Door lock control method and device and intelligent door lock

Abstract

The invention discloses a door lock control method and device and an intelligent door lock, the method is applied to the intelligent door lock, and the method comprises the following steps: obtaining a to-be-recognized image, and analyzing the to-be-recognized image to obtain facial feature data of a global face; and processing the facial feature data to obtain a blood perfusion index. Acquiring to-be-processed pulse data of the target user, and processing the to-be-processed pulse data to obtain pulse characteristic parameters. And inputting the blood perfusion index and the pulse characteristic parameter into a preset characteristic fusion model to obtain a control instruction of the intelligent door lock. According to the control instruction, the intelligent door lock is controlled to give out an early warning prompt. According to the door lock control method, signal acquisition and analysis from multiple dimensions are achieved, so that the intelligent door lock is controlled more comprehensively and accurately, finally, early warning prompt is triggered based on the control instruction, accurate and normalized monitoring of the home state of a user is achieved, and effective guarantee is provided for home safety of the user.

Description

Technical Field

[0001] This invention relates to the field of smart door lock technology, and in particular to a door lock control method, device, and smart door lock. Background Technology

[0002] With the continuous upgrading of home security and scenario-based monitoring needs, smart door locks, as high-frequency interactive entry devices, are gradually becoming an important carrier for multi-dimensional signal acquisition and early warning functions. However, existing smart door locks are too simplistic in signal acquisition and still have significant limitations in data analysis, making it difficult to meet the refined usage needs of a wide range of users.

[0003] Therefore, this application is hereby filed. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a door lock control method, device, and smart door lock, which can precisely control the door lock and improve the accuracy of door lock warnings.

[0005] According to a first aspect of the present invention, a door lock control method is provided, applied to a smart door lock, the method comprising: Obtain the image to be identified, which is an image containing the global face of the target user; The image to be identified is analyzed to obtain the facial feature data of the global face; The facial feature data is processed to obtain the blood perfusion index; Obtain the pulse data to be processed from the target user, process the pulse data to be processed, and obtain pulse feature parameters; The blood perfusion index and the pulse characteristic parameters are input into a preset feature fusion model to obtain the control command of the smart door lock; Based on the control command, the smart door lock is controlled to issue an early warning.

[0006] In some embodiments, the image to be identified includes a visible light image and an infrared light image, and the step of parsing the image to be identified to obtain the facial feature data of the global face includes: The visible light image and the infrared light image are analyzed to obtain visible light facial data and infrared light facial data; Obtain the first ambient light intensity corresponding to the visible light image and the second ambient light intensity corresponding to the infrared light image; The weights of the visible light facial data are determined based on the first ambient light intensity. The weights of the infrared facial data are determined based on the second ambient light intensity. The visible light facial data and the infrared facial data are weighted and calculated to obtain the facial feature data.

[0007] In some embodiments, processing the facial feature data to obtain the blood perfusion index includes: Based on a preset standard facial database, target facial data that matches the facial feature data is determined; Based on the target facial data, key facial regions in the global face are determined, and features are extracted from these key facial regions to obtain facial blood flow feature parameters. The blood flow characteristic parameters of the face are calculated to obtain the blood perfusion index.

[0008] In some implementations, the step of extracting features from the key facial regions to obtain facial blood flow feature parameters includes: The key facial areas are smoothed to obtain corrected data for the key facial areas, wherein the smoothing process includes smoothing of age spots and smoothing of wrinkles. The corrected data is converted to facial blood flow data. Feature extraction is performed on the facial blood flow data to obtain the facial blood flow feature parameters.

[0009] In some implementations, acquiring the pulse data to be processed collected by the smart door lock includes: Based on the smart door lock, pulse data from multiple touch points on the target user's palm are obtained; The pulse data from multiple contact points are subjected to reliability screening to obtain the pulse data to be processed.

[0010] In some embodiments, the pulse characteristic parameters include heart rate parameters, heart rate variability parameters, and pulse wave propagation time parameters. After processing the pulse data to obtain the pulse characteristic parameters, the method further includes: If any one of the heart rate parameter, the heart rate variability parameter, and the pulse wave transit time parameter exceeds the corresponding preset threshold range, the pulse data of the target user is reacquired.

[0011] In some implementations, the step of inputting the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control commands for the smart lock includes: The perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave transit time parameter are mapped to obtain the perfusion index score, heart rate score, heart rate variability score, and pulse wave transit time score, respectively. Based on the initial control parameters preset by the target user, determine the weights of the blood perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave conduction time parameter. The control parameters of the smart door lock are obtained by weighting the blood perfusion index score, the heart rate score, the heart rate variability score, and the pulse wave transit time score according to the weight of the blood perfusion index, the heart rate score, the heart rate variability score, and the pulse wave transit time score. Based on the control parameters, control commands for the smart door lock are generated.

[0012] In some implementations, the warning prompts include level one, level two, and level three warnings, and controlling the smart door lock to issue warning prompts based on the control commands includes: When the warning is a Level 1 warning, the smart door lock is controlled to provide an audible alert. When the warning is a level two warning, the smart lock is controlled to provide sound and light alerts and send warning information to the target device bound to the smart lock. When the warning prompt is a Level 3 warning, the smart door lock is controlled to sound an alarm, and a temporary rescue password is generated and sent to the target device.

[0013] According to a second aspect of the present invention, a door lock control device is provided for use in a smart door lock, the device comprising: The acquisition module is used to acquire an image to be identified, which is an image containing the global face of the target user, and is also used to acquire the target user's pulse data to be processed. The parsing module is used to parse the image to be identified and obtain the facial feature data of the global face; The processing module is used to process the facial feature data to obtain the blood perfusion index, and also to process the pulse data to be processed to obtain pulse feature parameters; The instruction generation module is used to input the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control instructions for the smart door lock. The control module is used to control the smart door lock to issue an early warning prompt based on the control command.

[0014] According to a third aspect of the present invention, a smart lock is provided, the smart lock including the lock control device in the above embodiments, the lock control device being used to execute the lock control method described in any of the above embodiments or implementation methods.

[0015] The door lock control method provided in this application obtains facial feature data by parsing the collected image to be identified, processes it to extract the blood perfusion index, and performs targeted processing on the pulse data to obtain pulse feature parameters, enabling precise capture of user information. Based on a feature fusion model, the above-mentioned multi-dimensional user data is comprehensively analyzed and control commands are output to control the smart door lock, avoiding the limitations of single data analysis, thus achieving more comprehensive and accurate control of the smart door lock. Finally, based on the control commands, an early warning prompt is triggered, enabling the smart door lock to provide more accurate intelligent early warnings, achieving convenient, accurate, and routine monitoring of the user's home status, and providing effective protection for the user's home security. Attached Figure Description

[0016] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart illustrating the steps of a door lock control method provided by the present invention; Figure 2 This is a schematic diagram of the steps for obtaining facial feature data by parsing an image, as provided by the present invention. Figure 3 This is a schematic diagram of the structure of a door lock control device provided by the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions of this application, the application will be described in detail below with reference to the accompanying drawings and specific embodiments. Although the accompanying drawings and specific embodiments describe exemplary embodiments of the present invention, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0018] The terms "first," "second," and similar words used in this application do not indicate any order, quantity, or importance, but are merely used for distinction. The terms "including," and similar words used in this application mean that the element preceding the word encompasses the elements listed after the word, and do not exclude the possibility of including other elements. The technical solutions of this application are not limited to the execution order described in the embodiments. The steps in the execution order can be combined, decomposed, or their order can be changed, as long as the logical relationship of the execution content is not affected.

