AI-based intelligent door lock control method, intelligent access control, storage medium and device

Abstract

The invention relates to the technical field of intelligent door locks, in particular to an AI-based intelligent door lock control method, which comprises the following steps: collecting biological characteristic information of a user in real time, and respectively generating confidence indexes of fingerprint characteristics, face characteristics and voiceprint characteristics based on signal credibility and characteristic integrity of the biological characteristic information; acquiring real-time environment interference parameters of the door lock; determining the priority of the authentication modes based on the confidence index, and obtaining the matching degree evaluation value of each authentication mode to determine a comprehensive risk index of authentication identification; judging whether the authentication is passed based on the comprehensive risk index, and judging whether the environment has an influence on a confidence index based on an environment interference parameter under the condition that the authentication is not passed; and executing a targeted interference elimination and authentication recovery strategy based on the category of the environmental interference parameter under the condition that the environment has an influence on the confidence index. The authentication success rate and safety of the intelligent door lock in a complex real environment are improved.

Description

Technical Field

[0001] This invention relates to the field of smart door lock technology, and in particular to a control method, smart access control, storage medium and device for an AI-based smart door lock. Background Technology

[0002] With the deep integration of IoT technology and artificial intelligence algorithms, the basic security field is undergoing a critical transformation from electronic to intelligent. Traditional electronic door lock systems mainly rely on static credentials such as preset passwords, biometric recognition, or RFID cards for identity verification, presenting a simple binary judgment mode in the verification logic. This architecture is gradually revealing its inherent limitations when dealing with complex real-world security scenarios. The system lacks the ability to dynamically perceive the operating environment and cannot assess whether there are potential threats such as tailgating, mechanical damage, or abnormal gathering at the moment of unlocking, leading to a disconnect between security decisions and actual risks. At the same time, access control mechanisms usually adopt fixed rule configurations, making it difficult to adapt to the flexible needs of modern management scenarios such as temporary authorization, regional collaborative control, and emergency response.

[0003] Chinese Patent Application Publication No. CN118629109A discloses a control method for a smart door lock and a smart door lock. The control method includes: acquiring door opening information of the smart door lock; acquiring the remaining battery power of the smart door lock; predicting a first time period when the remaining battery power is consumed to a first preset level based on a preset power consumption model in the smart door lock; if door opening information is not acquired within the first time period, controlling the smart door lock to activate a super power saving mode after the first time period ends; if door opening information is acquired within the first time period, updating the remaining battery power and re-predicting the first time period.

[0004] However, existing technologies have the following problems: In complex real-world environments, traditional smart door locks rely on static, single, or simple combinations of biometric authentication, which makes them susceptible to environmental interference and fluctuations in feature quality, resulting in insufficient accuracy, security, and adaptability of the overall authentication. Summary of the Invention

[0005] To address this, the present invention provides an AI-based smart lock control method, smart access control system, storage medium, and device to overcome the problems of low recognition reliability and insufficient security caused by environmental interference and static authentication strategies in the prior art.

[0006] To achieve the above objectives, the present invention provides an AI-based smart lock control method, comprising: Step S1: Collect the user's biometric information in real time, including fingerprint image, facial image, facial video stream and ambient background voiceprint. Based on the signal confidence and feature completeness of the biometric information, generate confidence indices for fingerprint features, facial features and voiceprint features respectively. Step S2: Obtain the real-time environmental interference parameters of the door lock. The environmental interference parameters include the acoustic interference index obtained by spectral analysis of the ambient background acoustic pattern and the optical interference index obtained by real-time frame analysis of the facial video stream. Step S3: Determine the priority of the authentication mode based on the confidence index, and obtain the matching degree evaluation value of each authentication mode. Determine the comprehensive risk index of authentication recognition based on the matching degree evaluation value of the first authentication mode and the second authentication mode. The authentication modes include fingerprint authentication, face authentication and voiceprint authentication. Step S4: Determine whether the certification is passed based on the comprehensive risk index. If the certification is not passed, determine whether the environment has an impact on the confidence index based on the environmental interference parameters. Step S5: Under the condition that the environment affects the confidence index, implement targeted interference elimination and authentication recovery strategies based on the category of environmental interference parameters, including improving the matching degree evaluation value of voiceprint features, or automatically activating the infrared fill light of the door lock or adjusting the camera exposure parameters.

[0007] Furthermore, the confidence index of each biometric feature in step S1 is obtained in the following manner, wherein, The confidence index of fingerprint features is determined by the density of effective feature points and the sharpness of ridges in the fingerprint. The confidence index of facial features is determined by the accuracy of key feature point localization and the uniformity of image illumination. The confidence index of voiceprint features is determined by the signal-to-noise ratio and spectral stability index within the audio segment.

