absolute value of the change in the cervical spine forward tilt angle is greater than the preset response threshold, then the prompt is considered to have generated a valid response. Record the prompt type, timestamp, current environmental parameters, and user physiological state corresponding to valid responses; The recorded data is added as new training samples to the training set of the personal health digital twin model; An incremental learning algorithm is used to update the posture correction response function, which is used to predict the corrective effect of different guidance strategies on a specific user. Through continuous closed-loop feedback, the system gradually adapts to individual user behavior habits, thereby improving the effectiveness of long-term intervention.

[0013] As a preferred embodiment of the AI-based posture and light adaptive adjustment desk lamp of the present invention, wherein: the personal health digital twin model is stored in the desk lamp's local encrypted storage unit, and the specific steps are as follows: All user behavior data, physiological data, and model parameters are processed by AES-256 encryption algorithm before being written to the storage unit. The data upload process is only allowed to be initiated when the user presses and holds a physical button for three seconds or speaks a preset voice command; Before uploading, the data is anonymized by removing identity information and retaining only the group behavior feature vector. The cloud server aggregates the de-identified feature vectors uploaded by multiple desk lamps and trains the shared weights of the global multimodal fusion network. After training is complete, an incremental update package for the model is generated and pushed to each lamp via a TLS 1.3 secure channel; When the desk lamp is in standby mode at night and connected to a power source, it automatically loads an incremental update package to update the parameters of the underlying multimodal fusion network.

[0014] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the posture and light adaptive adjustment desk lamp based on artificial intelligence as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the posture and light adaptive adjustment desk lamp based on artificial intelligence as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By combining millimeter-wave radar for non-intrusive physiological monitoring, a posture quantification model based on skeletal point geometry, a dynamic light guidance mechanism, and a localized health digital twin model, end-to-end adaptive adjustment is achieved. By using a geometric method to calculate the cervical spine forward tilt angle based on the midpoint of the thoracic spine, errors caused by differences in individual height and sitting posture are effectively eliminated. Furthermore, by utilizing a light environment collaborative optimization strategy that links the spatial projection of the guiding light spot with physiological state, non-invasive health intervention is achieved without interfering with the user's concentration. At the same time, relying on local encrypted storage and a collective intelligent incremental update mechanism, personalized correction accuracy is continuously improved while ensuring privacy and security, effectively enhancing the proactive protection capabilities of adolescents and office workers against long-term eye strain and spinal health. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the posture and light adaptive adjustment desk lamp based on artificial intelligence in Example 1. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, referring to Figure 1 As one embodiment of the present invention, this embodiment provides an artificial intelligence-based posture and light adaptive desk lamp, comprising: The system includes a visual perception module, a physiological perception module, an ambient light perception module, a state assessment module, a light control execution module, and a generative interaction module.

[0023] S1, the visual perception module is used to collect image data of the user's upper body and send the image data to the status assessment module.

[0024] Furthermore, the visual perception module includes a wide-angle RGB camera and an infrared depth sensor, both mounted on the same optical axis below the lamp head. Upon startup, it synchronously acquires color images and depth maps at a frequency of 30 frames per second. It performs face detection on the color images, and if a face is detected, it crops out the upper body region, including the shoulders and upper back, based on the center of the face. The cropped upper body color image is then registered and fused with the depth map of the corresponding region to form upper body image data with depth information. This image data is compressed using H.264 and transmitted to the status evaluation module via an internal high-speed bus.

[0025] It should be noted that by simultaneously acquiring color images and depth information, and intelligently cropping the upper body area centered on the face, the system effectively focuses on the key parts required for pose analysis, avoiding interference from irrelevant backgrounds. At the same time, it adopts a co-optical axis design and data registration fusion strategy to ensure spatial consistency and improve the accuracy of subsequent skeletal point positioning. After compression, the data is transmitted at high speed, balancing real-time performance and system resource consumption, providing a high-quality, low-latency visual input foundation for state assessment.

[0026] S2, the physiological sensing module is used to acquire the user's heart rate and respiratory rate signals through millimeter-wave radar, and send the heart rate and respiratory rate signals to the status assessment module.