[0019] All terms used in this application (including technical or scientific terms) have the same meaning as understood by one of ordinary skill in the art to which this application pertains, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant art, and not as idealized or highly formalized, unless expressly defined herein. Technologies and equipment known to one of ordinary skill in the art may not be discussed in detail, but where appropriate, such technologies and equipment should be considered part of the specification.

[0020] With the continuous upgrading of home security and scenario-based monitoring needs, smart door locks, as high-frequency interactive entry devices, are gradually becoming important carriers for multi-dimensional signal acquisition and early warning functions. However, existing smart door locks are too limited in signal acquisition and have significant limitations in data analysis, making it difficult to meet the refined usage needs of a wide range of users. Therefore, there is an urgent need for a door lock control method and smart door lock that can acquire and analyze signals from multiple dimensions to more comprehensively and accurately assess the user's current state, and can issue timely warnings when the user's state is abnormal, achieving a seamless experience of monitoring upon unlocking and issuing warnings upon anomalies.

[0021] To address the aforementioned issues, this application provides a door lock control method applied to smart door locks. This method enables multi-dimensional signal acquisition and analysis, thereby providing a more comprehensive and accurate assessment of the user's current state and meeting the refined usage needs of different user groups. For example... Figure 1 As shown, Figure 1 This is a schematic flowchart of a door lock control method provided by the present invention. The method may include the following steps S110-S150.

[0022] S110: Obtain the image to be identified, which is an image containing the global face of the target user.

[0023] In this embodiment, the image to be identified can be an image captured by the built-in camera of the smart lock after the target user triggers the unlocking process (such as pressing the wake-up button or approaching the sensing area), which can fully present the key areas of the target user's face (including the forehead, eyes, nose, mouth, and cheeks).

[0024] The target users can be those who have already registered their identity with the smart lock (such as through fingerprint, password or facial recognition), and are usually permanent residents of the residence where the smart lock is located.

[0025] A global face image can be the complete facial area of ​​the target user, from the hairline to the bottom of the chin and from the left and right sides to the edges of the cheeks. This avoids the loss of facial features due to occlusion (such as a mask completely covering the mouth and nose or a hat covering the forehead). For example, when the target user is standing 30-50 cm away from the smart lock and facing the lock panel, the camera can capture a global face image without obvious occlusion. If the detected facial occlusion area exceeds 30%, the smart lock can prompt the user to adjust their posture with voice prompts such as "Please adjust your face position to ensure no occlusion" until a suitable image for recognition is captured.

[0026] Specifically, the smart lock panel can integrate an image acquisition module (such as a camera or video module), whose activation timing is linked to the target user's unlocking behavior. When the smart lock detects a target user standing within 30-50cm of the lock and intending to unlock it (e.g., hand near the lock handle) via its built-in millimeter-wave radar or infrared human sensor, it automatically activates the image acquisition module. The image acquisition module uses hardware configuration adapted to the user's unlocking scenario (e.g., at least 3 megapixels, supporting common lighting environments) to acquire a frontal image of the target user's face (the acquisition angle is adapted to the target user's facial height when standing to unlock, and the acquisition time is synchronized with the unlocking process, ensuring that at least one clear frame of facial data is obtained without significant obstruction (e.g., no large area of ​​the face is covered by a mask or hat).

[0027] S120: Analyze the image to be identified to obtain the facial feature data of the global face.

[0028] In this step, to improve the reliability of facial feature data, preprocessing (such as noise reduction, region filtering, etc.) and blood flow feature extraction can be performed on the image to be recognized. The purpose is to eliminate environmental interference (such as uneven lighting, facial wrinkles or age spots, etc.) and highlight the signals that reflect the blood flow status of facial capillaries.

[0029] Specifically, the image to be recognized can first be preprocessed, such as by Gaussian filtering (kernel size 3×3), to filter out local pixel noise caused by blemishes and wrinkles. Then, a face detection algorithm (such as Adaboost) is used to locate key facial regions, selecting areas with thinner skin, denser capillary distribution, and less interference, such as the upper middle part of the forehead and the tip of the nose, as regions of interest (ROIs), while excluding occluded areas such as hair and eyeglass frames. Finally, skin color normalization is performed to eliminate baseline differences in skin color (lighter, yellower, darker) among different users, ensuring consistency in subsequent facial feature data extraction.

[0030] In this embodiment, facial feature data can be feature parameters obtained through image analysis that reflect the distribution of facial skin and blood, including the average grayscale value of each facial region (forehead, left cheek, right cheek), the pixel value distribution of the three primary colors (Red, Green, Blue, RGB), and skin texture variation coefficients. For example, for a frame of the image to be identified, 68 facial key points are located using an algorithm. Then, taking the forehead region (the region consisting of key points 1-10) as the analysis object, the average value of the R channel (i.e., the average value of the red component) of all pixels in this region is calculated. This average value is one of the facial feature data of the forehead region.

[0031] S130: Process the facial feature data to obtain the blood perfusion index.

[0032] In this embodiment, the perfusion index (PI) is a parameter that quantifies the richness of blood flow in the subcutaneous capillaries of the face. A higher value usually indicates more sufficient blood flow to the face. The value range is generally 0.1-2.0 (for a user's normal resting state, the PI is usually between 0.8 and 1.5).

[0033] Feature extraction was performed on the preprocessed facial feature data. Based on the preprocessed ROI region image, the skin's reflectivity to specific wavelengths of light was analyzed using photoplethysmography (PPG). Since the absorption of visible light (such as 550nm green channel light) by hemoglobin fluctuates with blood flow, the perfusion index was obtained by calculating the ratio of the fluctuating to the static blood flow component in the ROI region image.

[0034] S140: Obtain the pulse data to be processed from the target user, process the pulse data to be processed, and obtain pulse feature parameters.

[0035] In this embodiment, the pulse data to be processed refers to signal data that can reflect the cardiovascular pulsation state of the target user, specifically the blood flow volume change signal generated when the palmar artery of the target user pulsates. This signal can be used to extract key parameters such as heart rate and heart rate variability.

[0036] Specifically, a pulse acquisition module can be integrated into the handle grip area of ​​the smart door lock. This module can employ a photoplethysmography (PPG) sensor. For example, to accommodate the physiological characteristics of users with uneven grip strength and varying hand sizes, the sensor is typically designed with an arc-shaped structure that conforms to the hand grip posture. The activation timing of the pulse acquisition module is synchronized with the facial data acquisition in step S110. When the target user grips the door lock handle, the pressure sensor built into the handle (range 0-100N) detects a contact pressure greater than or equal to 5N (the lower limit of the user's normal grip strength, ensuring effective contact), and automatically activates the PPG sensor. The PPG sensor emits light of a specific wavelength (such as 660nm red light and 940nm near-infrared light) that penetrates the skin of the hand, receives the reflected light signal, and converts it into an electrical signal. This electrical signal is the pulse data to be processed. The sampling rate is no less than 100Hz (to ensure that the complete pulse fluctuation cycle can be captured), and the acquisition duration is consistent with the facial data acquisition (e.g., 1-2 seconds).

[0037] In this embodiment, the processing of pulse data refers to the filtering, denoising, and feature extraction operations performed on the acquired raw pulse electrical signals. The purpose is to eliminate interference signals such as hand tremors and grip strength fluctuations, and to extract key parameters that reflect cardiovascular status. Pulse feature parameters refer to physiological parameters strongly correlated with the user's status extracted from the pulse data to be processed, including at least heart rate (HR), and may further include heart rate variability (HRV), pulse wave velocity (PWV), etc., as needed.