[0008] Furthermore, in step S3, the corresponding authentication modes are determined in descending order of the confidence indices of each biometric feature, namely the first authentication mode, the second authentication mode, and the third authentication mode.

[0009] Furthermore, the comprehensive risk index in step S3 is determined by the matching degree evaluation value between the first authentication mode and the second authentication mode, wherein, The matching accuracy evaluation value for fingerprint authentication is the ratio of the number of matching fingerprint points to the total number of preset template points; The matching degree evaluation value of facial features is the ratio of the similarity of facial feature vectors to the similarity threshold; The matching degree evaluation value of voiceprint features is the ratio of voiceprint log-likelihood to the likelihood threshold.

[0010] Furthermore, in step S4, the authentication is determined based on the comprehensive risk index. If the comprehensive risk index is less than the first preset risk index, the authentication is determined to be successful, and an unlocking command is generated to control the door lock to open. If the comprehensive risk index is greater than or equal to the first preset risk index and less than the second preset risk index, the certification is deemed to have failed, and the environmental interference parameters are used to determine whether the environment has an impact on the confidence index. If the comprehensive risk index is greater than or equal to the second preset risk index, the authentication is deemed unsuccessful, the unlocking is rejected, all context snapshots of this high-risk event are recorded, and an alarm message is sent to the preset security contact.

[0011] Furthermore, the impact of the environment on the confidence index is determined based on environmental interference parameters. If the acoustic interference index is greater than or equal to the preset acoustic interference index, it is determined that the environment has an impact on the confidence index, and the priority of the voiceprint authentication mode is forcibly adjusted to the lowest level. At the same time, the likelihood threshold is reduced according to the difference between the acoustic interference index and the preset acoustic interference index. If the acoustic interference index is less than the preset acoustic interference index, it is determined that the environment has no impact on the confidence index, and the system allows the user to temporarily adjust the priority order of each authentication mode through voice commands. If the optical interference index is greater than or equal to the preset optical interference index, it is determined that the environment has an impact on the confidence index, and the door lock's infrared fill light is automatically activated or the camera's exposure parameters are adjusted.

[0012] Furthermore, the reduction in the likelihood threshold is positively correlated with the acoustic interference index difference, which is the difference between the acoustic interference index and the preset acoustic interference index.

[0013] The present invention also provides an intelligent access control system, comprising: The biometric acquisition module is used to collect the user's fingerprint image, facial image, facial video stream, and ambient background voiceprint to generate confidence indices for fingerprint features, facial features, and voiceprint features. An environmental perception module is used to perform real-time spectrum analysis on the background acoustic signature of the environment to generate an acoustic interference index, and to perform real-time frame analysis on the acquired facial video stream to generate an optical interference index. A voice interaction module, which is connected to the biometric data acquisition module, is used to receive and parse the user's voice commands and verify the source of the commands. The data processing module, which is connected to the biometric acquisition module, the environmental perception module and the voice interaction module respectively, is used to determine the priority of the authentication mode based on the confidence index, determine the comprehensive risk index of authentication recognition based on the matching degree evaluation value of the authentication mode, and determine whether the authentication is successful based on the comprehensive risk index. It is also used to dynamically adjust the authentication process according to the legal instructions verified by the voice interaction module. The door lock control module is connected to the environmental perception module and the data processing module respectively. It is used to execute targeted interference elimination and authentication recovery strategies based on the category of environmental interference parameters, under the condition that the environment has an impact on the confidence index based on the environmental interference parameters.

[0014] The present invention also provides a storage medium having a computer program stored thereon, wherein the computer executes the control method of the smart door lock when running the computer program.

[0015] The present invention also provides an intelligent door lock device, including a lock body mechanism, a sensor group, a controller and a communication unit.

[0016] Compared with existing technologies, the advantages of this invention lie in its ability to dynamically sense the impact of environmental interference by acquiring acoustic and optical interference indices in real time. When authentication is hindered, the system automatically executes targeted recovery strategies, such as activating infrared fill lights or adaptively adjusting the voiceprint matching threshold. This design effectively overcomes the interference of adverse environments such as strong light, low light, and background noise on the quality of biometric data acquisition, thereby maintaining a high authentication pass rate under various harsh conditions and enhancing the product's environmental adaptability and user experience.

[0017] Furthermore, this invention dynamically determines the execution order of authentication modes based on real-time collected confidence indices, prioritizing the use of the highest-quality biometric features for identification. Simultaneously, it calculates a comprehensive risk index by combining the matching evaluation values ​​of the first and second authentication modes, and adopts differentiated response measures based on multi-level risk thresholds. This mechanism fully utilizes the complementary advantages of multimodal information while effectively resisting the risk of imitation or failure of a single modality, forming an adaptive, three-dimensional security protection system.