[0027] Furthermore, the physiological sensing module employs a 60GHz continuous wave millimeter-wave radar chip, integrated into the front edge of the lamp base. The radar emits low-power electromagnetic waves that penetrate clothing to illuminate the user's chest area, receiving echo signals reflected by slight chest movements. The echo signals undergo Fast Fourier Transform to extract the Doppler frequency shift components. Among these, the low-frequency component (0.1-0.5Hz) corresponds to respiratory movements, and the high-frequency component (0.8-2.5Hz) corresponds to heartbeats. Bandpass filtering, peak detection, and time-domain smoothing are performed on the two frequency band signals respectively, outputting stable heart rate and respiratory rate signals. The signals are packaged and sent to the status assessment module at a sampling rate of once per second.

[0028] It should be noted that using millimeter-wave radar to non-contactly penetrate clothing to sense subtle chest movements allows for continuous acquisition of physiological signals without the user's awareness, avoiding discomfort or decreased compliance caused by wearing sensors. By decomposing the echo signal in the frequency domain and performing targeted filtering, respiratory and heartbeat components are effectively separated, improving signal stability and anti-interference capabilities. This method achieves highly reliable physiological state monitoring while protecting privacy, providing key dimension support for multimodal health assessment.

[0029] S3, the ambient light sensing module is used to detect the ambient illuminance and color temperature values, and send the illuminance and color temperature values ​​to the status assessment module.

[0030] Furthermore, the ambient light sensing module includes a high dynamic range digital ambient light sensor, installed on the top outer side of the lampshade to avoid direct illumination from its own light source. The sensor incorporates a dual photodiode array, covering the short-wavelength and long-wavelength response ranges of the visible spectrum. It calculates the current ambient color temperature by measuring the photocurrent ratio of the two channels, and then converts this into an illuminance value by combining the total luminous flux integral. The sensor collects a set of illuminance-color temperature data pairs every 500 milliseconds, which, after temperature compensation and nonlinear correction, are then processed via I... 2 The C interface transmits the values ​​to the status assessment module.

[0031] It should be noted that the ambient light sensor is placed on the top of the outside of the lampshade, avoiding direct light from its own light source, to ensure that the measured illuminance and color temperature truly reflect the external environment rather than the output of the desk lamp; the use of a dual-channel photoelectric response structure combined with the ratio method to calculate the color temperature effectively eliminates measurement deviations caused by light source aging or temperature drift; high-frequency sampling combined with a correction algorithm enables the environmental perception to have high dynamic adaptability, providing an accurate basis for the collaborative optimization of the light environment.

[0032] S4, the status assessment module is used to calculate the spatial position of the user's key skeletal points based on image data, generate a sitting posture status vector including cervical spine forward tilt angle, trunk tilt angle and eye-to-book distance, fuse the sitting posture status vector, heart rate signal, respiratory rate signal, illuminance value and color temperature value, and combine it with the locally stored personal health digital twin model to output dynamic light guidance instructions, light environment collaborative optimization instructions and personalized health suggestion text.

[0033] Furthermore, the upper body image output by the visual perception module is processed for human two-dimensional pose estimation to obtain the pixel positions of the head, left shoulder, right shoulder, midpoint of the thoracic vertebrae, upper point of the lumbar vertebrae, and center of the eyes in the image coordinate system. Combined with the corresponding depth map obtained by the depth sensor, the pixel positions are back-projected to the table lamp body coordinate system using a pre-calibrated camera intrinsic parameter matrix to obtain the three-dimensional spatial coordinates of six key skeletal points. With the midpoint of the thoracic vertebrae as the reference origin, a local trunk coordinate system is constructed, and the head and neck direction vector and the trunk principal axis direction vector are defined.