[0038] Specifically, the pulse data to be processed is first preprocessed, including bandpass filtering of the raw pulse electrical signal to remove noise such as respiratory interference and muscle tremors. Then, the signal baseline drift (the overall signal shift caused by changes in grip strength) is corrected using a moving average method to obtain a stable pulse signal waveform. Next, heart rate features are extracted. Peak detection algorithms (such as adaptive thresholding) are used to identify each pulsation peak in the pulse signal waveform, and the time interval between two adjacent peaks (pulse cycle) is calculated. Then, based on: Heart rate = 60 / pulse cycles, which is used to calculate heart rate in beats per minute (bpm).

[0039] In some other embodiments, if heart rate variability (HRV) needs to be extracted, the standard deviation of consecutive normal pulse cycles is calculated. If pulse wave conduction time (PWV) needs to be extracted, the rising edge inflection point of the pulse signal waveform is located and estimated in conjunction with a preset arterial conduction distance (based on the user's average physiological parameters).

[0040] S150: Input the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control command of the smart door lock.

[0041] Specifically, the blood perfusion index and the pulse characteristic parameters can be input into a preset feature fusion model to obtain control commands for the smart lock. These control commands are operational instructions used to regulate the core functions of the lock (including the early warning module, unlocking process, and data recording module). The lock can execute specific actions through these control commands.

[0042] In some implementations, the blood perfusion index and pulse characteristic parameters can be processed using a preset feature fusion model to obtain control parameters, and then control commands can be generated based on these control parameters. The control parameters can be quantitative indicators of the user's current state. To reflect the quantitative value of the target user's current state, a range of 0-100 can be set, with higher values ​​indicating better user condition (e.g., 80-100 indicates a normal state, 60-79 indicates a slight abnormality, and less than 60 indicates a significant abnormality). If the control command is obtained through conversion of the control parameters, the command type is directly related to the "normal / abnormal" determination result of the control parameters.

[0043] Furthermore, the control parameter calculation involves using the blood perfusion index obtained in step S130 and the pulse characteristic parameters obtained in step S140 as model inputs. The model outputs control parameters through weight allocation (e.g., blood perfusion index accounts for 40%, heart rate accounts for 30%, HRV accounts for 20%, and PWV accounts for 10%) and feature calculation.

[0044] It should be noted that the feature fusion model is trained based on multiple historical blood perfusion indices and multiple historical pulse characteristic parameters. The feature fusion model can be a machine learning model capable of comprehensively analyzing blood perfusion indices and pulse characteristic parameters. Its core function is to transform multi-dimensional physiological parameters into a single, quantifiable control parameter. The multiple historical blood perfusion indices and multiple historical pulse characteristic parameters refer to the sample data used during the model training phase. This sample needs to cover users of different ages and health conditions to ensure the model's adaptability to different user groups.

[0045] Specifically, the model training involves using a lightweight model structure suitable for edge computing (smart locks have limited computing power) (such as a weighted fusion model, a shallow neural network model, etc.). The training samples are derived from the physiological data collection of more than 1,000 users (blood perfusion index and pulse characteristic parameters are collected for each user under different scenarios such as resting and after activity, with 10-20 sets of data collected for each scenario).

[0046] S160: Based on the control command, control the smart door lock to issue a warning.

[0047] It should be noted that in some other embodiments, the smart lock can also send alert information to the target device and / or the bound device. The target device can refer to a terminal device used to receive information from the lock, typically a mobile or fixed terminal associated with the target user. The bound device refers to a device that has been paired with the smart lock through its accompanying application (APP). The pairing process requires verifying the user's identity (e.g., entering the smart lock's management password or scanning the QR code generated by the lock) to ensure the security of information transmission.

[0048] In this embodiment, the warning prompt includes at least the control parameter value and the status description (such as "control parameter 92 points, current status is normal", "control parameter 58 points, current status is abnormal, attention is recommended", etc.). The specific values ​​of key physiological parameters (such as heart rate and blood perfusion index) can be added as needed.

[0049] Specifically, smart door locks can use built-in communication modules (such as Bluetooth or WiFi modules) to transmit warning information to the target device in an encrypted manner (such as AES encryption) according to preset transmission rules (such as real-time transmission or timed summary transmission) upon receiving a control command. After receiving the physiological information, the target device will remind the user to check through APP pop-ups, message notifications, etc. (If the target device is the user's child's smartphone, a simple prompt of "target user's status is normal / abnormal" can also be sent simultaneously).

[0050] The door lock control method provided in this embodiment obtains facial feature data by parsing the acquired image to be identified, and processes it to extract the blood perfusion index. It also performs targeted processing on the pulse data to obtain pulse feature parameters, enabling precise capture of user information. Based on a feature fusion model, the multi-dimensional user data is comprehensively analyzed, and control commands are output to control the smart door lock. This avoids the limitations of single data analysis, thus providing more comprehensive and accurate control of the smart door lock. Finally, based on the control commands, an early warning prompt is triggered, enabling the smart door lock to provide more accurate intelligent warnings. This achieves convenient, accurate, and routine monitoring of the user's home status, providing effective protection for the user's home security.

[0051] In some embodiments, such as Figure 2 As shown, Figure 2 This is a schematic flowchart illustrating the steps for obtaining facial feature data by analyzing an image, as provided by the present invention. The image to be identified includes a visible light image and an infrared light image. The step of analyzing the image to be identified to obtain the facial feature data of the global face includes the following steps S210-S240.

[0052] S210: Analyze the visible light image and the infrared light image to obtain visible light facial data and infrared light facial data.

[0053] In this embodiment, the visible light global facial data can be facial image data acquired using a visible light source, typically ranging from 400-700nm (covering the red, green, and blue light perceptible to the human eye). This data clearly reflects details such as facial skin texture and skin tone, and especially accurately captures skin tone changes caused by blood flow in facial capillaries under normal lighting conditions. The infrared light global facial data can be facial image data acquired using an infrared light source, typically ranging from 700-1100nm (invisible to the human eye). This data is less affected by ambient light interference and can stably reflect the blood flow distribution characteristics of subcutaneous facial tissue even in low-light or backlight conditions.

[0054] Specifically, the image acquisition module of a smart door lock can integrate a dual-light source camera or two independent cameras (one adapted for visible light acquisition, and one adapted for infrared light acquisition). The visible light acquisition unit uses an RGB camera, and the infrared light acquisition unit uses an infrared camera (e.g., equipped with an 850nm infrared LED supplement light). When a target user is detected approaching the door lock, both acquisition units start simultaneously, acquiring the target user's facial image within 1-2 seconds. The visible light acquisition unit outputs visible light facial data in RGB format, and the infrared light acquisition unit outputs infrared facial data in grayscale format. The acquisition areas and frame rates of both are consistent to ensure a one-to-one correspondence of image pixels during subsequent weighted calculations.

[0055] For example, when a target user unlocks the door in the hallway in the evening (when the ambient light is weak), the RGB camera of the smart door lock captures a visible light image containing the facial area, while the infrared camera captures an infrared image of the same area with the assistance of an 850nm fill light. The visible light image has slight noise due to insufficient light, while the infrared image clearly shows the facial contours and subcutaneous blood flow distribution.

[0056] In some other implementations, the image acquisition module of the smart door lock can also integrate multiple independent cameras to acquire visible light and infrared light, and this application embodiment does not specifically limit it.

[0057] S220: Obtain the first ambient light intensity corresponding to the visible light image and the second ambient light intensity corresponding to the infrared light image.