[0018] Furthermore, based on a comprehensive risk index-based hierarchical response design, this invention constructs a gradient and intelligent security decision-making mechanism, achieving an effective balance between security and convenience. When the risk is low, passage is quickly granted to ensure a good user experience. When the risk is in the intermediate range, the system does not directly reject access but initiates environmental interference analysis to provide a basis for subsequent targeted recovery strategies, demonstrating dynamic adaptability. When the risk reaches a high-risk threshold, an active defense process including on-site snapshot recording and real-time alarms is triggered, not only blocking potential risks but also providing a complete data chain for post-event traceability and system optimization. This significantly improves the overall ability of smart locks to cope with complex scenarios and potential threats. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the AI-based smart door lock control method according to an embodiment of the present invention; Figure 2This is a flowchart illustrating how the certification process is determined based on a comprehensive risk index, according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the module connection of the intelligent access control system according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the frame structure of the smart door lock device according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0021] It should be noted that the data in this embodiment are all derived from a comprehensive analysis and evaluation of historical test data and corresponding historical test results from the three months prior to this test. Those skilled in the art will understand that the determination of the above-mentioned parameters for any single item in this invention can be achieved by selecting the value with the highest percentage based on the data distribution as the preset standard parameter, using weighted summation to obtain the value as the preset standard parameter, substituting each historical data point into a specific formula and using the value obtained from that formula as the preset standard parameter, or other selection methods, as long as the invention can clearly define different specific situations in the single-item judgment process through the obtained values.

[0022] Please see Figure 1 The diagram shows a flowchart of an AI-based smart lock control method according to an embodiment of the present invention. The AI-based smart lock control method according to an embodiment of the present invention includes: Step S1: Collect the user's biometric information in real time, including fingerprint image, facial image, facial video stream and ambient background voiceprint. Based on the signal confidence and feature completeness of the biometric information, generate confidence indices for fingerprint features, facial features and voiceprint features respectively. Step S2: Obtain the real-time environmental interference parameters of the door lock. The environmental interference parameters include the acoustic interference index obtained by spectral analysis of the ambient background acoustic pattern and the optical interference index obtained by real-time frame analysis of the facial video stream. Step S3: Determine the priority of the authentication mode based on the confidence index, and obtain the matching degree evaluation value of each authentication mode. Determine the comprehensive risk index of authentication recognition based on the matching degree evaluation value of the first authentication mode and the second authentication mode. The authentication modes include fingerprint authentication, face authentication and voiceprint authentication. Step S4: Determine whether the certification is passed based on the comprehensive risk index. If the certification is not passed, determine whether the environment has an impact on the confidence index based on the environmental interference parameters. Step S5: Under the condition that the environment affects the confidence index, implement targeted interference elimination and authentication recovery strategies based on the category of environmental interference parameters, including improving the matching degree evaluation value of voiceprint features, or automatically activating the infrared fill light of the door lock or adjusting the camera exposure parameters.

[0023] Specifically, the confidence index of each biometric feature in step S1 is obtained in the following manner, wherein, The confidence index of fingerprint features is determined by the density of effective feature points and the sharpness of ridges in the fingerprint. The confidence index of facial features is determined by the accuracy of key feature point localization and the uniformity of image illumination. The confidence index of voiceprint features is determined by the signal-to-noise ratio and spectral stability index within the audio segment.

[0024] Specifically, the confidence index of fingerprint features = density weighting coefficient × effective feature point density / preset feature point density + sharpness weighting coefficient × ridge sharpness / preset sharpness, where the density weighting coefficient is 0.4 and the preset feature point density is 60 feature points / cm. 2 The resolution weighting coefficient is set to 0.6, and the preset resolution is set to 0.85.

[0025] Specifically, effective feature point density refers to the number of stable minutiae within a unit area of ​​a fingerprint image that can be used for identity recognition. These stable minutiae are mainly ridge endpoints and bifurcation points. This indicator directly reflects the information richness and uniqueness of fingerprint features and is a key quantitative basis for evaluating fingerprint image quality and authentication value. The acquisition process is as follows: First, the preprocessed fingerprint image is divided into effective regions, excluding invalid regions with incomplete boundaries, blurred edges, or severe contamination. Then, within these effective regions, a minutiae extraction algorithm is used to locate and count the total number of all ridge endpoints and bifurcation points that meet the quality requirements. Simultaneously, based on the physical resolution of the fingerprint sensor, the pixel area occupied by this effective region is converted to a standard physical area. Finally, the effective feature point density is calculated by dividing the total number of counted feature points by the physical area of ​​the effective region.