[0034] The cervical spine forward tilt angle is calculated as the angle between the head and neck direction vector and the vertically upward unit vector; the trunk tilt angle is calculated as the angle between the trunk principal axis direction vector and the vertically upward unit vector; the eye-to-book distance is calculated as the vertical distance from the center point of both eyes to the desktop working plane, which is determined by the lamp base installation height and horizontal calibration parameters; the cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance are combined to form a sitting posture vector; where, the cervical spine forward tilt angle... The calculation uses the following formula: ; In the formula, This represents the spatial vector pointing from the midpoint of the thoracic vertebrae to the head, which is obtained by subtracting the three-dimensional coordinates of the head from the three-dimensional coordinates of the midpoint of the thoracic vertebrae. This represents the unit vector along the direction of gravity in the coordinate system of the lamp body, with the direction vertically upward; Represents vector The Euclidean norm.

[0035] The cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance from the sitting posture state vector, together with the heart rate and respiratory rate signals output by the physiological perception module, and the illuminance and color temperature values ​​output by the ambient light perception module, constitute a multimodal input vector. This multimodal input vector is then fed into a locally deployed hierarchical artificial intelligence processing engine, which includes a bottom-level multimodal fusion network and an upper-level behavior reasoning module. The bottom-level multimodal fusion network employs a lightweight convolution-attention hybrid structure to extract features from different modalities and perform cross-modal alignment, outputting a structured comprehensive state vector.

[0036] The upper-level behavioral reasoning module compares the structured comprehensive state vector with the historical state sequence stored in the personal health digital twin model in a time sequence to identify the degree of deviation of the current state from the health baseline. Based on the degree of deviation, user age tag and historical correction response records, it matches and generates dynamic light guidance instructions, light environment collaborative optimization instructions and personalized health suggestion text from the preset behavioral rule library.

[0037] It should be noted that by using the midpoint of the thoracic spine as the origin of the local coordinate system to construct geometric relationships, the shortcomings of traditional methods that use the head or desktop as a reference are easily affected by individual height and chair height, making the quantification of sitting posture more universal and comparable. By integrating visual, physiological and environmental multi-source information and combining it with a local health digital twin model for time-series comparison, it can not only identify the current abnormal state, but also determine the degree of deviation from the health baseline, thereby generating personalized intervention strategies with context awareness capabilities and improving the depth and accuracy of the assessment.

[0038] S5, the light control execution module, is used to project a guide light spot within the user's field of vision according to the dynamic light guidance command, and to adjust the brightness and color temperature of the light source according to the light environment collaborative optimization command.

[0039] Furthermore, the system analyzes the target guidance position and guidance trajectory type contained in the dynamic light guidance command; when the cervical spine forward tilt angle exceeds the health threshold, the target guidance position is set to be located in the space directly in front of the user and slightly above the current line of sight; the system controls the micro MEMS galvanometer in the programmable optical component to reflect the LED light source to the direction corresponding to the target guidance position.

[0040] The light spot moves along a vertical or horizontal path at a preset rate to form a continuous visual guidance path, guiding the user to naturally look up or tilt back. The brightness of the guiding light spot is dynamically adjusted according to the distance between the eyes and the book. When the distance between the eyes and the book is less than a safe threshold, the brightness of the light spot is increased to enhance the prompt intensity. The movement trajectory of the light spot avoids the area where the user is reading, so as not to interfere with normal study or work.

[0041] The system extracts target brightness and target color temperature from the light environment collaborative optimization instructions; reads the current ambient illuminance value and the user's heart rate signal; if the heart rate is higher than the resting baseline and the ambient illuminance is too low, it sets the target brightness to medium-high brightness and the target color temperature to the low color temperature range to create a relaxing atmosphere; if the heart rate is lower than the resting baseline and the ambient illuminance is too high, it sets the target brightness to low brightness and the target color temperature to the high color temperature range to maintain alertness; it drives an adjustable spectrum LED array to make the output luminous flux and color coordinates approximate the target brightness and target color temperature; the adjustment process is smooth to avoid visual discomfort caused by sudden changes in brightness or color temperature.

[0042] It should be noted that the spatial projection position and trajectory of the guiding light spot are dynamically generated based on the real-time sitting posture and actively avoid the reading content area to achieve non-disturbing visual guidance; the light environment adjustment not only responds to ambient light but also links with the user's physiological state, forming an active health adjustment mechanism where light changes with the person; the smooth transition design of brightness and color temperature conforms to the visual adaptation characteristics of the human eye, avoiding sudden stimulation and ensuring user comfort while improving the effectiveness of intervention.