[0058] In this embodiment, the first ambient light intensity and the second ambient light intensity are the same. Ambient light intensity (i.e., the first ambient light intensity and the second ambient light intensity) refers to the illumination intensity of the area where the target user's face is located, and the unit is lux. This parameter is used to evaluate the reliability of global facial data in visible light and infrared light (too strong or too weak ambient light will affect the quality of visible light data and infrared light data), and thus determine the weight allocation of visible light facial data and infrared light facial data.

[0059] Specifically, the smart door lock panel can integrate a digital light sensor next to the image acquisition module (no more than 5cm from the center of the camera). The sensor has a detection range of 0-65535 lux (covering common scenarios for users to unlock the door, such as about 10000-30000 lux under direct sunlight at noon, about 100-500 lux under indoor lighting, and about 10-50 lux in the corridor at night). The sampling frequency is synchronized with the acquisition of the image to be recognized (acquiring once every 100ms), and the average light intensity during the acquisition period of the image to be recognized is taken as the final ambient light intensity.

[0060] S230: Determine the weight of the visible light facial data based on the first ambient light intensity; determine the weight of the infrared light facial data based on the second ambient light intensity.

[0061] In this embodiment, the weights of visible light facial data and infrared facial data refer to the proportions of visible light facial data and infrared facial data in the final fusion result (the sum is 100%). The weight allocation is as follows: the closer the ambient light intensity is to the ideal range of visible light acquisition (typically 500-5000 lux), the higher the weight of visible light facial data. The further the ambient light intensity deviates from the ideal range (too strong or too weak), the higher the weight of infrared facial data, so as to compensate for the deficiencies of visible light data through the anti-interference characteristics of infrared light.

[0062] Specifically, the preset relationship between ambient light intensity and weight can be: When the ambient light intensity is greater than or equal to 10,000 lux (strong light or backlight scene, such as a midday balcony), the visible light facial data weight is 30%, and the infrared light facial data weight is 70% (to avoid blood flow signal loss caused by overexposure of visible light images).

[0063] When the ambient light intensity is greater than or equal to 5000 lux but less than 10000 lux (in strong light scenarios, such as indoor windows on a sunny day), the visible light facial data weight is 50%, and the infrared light facial data weight is 50%.

[0064] When the ambient light intensity is greater than or equal to 500 lux but less than 5000 lux (ideal lighting scenario, such as indoors on a cloudy day), the visible light facial data has a weight of 80%, and the infrared light facial data has a weight of 20% (prioritizing the preservation of the detail advantage of visible light).

[0065] When the ambient light intensity is greater than or equal to 100 lux but less than 500 lux (low-light scene, such as a corridor at dusk), the weight of visible light facial data is 60%, and the weight of infrared light facial data is 40%.

[0066] When the ambient light intensity is less than 100 lux (dark light scene, such as late at night), the visible light facial data weight is 20%, and the infrared light facial data weight is 80% (mainly relying on infrared light data).

[0067] S240: Based on the weights of the visible light facial data and the infrared facial data, perform a weighted calculation on the visible light facial data and the infrared facial data to obtain the facial feature data.

[0068] In this embodiment, the purpose of weighted calculation is to combine the advantages of visible light facial data and infrared light facial data (the richness of detail in visible light and the anti-interference capability of infrared light) to generate more reliable facial data, providing high-quality input for subsequent blood perfusion index calculation.

[0069] Specifically, image alignment is the first step. Since visible light facial data and infrared facial data may have slight differences in acquisition angle, the two images are first aligned pixel-level by feature point matching (e.g., locating five key feature points such as the eyes and nose tip) to ensure that pixels at the same facial location have consistent coordinates in both images. Then, pixel value weighting is performed. For each pixel in the aligned image, if it is visible light data (RGB format), its green channel pixel value (G value, most sensitive to blood flow changes) is used. If it is infrared data (grayscale format), its grayscale value is used. Finally, the fused pixel value is calculated according to the weights. The fused pixel value = (visible light facial data pixel value × first weight) + (infrared light facial data pixel value × second weight).

[0070] Finally, image enhancement is performed. The fused image is contrast-stretched (pixel values ​​are normalized to 0-255) to further highlight facial blood flow-related feature areas.

[0071] For example, in a hallway scene at dusk, a pixel in the aligned visible light image has a G value of 120, while the pixel at the same location in the infrared image has a grayscale value of 180. Calculated with weights of 20% and 80%, the fused pixel value = 120 × 0.2 + 180 × 0.8 = 24 + 144 = 168. After performing the same calculation and enhancement on all pixels, the final facial data is obtained. This data preserves local details in the visible light while suppressing noise in low-light environments through infrared light.

[0072] Through the method of this embodiment, the smart door lock can adaptively fuse visible light facial data and infrared light facial data according to different ambient light intensities, which solves the problem of poor data quality of a single light source in backlight, low light and other scenarios, reduces the calculation error of the subsequent blood perfusion index to within 10%, and significantly improves the reliability of user status detection in complex lighting environments.

[0073] In some embodiments, processing the facial feature data to obtain the blood perfusion index includes the following steps S310-S330.

[0074] S310: Determine target facial data that matches the facial feature data based on a preset standard facial database.

[0075] In this embodiment, a preset standard facial database can be stored in the local storage of the smart lock. This database may contain facial data (including coordinates of 68 key points and RGB mean values ​​for each region) of 2000 users of different ages and skin tones, categorized by age and skin tone and indexed, with support for updates. The target facial data can be a single standard facial data entry selected from the standard facial database that has the highest similarity to the facial feature data to be processed.

[0076] Specifically, the first step is to construct a facial feature template for the target user. When the target user uses the smart lock for the first time, the lock's image acquisition module captures 3-5 clear facial images from different angles (frontal view and 15° side view on each side). A feature point detection algorithm (such as the Dlib algorithm) is used to extract the coordinate information of 68 key facial feature points in each image, generating the user's facial feature template and storing it locally on the lock (using AES-256 encryption to prevent data leakage).

[0077] Furthermore, when processing facial feature data, the system first filters the corresponding subset of the database based on the user's pre-entered basic information (basic information refers to the user's age range, skin color, and facial feature templates, etc., which are pre-stored by the smart lock during the initial registration phase of the target user; the templates contain key facial feature information of the target user under different postures and lighting conditions). The cosine similarity algorithm is used to calculate the similarity between the data to be processed and the data in the subset. Standard data with a similarity greater than or equal to 0.85 is selected as the target facial data. If there is no matching data, the average data in the database is used as a temporary substitute.

[0078] Specifically, when performing face comparison of a target user, facial feature data is input into the target user's facial feature template. The facial feature template is then compared with the target user's facial feature template using a feature point matching algorithm (such as Euclidean distance calculation). When the matching similarity is greater than or equal to 90%, the user is determined to be an authorized target user.

[0079] The cosine similarity calculation uses the cosine similarity algorithm. Specifically, the key feature vectors of the facial feature data to be processed and a standard facial data set from the standard facial database are extracted first. Each key feature vector contains the x-coordinate difference (with the left pupil as the origin, the difference between the x-coordinate of each point and the origin's x-coordinate), the y-coordinate difference (the difference between the y-coordinate of each point and the origin's y-coordinate), and the RGB channel mean values ​​of 10 typical regions (forehead, left and right cheeks, bridge of the nose, chin, etc.), totaling 68×2 + 10×3 = 166 dimensions. Then, the cosine similarity between the two 166-dimensional feature vectors is calculated. The similarity value ranges from 0 to 1, with values ​​closer to 1 indicating higher similarity. To improve matching efficiency, the system first filters the data subset corresponding to the category from the standard facial database based on the user's basic information (such as age range and skin color type pre-entered by the user through the APP). For example, if the user enters an age range of 31-45 years old and a skin color type of 5, only 200 data entries under that category are retrieved. Then, the cosine similarity between each data entry and the facial feature data to be processed is calculated in the data subset. The standard facial data with the highest similarity and a value ≥0.85 is determined as the target facial data. If the similarity of all data in the data subset is <0.85, the filtering range is expanded to the category data of adjacent age ranges (such as 26-50 years old), and the similarity is recalculated. If no data meets the conditions, the average facial data of the standard facial database (the average value of the key feature vectors of all standard facial data) is used as the temporary target facial data, and marked as "low matching accuracy" in the data recording module for subsequent database optimization.