[0043] S6, the generative interaction module, is used to convert personalized health advice text into voice or graphic prompts, and write the user's feedback on the voice or graphic prompts into the personal health digital twin model.

[0044] Furthermore, the system receives personalized health advice text from the status assessment module, including the type of posture deviation and a description of specific corrective actions. It then calls a local miniaturized generative language model to map the text into an expression that matches the user's age and cognitive level. If the user is under 12 years old, a cartoon icon is used in conjunction with a short voice broadcast. If the user is 12 years old or older, scrolling text prompts and natural speech synthesis are used. The voice or graphic prompts clearly point to specific body parts and directions of movement.

[0045] Within three seconds of issuing a voice or text prompt, continuously monitor changes in the sitting posture vector, including changes in cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance. If the absolute value of the change in cervical spine forward tilt angle is greater than a preset response threshold, the prompt is considered to have generated a valid response. Record the prompt type, timestamp, current environmental parameters, and user physiological state corresponding to the valid response.

[0046] Recorded data is added as new training samples to the training set of the personal health digital twin model; the sitting posture correction response function is updated using an incremental learning algorithm, which is used to predict the corrective effect of different guidance strategies on specific users; through continuous closed-loop feedback, the system gradually adapts to the individual user's behavioral habits, improving the effectiveness of long-term intervention.

[0047] The personal health digital twin model is stored in the local encrypted storage unit of the desk lamp. All user behavior data, physiological data, and model parameters are processed by the AES-256 encryption algorithm before being written to the storage unit. The data upload process is only allowed when the user presses and holds a physical button for three seconds or speaks a preset voice command. Before uploading, the data is anonymized, removing identity information and retaining only the group behavior feature vector. The cloud server aggregates the anonymized feature vectors uploaded by multiple desk lamps and trains the shared weights of the global multimodal fusion network. After training, an incremental update package is generated and pushed to each desk lamp through a TLS 1.3 secure channel. When the desk lamp is in standby mode at night and connected to a power source, it automatically loads the incremental update package to update the underlying multimodal fusion network parameters.

[0048] It should be noted that the interactive content automatically adapts its expression to the user's age, balancing child-friendliness with adult professionalism, thereby improving the acceptance of prompts and the willingness to follow up; through a closed-loop feedback mechanism, the system automatically records user response behavior and updates the individual health model, enabling the system to have continuous learning and personalized evolution capabilities; all sensitive data is stored locally in encrypted form, and uploading requires explicit user authorization and is anonymized, thus strictly protecting user privacy and security while achieving collective intelligent collaborative optimization.

[0049] This embodiment also provides a computer device suitable for an AI-based posture and light adaptive adjustment desk lamp, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the AI-based posture and light adaptive adjustment desk lamp as proposed in the above embodiment.

[0050] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0051] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements an AI-based posture and light adaptive adjustment desk lamp as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0052] In summary, this invention combines millimeter-wave radar for non-invasive physiological monitoring, a posture quantification model based on skeletal point geometry, a dynamic light guidance mechanism, and a localized health digital twin model to achieve end-to-end adaptive adjustment. By using a geometric method to calculate the cervical spine forward tilt angle based on the midpoint of the thoracic spine, it effectively eliminates errors caused by differences in individual height and sitting posture. Furthermore, by utilizing a light environment optimization strategy that links the spatial projection of the guiding light spot with physiological state, it achieves non-invasive health intervention without interfering with the user's concentration. At the same time, relying on local encrypted storage and a collective intelligent incremental update mechanism, it continuously improves the accuracy of personalized correction while ensuring privacy and security, effectively enhancing the proactive protection capabilities of adolescents and office workers against long-term eye strain and spinal health.