[0080] S320: Based on the target facial data, determine the key facial regions in the global face, extract features from the key facial regions, and obtain facial blood flow feature parameters.

[0081] In this embodiment, the key facial region can be a highly compatible blood flow extraction area marked by a facial template matching the target user selected from a standard facial database. For example, for elderly users' facial features (loose skin, many age spots, and easy to wear reading glasses), the key facial regions are preferentially the upper middle part of the forehead (located 2-3 cm above the eyebrows, avoiding the hairline and areas with dense forehead wrinkles) and the tip of the nose (avoiding the common distribution area of ​​age spots on both sides of the nose). The area of ​​both regions is set to 20×20 pixels (balancing signal strength and anti-interference ability). If there is local occlusion in the real-time facial feature data (such as the edge of reading glasses obscuring the edge of the forehead), the boundary of the key facial region is automatically adjusted to ensure that the selected area is unobstructed. For ordinary users, the upper middle part of the forehead and the middle of the left and right cheeks are preferentially selected as key facial regions because the skin is thin (estimated thickness < 0.5 mm), the blood vessels are dense (density value > 0.7), and there is little occlusion. For the aforementioned region, RGB channel pixel value change curves were extracted from 30 frames of images to be identified. The fluctuation amplitude and period of the curves were calculated to obtain 18 facial blood flow feature parameters. Finally, a weighted fusion algorithm was used to first calculate the local index of each region according to "local index = (fluctuation amplitude / 255) × (10 / period mean)", and then combine the regional score (combined score of skin thickness and blood vessel density) to calculate the global blood flow perfusion index, ensuring that the result is within the normal range of 0.1-10.

[0082] Feature extraction of key facial regions involves extracting a set of pixels within that region, containing information about the skin's reflection or absorption of light, which can indirectly reflect changes in blood flow in subcutaneous capillaries. Facial blood flow feature parameters refer to quantitative indicators extracted from local facial data that are directly related to blood flow status, including at least the blood flow fluctuation component (AC) and the blood flow static component (DC). The AC component reflects the periodic changes in blood flow with the heartbeat, while the DC component reflects the baseline value of light absorption by static structures such as skin tissue and blood vessel walls. Together, they constitute the core basis for calculating the blood perfusion index.

[0083] Specifically, the processing can begin by cropping the target facial data. Based on the feature extraction region coordinates determined in the above process, corresponding local images (such as 20×20 pixel images of the upper part of the forehead and 20×20 pixel images of the tip of the nose) are cropped from the target facial data, and the local data of the two regions are processed separately (the average value is then taken to improve stability).

[0084] Next, noise filtering is performed. The key facial areas are preprocessed. First, a 3×3 kernel Gaussian filter is used to eliminate local pixel noise caused by age spots and wrinkles. Then, adaptive histogram equalization is used to enhance image contrast and highlight pixel value differences caused by blood flow changes.

[0085] Finally, component extraction was performed. Blood flow characteristic parameters were extracted using photoplethysmography (PPG). For the preprocessed key facial regions, the pixel value change curve for each pixel was calculated within the acquisition period (1-2 seconds). The average of all pixel change curves was then taken to obtain the average pixel value time series for that region. A Fourier transform was performed on this time series to separate the periodically fluctuating AC component (corresponding to blood flow pulsation) and the constant DC component (corresponding to static tissue absorption). The AC component was calculated as the difference between the peak and trough values ​​of the fluctuation, while the DC component was calculated as the average of the time series.

[0086] S330: Calculate the facial blood flow characteristic parameters to obtain the blood perfusion index.

[0087] In this embodiment, the Perfusion Index (PI) is calculated using facial blood flow characteristic parameters (AC and DC components). It is a quantitative indicator reflecting the richness of blood flow in the subcutaneous capillaries of the face, and its value directly corresponds to the adequacy of blood flow. A higher value indicates more adequate blood flow (the PI is typically 0.8-1.5 for normal users at rest), while a low value suggests potential physiological risks such as hypoxia and peripheral circulatory disorders.

[0088] Specifically, the baseline PI value can be calculated first. The industry-standard blood perfusion index calculation formula is used: PI = AC component / DC component, where the AC component is the average of the blood flow fluctuation components obtained in step S320, and the DC component is the average of the blood flow static components. Then, calibration correction is performed. Considering that differences in light absorption due to different skin tones (fair, yellow, dark) among elderly users may affect the calculation results, a skin tone calibration coefficient is introduced. This coefficient is determined based on the skin tone baseline value of the target user during initial registration (the average skin tone of the user at rest is collected during the registration phase and compared with a standard skin tone template to obtain the calibration coefficient, ranging from 0.9 to 1.1). The baseline PI value is multiplied by the calibration coefficient to obtain the final blood perfusion index, thus eliminating errors caused by skin tone differences.

[0089] This embodiment introduces a pre-defined standard facial database, providing a precise reference benchmark for processing facial feature data. This effectively solves the problem in existing technologies where the lack of a unified reference standard leads to significant differences in the processing results of facial feature data from different users. By matching target facial data from the standard facial database, the appropriate processing criteria can be determined based on the user's individual facial features (such as age group, skin color, and facial structure). This avoids the bias that occurs when using a unified standard to process data from different users, significantly improving the targeting of subsequent key facial regions. Simultaneously, by determining key facial regions based on the target facial data, regions with thin skin, dense blood vessels, and minimal interference are prioritized. This ensures that the extracted facial blood flow feature parameters more accurately reflect the user's blood flow status, reducing the interference of irrelevant region data and further improving data effectiveness. Furthermore, by weighted fusion of blood flow feature parameters from multiple key regions to calculate the blood perfusion index, the blood flow information from different regions is fully integrated, avoiding the limitations of single-region data. The weighting process incorporates the region's own scoring (based on factors such as skin thickness and blood vessel density), making the calculation results more scientific and accurate. This solution significantly improves the accuracy and reliability of blood perfusion index calculation through standardized database matching, personalized key area identification, and multi-region data fusion calculation. This lays a solid foundation for generating accurate control parameters for subsequent feature fusion models, thereby ensuring that smart locks can execute subsequent control commands more accurately. It effectively improves the performance of smart locks in the data processing stage, better meets the usage needs of a wide range of users in different scenarios, and also provides an scalable implementation path for optimizing smart lock data processing technology. This facilitates further improvement of the standard facial database by combining more user data, continuously enhancing the processing effect.

[0090] In some embodiments, the step of extracting features from the key facial regions to obtain facial blood flow feature parameters includes: smoothing the key facial regions to obtain corrected data for the key facial regions, wherein the smoothing includes smoothing age spots and wrinkles; performing facial blood flow degree transformation on the corrected data to obtain facial blood flow data; and extracting features from the facial blood flow data to obtain the facial blood flow feature parameters.

[0091] In this embodiment, the key facial areas are first smoothed. A combined algorithm of "3×3 kernel Gaussian filter (standard deviation 1.2) + bilateral filter (spatial domain standard deviation 1.5, gray domain standard deviation 20)" can be used. The Gaussian filter first weakens the isolated pixel abrupt changes of age spots, and the bilateral filter smooths the light and dark stripes caused by wrinkles while preserving the edges of blood flow signals, finally obtaining locally corrected facial data that eliminates the interference of age spots and wrinkles.

[0092] Next, facial blood flow conversion is performed. The green channel pixel values ​​of the corrected data of key facial areas can be extracted (hemoglobin absorbs 550nm green light most significantly), the dynamic change rate of "(current frame pixel value - reference frame pixel value) / reference frame pixel value" can be calculated, and then multiplied by the experimentally calibrated blood flow concentration conversion coefficient of 0.85 to obtain facial blood flow data reflecting subcutaneous blood flow concentration (in units of relative blood flow concentration).

[0093] Finally, facial blood flow characteristic parameters are extracted. A sequence of 30-60 facial blood flow data points within 1-2 seconds is obtained. After low-pass filtering at 0.5Hz, a trend line is extracted using a moving average method (window of 5 data points) as the static blood flow component (DC). The trend line is then subtracted from the original sequence, and the difference between the peak and trough values ​​of the residual sequence is taken as the fluctuation blood flow component (AC).

[0094] This solution addresses the interference caused by age spots and wrinkles on the faces of elderly users. It employs a combined smoothing algorithm to accurately eliminate noise while preserving blood flow signals, overcoming the signal loss problem inherent in traditional filtering. Based on the light absorption characteristics of hemoglobin and experimental coefficients, pixel values ​​are transformed into intuitive blood flow data, enhancing the scientific rigor of the quantification. Finally, the AC and DC components are separated to provide high-quality parameters for calculating the blood perfusion index, keeping the calculation error within a certain range. The entire process is adapted to the physiological characteristics of elderly users, with a short processing time on the door lock edge chip, not affecting the unlocking experience, and balancing monitoring accuracy with practicality.

[0095] In some embodiments, obtaining the pulse data to be processed collected by the smart door lock includes: obtaining multiple touch point pulse data of the target user's palm based on the smart door lock; performing reliability screening on the multiple touch point pulse data to obtain the pulse data to be processed.

[0096] In this embodiment, the pulse data of multiple touch points on the target user's palm can refer to the pulse signal data of different parts of the user's palm that are simultaneously collected through multiple sensor touch points in the grip area of ​​the smart door lock handle.

[0097] Specifically, the handle of a smart door lock can be designed with an arc shape to fit the size of a user's hand. Four independent PPG pulse sensor contacts are evenly distributed in the grip area (e.g., the main contact is located in the corresponding position in the palm, and the auxiliary contacts are located in the corresponding areas at the base of the index, middle, and ring fingers). Each contact is equipped with a pressure sensing unit (e.g., detecting whether the contact pressure is greater than or equal to 5N) and a photoplethysmography module (e.g., emitting 660nm red light and 940nm near-infrared light). When the target user grips the handle, the pressure sensing unit synchronously triggers the PPG modules of all contacts, which can collect raw pulse signals for 1-2 seconds at a 100Hz sampling rate (each contact outputs an independent voltage fluctuation sequence), forming pulse data from multiple contacts.

[0098] It should be noted that in some other embodiments, the number of PPG pulse sensor contacts may also be other, and this application embodiment does not specifically limit them.

[0099] Reliability screening of multiple contact point pulse data can be performed by evaluating the quality of each group of data using preset indicators, eliminating invalid data due to interference, and retaining the most stable valid data. For example, screening indicators may include signal strength (peak-to-peak value greater than or equal to 0.5V to ensure sufficient signal clarity), signal-to-noise ratio (greater than or equal to 15dB to exclude signals drowned out by noise), and waveform integrity (the number of identifiable peaks in 5 consecutive pulse cycles greater than or equal to 4 to avoid waveform breaks). The specific screening process can involve first calculating the above indicators for each group of contact point data, and then eliminating data that fails to meet any of these indicators. If there are 2 or more remaining groups of data, the fluctuation coefficient of each group of data is further calculated (standard deviation / mean of consecutive pulse cycles), and the group with the smallest fluctuation coefficient is selected as the target contact point pulse data (the smaller the fluctuation coefficient, the more stable the signal). If only 1 group of data remains, it is directly used as the target data.

[0100] This embodiment, by setting multiple pulse acquisition contacts in the smart door lock handle and performing reliability screening, can adapt to the physiological characteristics of users with uneven grip strength and large differences in hand size. It avoids data loss caused by grip position deviation at a single contact point, significantly improving the success rate of pulse data acquisition. Simultaneously, through a multi-index screening mechanism, interference signals caused by hand tremors or poor contact can be effectively eliminated, ensuring the stability and integrity of the target contact point's pulse data. This provides high-quality input for the accurate extraction of subsequent pulse feature parameters, thereby improving the reliability of data during user status monitoring. Furthermore, the entire process requires no conscious adjustment of the user's grip posture, perfectly meeting the user's need for simplified operation and balancing technical practicality with user experience.

[0101] In some embodiments, the pulse feature parameters include heart rate parameters, heart rate variability parameters, and pulse wave transit time parameters. After processing the pulse data to obtain the pulse feature parameters, the method further includes: when any one of the heart rate parameters, the heart rate variability parameters, and the pulse wave transit time parameters exceeds the corresponding preset threshold range, re-acquiring the target user's pulse data.

[0102] In this embodiment, pulse characteristic parameters include heart rate parameters, heart rate variability parameters, and pulse wave propagation time parameters. The heart rate parameter refers to the number of heartbeats per minute (reflecting the frequency of heartbeats), the heart rate variability parameter refers to the degree of fluctuation in the interval between consecutive heartbeats (reflecting the autonomic nervous system's regulatory function), and the pulse wave propagation time parameter refers to the time it takes for the pulse wave to travel from the heart to the arteries in the hand (indirectly reflecting vascular elasticity).

[0103] The preset threshold ranges can be set by the target user based on their own situation, or they can be set based on the normal fluctuation ranges of various parameters according to the physiological characteristics of the user group. Specifically, the preset range for blood perfusion index can be 0.7-1.6, the preset range for heart rate parameter can be 50-100 beats / minute, the preset range for heart rate variability parameter can be 20-100 milliseconds, and the preset range for pulse wave conduction time parameter can be 80-150 milliseconds. These thresholds are implemented through a parameter configuration table stored locally on the smart lock, and can be personalized and fine-tuned based on the target user's basic data (such as routine physical examination information entered during registration).

[0104] After processing the pulse data to obtain the aforementioned parameters, the system automatically compares each parameter with its corresponding preset threshold range. If any parameter exceeds the range, a re-acquisition mechanism is triggered. The smart door lock can emit a slight vibration through the vibration module built into the handle while keeping the pulse acquisition module active, waiting for the target user to grasp the handle again. The re-acquired pulse data is processed using the same procedure to obtain new parameters. If all new parameters are within the preset range, the system proceeds to the subsequent feature fusion step. If any parameters still exceed the range after re-acquisition, the current parameters are treated as valid data and processing continues.

[0105] This embodiment effectively eliminates abnormal data caused by accidental factors (such as a user gripping the handle forcefully or slight hand tremors during data acquisition) by setting preset threshold ranges for each physiological parameter and triggering re-acquisition when a parameter exceeds the limit. This reduces the impact of single measurement errors on subsequent user status assessments and improves the reliability of pulse characteristic parameters. Simultaneously, the re-acquisition mechanism alerts the user with a slight vibration, requiring no complex operation and adapting to user habits. This ensures data quality while avoiding misjudgments of user status due to directly including single abnormal parameters in the assessment, further enhancing the accuracy and rigor of user physiological monitoring and providing more reliable basic data support for the precise calculation of subsequent control parameters.