[0053] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A desk lamp with adaptive posture and light adjustment based on artificial intelligence, characterized in that: include: The system includes a visual perception module, a physiological perception module, an ambient light perception module, a state assessment module, a light control execution module, and a generative interaction module. The visual perception module is used to collect image data of the user's upper body and send the image data to the status assessment module; The physiological sensing module is used to acquire the user's heart rate and respiratory rate signals through millimeter-wave radar, and send the heart rate and respiratory rate signals to the status assessment module. The ambient light sensing module is used to detect the ambient illuminance value and color temperature value, and send the illuminance value and color temperature value to the status assessment module; The status assessment module is used to calculate the spatial position of the user's key skeletal points based on image data, generate a sitting posture state vector including cervical spine forward tilt angle, trunk tilt angle and eye-to-book distance, fuse the sitting posture state vector, heart rate signal, respiratory rate signal, illuminance value and color temperature value, and combine it with the locally stored personal health digital twin model to output dynamic light guidance instructions, light environment collaborative optimization instructions and personalized health suggestion text. The light control execution module is used to project a guide light spot within the user's field of vision according to the dynamic light guidance command, and to adjust the brightness and color temperature of the light source according to the light environment collaborative optimization command. The generative interaction module is used to convert personalized health advice text into voice or graphic prompts, and to write the user's feedback on the voice or graphic prompts into the personal health digital twin model.

2. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 1, characterized in that: The status assessment module calculates the spatial location of the user's key skeletal points based on the image data, and the specific steps are as follows: Perform human two-dimensional pose estimation processing on the upper body image output by the visual perception module to obtain the pixel positions of the head, left shoulder, right shoulder, midpoint of the thoracic vertebrae, upper point of the lumbar vertebrae, and center of the eyes in the image coordinate system; By combining the corresponding depth map obtained by the depth sensor, the pixel position is back-projected onto the table lamp body coordinate system using the pre-calibrated camera intrinsic parameter matrix to obtain the three-dimensional spatial coordinates of six key skeletal points. Using the midpoint of the thoracic vertebrae as the reference origin, a local trunk coordinate system is constructed, and the head and neck direction vector and the trunk principal axis direction vector are defined. The cervical lordosis angle is calculated as the angle between the head and neck direction vector and the vertically upward unit vector. The trunk tilt angle is calculated as the angle between the trunk principal axis direction vector and the vertically upward unit vector. The eye-to-book distance is calculated as the vertical distance from the center point of both eyes to the desktop working plane, which is determined by the installation height of the desk lamp base and the horizontal calibration parameters. The cervical spine forward tilt angle, trunk tilt angle and eye-to-book distance are combined to form a sitting posture state vector; Among them, cervical anteversion angle The calculation uses the following formula: ; In the formula, This represents the spatial vector pointing from the midpoint of the thoracic vertebrae to the head, which is obtained by subtracting the three-dimensional coordinates of the head from the three-dimensional coordinates of the midpoint of the thoracic vertebrae. This represents the unit vector along the direction of gravity in the coordinate system of the lamp body, with the direction vertically upward; Represents vector The Euclidean norm.

3. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 2, characterized in that: The state assessment module integrates the sitting posture vector, heart rate signal, respiratory rate signal, illuminance value, and color temperature value. The specific steps are as follows: The cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance in the sitting posture state vector, together with the heart rate signal and respiratory rate signal output by the physiological perception module, and the illuminance value and color temperature value output by the ambient light perception module, constitute a multimodal input vector. The multimodal input vectors are fed into a locally deployed hierarchical artificial intelligence processing engine, which includes a lower-level multimodal fusion network and an upper-level behavior reasoning module. The underlying multimodal fusion network adopts a lightweight convolution-attention hybrid structure to extract features and perform cross-modal alignment on data from different modalities, and outputs a structured comprehensive state vector; The upper-level behavioral reasoning module compares the structured integrated state vector with the historical state sequence stored in the personal health digital twin model in a time sequence to identify the degree of deviation of the current state from the health baseline. Based on the degree of deviation, user age tags, and historical correction response records, dynamic light guidance instructions, light environment collaborative optimization instructions, and personalized health suggestion texts are generated by matching from a preset behavior rule library.

4. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 3, characterized in that: The light modulation execution module projects a guide light spot within the user's field of vision according to the dynamic light guidance command. The specific steps are as follows: Analyze the target guidance position and guidance trajectory type contained in the dynamic optical guidance command; When the forward tilt angle of the cervical spine exceeds the healthy threshold, the target guidance position is set in the space directly in front of the user and slightly above the current line of sight. Control the miniature MEMS mirror in the programmable optical component to reflect the LED light source to the direction corresponding to the target guidance position; The light spot moves along a vertical or horizontal path at a preset rate to form a continuous visual guidance path, guiding the user to naturally look up or tilt back. The brightness of the guiding light spot is dynamically adjusted according to the distance between the eyes and the book. When the distance between the eyes and the book is less than the safe threshold, the brightness of the light spot is increased to enhance the prompt intensity. The light spot's movement trajectory avoids the area where the user is reading, so as not to interfere with normal study or work.

5. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 4, characterized in that: The light control execution module adjusts the brightness and color temperature of the light source according to the light environment collaborative optimization command. The specific steps are as follows: Extract target brightness and target color temperature from the light environment collaborative optimization command; Read the current ambient light level and the user's heart rate signal; If the heart rate is higher than the resting baseline and the ambient light is too low, set the target brightness to medium-high and the target color temperature to the low color temperature range to create a relaxing atmosphere. If the heart rate is below the resting baseline and the ambient light is too high, set the target brightness to low brightness and the target color temperature to the high color temperature range to maintain wakefulness. Drive the tunable spectrum LED array so that the output luminous flux and color coordinates approximate the target brightness and target color temperature; The adjustment process is smooth, avoiding visual discomfort caused by sudden changes in brightness or color temperature.

6. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 5, characterized in that: The generative interaction module converts personalized health advice text into voice or graphic prompts. The specific steps are as follows: Receive personalized health advice text output by the status assessment module, including the type of sitting posture deviation and a description of specific corrective actions; Call the local miniaturized generative language model to map the text into an expression that matches the user's age and cognitive level; If the user is under 12 years old, a cartoon icon will be used in conjunction with a short voice broadcast. If the user is 12 years or older, then text scrolling prompts and natural speech synthesis will be used. The voice or text prompts clearly point to specific body parts and directions of movement.

7. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 6, characterized in that: The generative interaction module writes the user's feedback on voice or text prompts into the personal health digital twin model. The specific steps are as follows: Within three seconds after the voice or text prompt is issued, continuously monitor the changes in the sitting posture vector, including changes in the cervical spine forward tilt angle, trunk tilt angle, and eye-to-book distance. If the absolute value of the change in the cervical spine forward tilt angle is greater than the preset response threshold, then the prompt is considered to have generated a valid response. Record the prompt type, timestamp, current environmental parameters, and user physiological state corresponding to valid responses; The recorded data is added as new training samples to the training set of the personal health digital twin model; An incremental learning algorithm is used to update the posture correction response function, which is used to predict the corrective effect of different guidance strategies on a specific user. Through continuous closed-loop feedback, the system gradually adapts to individual user behavior habits, thereby improving the effectiveness of long-term intervention.

8. The posture and light adaptive desk lamp based on artificial intelligence as described in claim 7, characterized in that: The personal health digital twin model is stored in the local encrypted storage unit of the desk lamp. The specific steps are as follows: All user behavior data, physiological data, and model parameters are processed by AES-256 encryption algorithm before being written to the storage unit. The data upload process is only allowed to be initiated when the user presses and holds a physical button for three seconds or speaks a preset voice command; Before uploading, the data is anonymized by removing identity information and retaining only the group behavior feature vector. The cloud server aggregates the de-identified feature vectors uploaded by multiple desk lamps and trains the shared weights of the global multimodal fusion network. After training is complete, an incremental update package for the model is generated and pushed to each lamp via a TLS 1.3 secure channel; When the desk lamp is in standby mode at night and connected to a power source, it automatically loads an incremental update package to update the parameters of the underlying multimodal fusion network.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the AI-based adaptive posture and light adjustment desk lamp as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the posture and light adaptive adjustment desk lamp based on any one of claims 1 to 8.

Images