[0106] In some embodiments, the step of inputting the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control command of the smart lock includes: mapping the blood perfusion index, the heart rate parameter, the heart rate variability parameter, and the pulse wave transit time parameter to obtain a blood perfusion index score, a heart rate score, a heart rate variability score, and a pulse wave transit time score, respectively; determining the weights of the blood perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave transit time parameter based on the initial control parameters preset by the target user; performing a weighted calculation on the blood perfusion index score, the heart rate score, the heart rate variability score, and the pulse wave transit time score based on the weights of the blood perfusion index, the heart rate parameter, the heart rate variability parameter, and the pulse wave transit time parameter to obtain the control parameters of the smart lock; and generating the control command of the smart lock based on the control parameters.

[0107] The step of inputting the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain control parameters includes: mapping the blood perfusion index, the heart rate parameter, the heart rate variability parameter, and the pulse wave transit time parameter to obtain blood perfusion index scores, heart rate scores, heart rate variability scores, and pulse wave transit time scores, respectively. Based on the preset control parameters of the target user, the weights of the blood perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave transit time parameter are determined. Based on the weights of the blood perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave transit time parameter, the blood perfusion index scores, heart rate scores, heart rate variability scores, and pulse wave transit time scores are weighted and calculated to obtain the control parameters.

[0108] In this embodiment, the mapping relationship can be a scoring rule that converts various physiological parameter values ​​into scores of 0-100, based on the normal physiological range of the user group. For the blood perfusion index, 100 points can be awarded based on 1.1 (the median of the normal range), and the score decreases linearly to 0 when deviating from 0.7 (lower limit) or 1.6 (upper limit). The heart rate parameter is awarded based on 75 beats / minute, and the score decreases linearly when deviating from 50 beats / minute or 100 beats / minute. The heart rate variability parameter is awarded based on 60 milliseconds, and the score decreases linearly when deviating from 20 milliseconds or 100 milliseconds. The pulse wave transit time parameter is awarded based on 115 milliseconds, and the score decreases linearly when deviating from 80 milliseconds or 150 milliseconds. Through this mapping relationship, each parameter value can be converted into the corresponding blood perfusion index score, heart rate score, heart rate variability score, and pulse wave transit time score.

[0109] The weights of each parameter can be determined based on the initial control parameters preset by the target user. These preset initial control parameters can be basic physiological data entered during user registration, including age, whether the user suffers from chronic diseases such as hypertension / coronary heart disease, and key indicators from past physical examinations (such as baseline heart rate and vascular elasticity), stored in the smart lock's local encrypted database. For example, for a 65-year-old user without underlying diseases, the weights for blood perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave transit time parameter are 25%. If the user has a history of hypertension, the weight of the heart rate parameter is increased to 35%, and the weights of the other parameters are adjusted accordingly to 20%, 25%, and 20% to strengthen the focus on key cardiovascular indicators.

[0110] The weighted calculation formula is: "Control Parameter = Perfusion Index Score × Perfusion Index Weight + Heart Rate Score × Heart Rate Parameter Weight + Heart Rate Variability Score × Heart Rate Variability Parameter Weight + Pulse Wave Transmission Time Score × Pulse Wave Transmission Time Parameter Weight". The result is a score of 0-100, with higher values ​​indicating better health. For example, a 70-year-old user with a history of hypertension, after mapping, has a Perfusion Index Score of 80, a Heart Rate Score of 75, a Heart Rate Variability Score of 85, and a Pulse Wave Transmission Time Score of 90, with corresponding weights of 20%, 35%, 25%, and 20%. Therefore, the Control Parameter = 80 × 20% + 75 × 35% + 85 × 25% + 90 × 20%, which yields the final Control Parameter.

[0111] This embodiment transforms various physiological parameters into standardized scores through mapping relationships, resolving the fusion problem caused by differences in the dimensions of different parameters. Based on the target user's preset initial control parameter information, the weights of each parameter are dynamically adjusted, enabling targeted strengthening of attention to the user's weak physiological indicators. This makes the control parameters more aligned with individual conditions, avoiding the neglect of differentiated needs by uniform weighting assessments. The final control parameters, obtained through weighted calculation, integrate the comprehensive value of multi-dimensional physiological information while reflecting the targeted nature of personalized assessments, significantly improving the accuracy of user status evaluation. This provides a more scientific basis for subsequent user status information delivery and further adapts to diverse user needs.

[0112] In some embodiments, the warning prompts include Level 1, Level 2, and Level 3 warnings. Controlling the smart lock to issue warning prompts based on the control command includes: when the warning prompt is Level 1, controlling the smart lock to issue an audible alert; when the warning prompt is Level 2, controlling the smart lock to issue both audible and visual alerts and send a warning message to the target device bound to the smart lock; when the warning prompt is Level 3, controlling the smart lock to issue an audible alarm, simultaneously generating a temporary rescue password and sending the temporary rescue password to the target device.

[0113] In this embodiment, the warning prompts can be divided into different risk levels based on the control parameters. Level 1 warnings correspond to control parameters of 70-89 points (minor abnormality, no urgent physiological risk), Level 2 warnings correspond to control parameters of 50-69 points (moderate abnormality, potential physiological risk), and Level 3 warnings correspond to control parameters of less than 50 points (severe abnormality, requiring emergency intervention). The correspondence between warning levels and control parameters is pre-stored in the warning configuration module of the smart door lock, and supports fine-tuning based on the target user's personalized physiological baseline (e.g., for a user with heart disease, a control parameter of 65 points would trigger a Level 2 warning).

[0114] At Level 1, the smart lock provides an audible alert via its built-in speaker, minimizing disruption to family members by not sending notifications to the target device. At Level 2, the smart lock provides both audible and visual alerts, while simultaneously sending a warning message to the target device. The visual alert is indicated by a yellow indicator light on the lock panel (flashing at 1Hz for 1 minute). The warning message sent to the target device includes control parameters, abnormal parameters (e.g., heart rate of 88 beats / minute, exceeding the normal range), and the warning level. This information is transmitted encrypted via the lock's Bluetooth or Wi-Fi module. Upon receiving the message, the target device (e.g., a family member's mobile phone) sends a notification via an app pop-up or SMS, allowing family members to monitor the situation promptly without requiring immediate action. At Level 3, the smart lock triggers an audible alarm and generates a temporary emergency code, which is then sent to the target device. The audible alarm can be a high-decibel sound (different from a regular alert, providing a warning without being harsh), such as a continuous 2-minute sound accompanied by a rapidly flashing red indicator light (2Hz), attracting the attention of those nearby. The temporary rescue password can be a 6-digit random number generated by the door lock's local algorithm. It is valid for 10 minutes (and will automatically expire after that time). After it is generated, it will be pushed to preset target devices such as family members and community elderly care service stations, so that rescuers can quickly enter the home with the password in case of emergency (such as when the user is suddenly unwell and unable to unlock the door). At the same time, the door lock positioning information will be attached when the password is pushed to ensure rescue efficiency.

[0115] This embodiment establishes a layered and progressive early warning response mechanism by linking early warning levels with control parameters and combining the smart lock's audio-visual module and target device push function. This avoids user anxiety caused by excessive early warnings for minor anomalies while enabling rapid linkage of rescue resources in the event of severe anomalies. The entire early warning process requires no additional user operation and relies entirely on the smart lock's hardware functions and linkage capabilities. It adapts to user habits and provides comprehensive home security protection, from daily reminders to emergency rescue, further enhancing the practicality and security of user status monitoring.

[0116] This application also provides a door lock control device for use in smart door locks, such as... Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a door lock control device provided by the present invention, comprising: The acquisition module is used to acquire an image to be identified, which is an image containing the global face of the target user, and is also used to acquire the target user's pulse data to be processed. The parsing module is used to parse the image to be identified and obtain the facial feature data of the global face; The processing module is used to process the facial feature data to obtain the blood perfusion index, and also to process the pulse data to be processed to obtain pulse feature parameters; The instruction generation module is used to input the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control instructions for the smart door lock. The control module is used to control the smart door lock to issue an early warning prompt based on the control command.

[0117] In this embodiment, the door lock control device accurately collects facial and pulse data through an acquisition module, analyzes the facial images through an analysis module, extracts key physiological parameters such as blood perfusion index, heart rate, and heart rate variability through a processing module, and then outputs quantitative control commands through a feature fusion model of the command generation module. Finally, the control module triggers layered early warnings (such as sound prompts, light prompts, and temporary rescue password generation) based on different states, achieving end-to-end protection from physiological data collection, analysis and evaluation to risk warning and emergency rescue. This integrated design not only avoids the complexity of using multiple devices in coordination but also fully utilizes the hardware foundation of smart door locks (such as cameras, handle sensors, and communication modules) to reduce the implementation cost of user status monitoring functions. It retains the core security value of smart door locks while providing users with a convenient, routine, and accurate home monitoring solution, effectively ensuring user home safety.

[0118] This application also provides a smart door lock, including a processor, a camera, and a sensor, wherein the camera and the sensor are both connected to the processor, and the processor is used to execute the door lock control method in any of the above embodiments or implementations.

[0119] The smart door lock provided by this invention, through the processor executing the door lock control method described in any of the above embodiments or implementation methods, deeply integrates the core security function of the smart door lock with the user's home monitoring needs. It eliminates the need for users to purchase, wear, or operate dedicated monitoring equipment. It can complete the non-intrusive data collection and analysis based solely on the daily high-frequency unlocking behavior, which greatly adapts to the physiological characteristics of user groups (especially the elderly) with declining operational abilities and memory loss, and effectively improves the daily usage rate and continuity of user status monitoring.

[0120] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Furthermore, the embodiments of this invention are not directed to any particular programming language.

[0121] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. Similarly, for the sake of brevity and to aid in understanding one or more aspects of the invention, in the description of exemplary embodiments of the invention above, various features of the embodiments are sometimes grouped together in a single embodiment, figure, or description thereof. The claims, which follow the detailed description, are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0122] Those skilled in the art will understand that the modules in the apparatus of the embodiments can be adaptively changed and placed in one or more apparatuses different from those of the embodiments. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components, except that at least some of such features and / or processes or units are mutually exclusive.

[0123] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware.

Claims

1. A door lock control method characterized by, Applied to smart door locks, the method includes: Obtain the image to be identified, which is an image containing the global face of the target user; The image to be identified is analyzed to obtain the facial feature data of the global face; The facial feature data is processed to obtain the blood perfusion index; Obtain the pulse data to be processed from the target user, process the pulse data to be processed, and obtain pulse feature parameters; The blood perfusion index and the pulse characteristic parameters are input into a preset feature fusion model to obtain the control command of the smart door lock; Based on the control command, the smart door lock is controlled to issue an early warning.

2. The method of claim 1, wherein, The image to be identified includes a visible light image and an infrared light image. The process of parsing the image to be identified to obtain the global facial feature data includes: The visible light image and the infrared light image are analyzed to obtain visible light facial data and infrared light facial data; Obtain the first ambient light intensity corresponding to the visible light image and the second ambient light intensity corresponding to the infrared light image; The weights of the visible light facial data are determined based on the first ambient light intensity. The weights of the infrared facial data are determined based on the second ambient light intensity. The visible light facial data and the infrared facial data are weighted and calculated to obtain the facial feature data.

3. The method of claim 1, wherein, The process of processing the facial feature data to obtain the blood perfusion index includes: Based on a preset standard facial database, target facial data that matches the facial feature data is determined; Based on the target facial data, key facial regions in the global face are determined, and features are extracted from these key facial regions to obtain facial blood flow feature parameters. The blood flow characteristic parameters of the face are calculated to obtain the blood perfusion index.

4. The method of claim 3, wherein, The feature extraction of the key facial regions to obtain facial blood flow feature parameters includes: The key facial areas are smoothed to obtain corrected data for the key facial areas, wherein the smoothing process includes smoothing of age spots and smoothing of wrinkles. The corrected data is converted to facial blood flow data. Feature extraction is performed on the facial blood flow data to obtain the facial blood flow feature parameters.

5. The method of claim 1, wherein, The process of acquiring the pulse data to be processed collected by the smart door lock includes: Based on the smart door lock, pulse data from multiple touch points on the target user's palm are obtained; The pulse data from multiple contact points are subjected to reliability screening to obtain the pulse data to be processed.

6. The method of claim 1, wherein, The pulse characteristic parameters include heart rate parameters, heart rate variability parameters, and pulse wave propagation time parameters. After processing the pulse data to obtain the pulse characteristic parameters, the method further includes: If any one of the heart rate parameter, the heart rate variability parameter, and the pulse wave transit time parameter exceeds the corresponding preset threshold range, the pulse data of the target user is reacquired.

7. The method of claim 6, wherein, The step of inputting the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control commands for the smart lock includes: The perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave transit time parameter are mapped to obtain the perfusion index score, heart rate score, heart rate variability score, and pulse wave transit time score, respectively. Based on the initial control parameters preset by the target user, determine the weights of the blood perfusion index, heart rate parameter, heart rate variability parameter, and pulse wave conduction time parameter. The control parameters of the smart door lock are obtained by weighting the blood perfusion index score, the heart rate score, the heart rate variability score, and the pulse wave transit time score according to the weight of the blood perfusion index, the heart rate score, the heart rate variability score, and the pulse wave transit time score. Based on the control parameters, control commands for the smart door lock are generated.

8. The method of claim 1, wherein, The warning prompts include Level 1, Level 2, and Level 3 warnings. The step of controlling the smart lock to issue warning prompts based on the control command includes: When the warning is a Level 1 warning, the smart door lock is controlled to provide an audible alert. When the warning is a level two warning, the smart lock is controlled to provide sound and light alerts and send warning information to the target device bound to the smart lock. When the warning prompt is a Level 3 warning, the smart door lock is controlled to sound an alarm, and a temporary rescue password is generated and sent to the target device.

9. A door lock control device characterized by comprising: Applications in smart door locks include: The acquisition module is used to acquire an image to be identified, which is an image containing the global face of the target user, and is also used to acquire the target user's pulse data to be processed. The parsing module is used to parse the image to be identified and obtain the facial feature data of the global face; The processing module is used to process the facial feature data to obtain the blood perfusion index, and also to process the pulse data to be processed to obtain pulse feature parameters; The instruction generation module is used to input the blood perfusion index and the pulse characteristic parameters into a preset feature fusion model to obtain the control instructions for the smart door lock. The control module is used to control the smart door lock to issue an early warning prompt based on the control command.

10. An intelligent door lock, characterized by The device includes a processor, a camera, and a sensor, wherein the camera and the sensor are both connected to the processor, and the processor is used to execute the door lock control method according to any one of claims 1-8.

Images