[0026] Specifically, ridge sharpness refers to a comprehensive quantitative index of the sharpness, continuity, and distinguishability of ridge structures in a fingerprint image. It directly determines the accuracy and reliability of feature extraction algorithms and is a core basis for evaluating the resolvable quality of a fingerprint image. The acquisition process includes: dividing the preprocessed fingerprint image into ridge and valley regions; using image segmentation techniques to divide pixels into ridge and valley representations; calculating the average gray value of all ridge pixels and the average gray value of all valley pixels; dividing the absolute value of the difference between the average gray values ​​of the ridges and valleys by their sum to obtain a normalized contrast value between zero and one. The closer this value is to one, the sharper the ridge-valley contrast and the higher the image sharpness; conversely, a value closer to one indicates a blurry image or insufficient contrast.

[0027] Specifically, the confidence index of facial features = positioning accuracy weight coefficient × key feature point positioning accuracy / preset positioning accuracy + illumination uniformity weight coefficient × image illumination uniformity / preset uniformity, where the positioning accuracy weight coefficient is 0.6, the illumination uniformity weight coefficient is 0.4, the preset positioning accuracy is 0.95, and the preset uniformity is 20.

[0028] Specifically, the process of obtaining the accuracy of key feature point localization includes: first, locating a predetermined number of key feature points such as the corners of the eyes, the tip of the nose, and the corners of the mouth in the facial image or video stream; then, quantifying the reliability of the localization by the confidence score output by the model or the stability of the feature point positions in consecutive frames. The higher the accuracy, the greater the contribution to the confidence score.

[0029] Specifically, the process of obtaining image illumination uniformity includes: dividing the detected face region into multiple sub-regions and calculating the average brightness value of each sub-region; and quantifying the uniformity of illumination distribution by statistically analyzing the standard deviation or maximum difference of the brightness values ​​of these sub-regions. The smaller this value, the more uniform the illumination, and the higher the corresponding score.

[0030] Specifically, the confidence index of voiceprint features = signal-to-noise ratio weighting coefficient × signal-to-noise ratio within the speech segment / preset signal-to-noise ratio + spectral stability weighting coefficient × spectral stability index / preset stability index, where the signal-to-noise ratio weighting coefficient is 0.5, the spectral stability weighting coefficient is 0.5, the preset signal-to-noise ratio is 20 dB, and the preset stability index is 0.9.

[0031] Specifically, the process of obtaining the signal-to-noise ratio (SNR) within a speech segment includes: performing endpoint detection on the acquired audio signal to divide the speech activity segment into a silent segment; extracting the spectral characteristics of the background noise from the silent segment and estimating its power spectral density; estimating the power of the clean speech signal within the speech activity segment based on spectral subtraction or a statistical model; calculating the ratio of the speech signal power to the background noise power and converting it to a decibel value to obtain the SNR quantization result within the speech segment.

[0032] Specifically, the process of obtaining the spectral stability index includes: dividing the speech activity segment into frames and performing pre-emphasis and windowing processing; extracting spectral feature vectors frame by frame; calculating the similarity of feature vectors between consecutive frames; and calculating the average or variance of the similarity of all consecutive frames to obtain the quantitative result of the spectral stability index. This index reflects the temporal consistency of acoustic features during the pronunciation process.

[0033] Specifically, in step S2, the acoustic interference index employs a noise classification model based on Mel frequency cepstral coefficients or a residual noise estimation method based on spectral subtraction to perform real-time analysis of the environmental background sound signature. The specific steps include: converting the sound signature signal into a frequency domain representation using short-time Fourier transform; calculating the equivalent sound pressure level of a preset noise characteristic frequency band (such as the 200Hz-4kHz human voice interference band), and quantifying the sound pressure level fluctuation using a time-series fluctuation detection algorithm; and synthesizing the equivalent sound pressure level and its fluctuation coefficient into an acoustic interference index in decibels (dB) using a linear weighted fusion model, the value of which is positively correlated with noise intensity and instability.

[0034] The Optical Interference Index (OI) is based on real-time frame analysis of facial video streams using an image quality assessment model and an illumination perception algorithm. The specific steps include: locating the facial region in the video frame using a face detection network; quantifying image detail loss, blurriness, and illumination unevenness within this region using local binary mode contrast analysis, gradient magnitude statistics, and histogram dispersion calculation; estimating the ambient illuminance value of the face in lux by combining camera exposure parameters and the image brightness histogram; and fusion of image quality and illuminance features using a multi-feature regression model to output the OI in lux, whose value is positively correlated with insufficient illumination and image degradation.

[0035] Specifically, in step S3, the corresponding authentication modes are determined in descending order of the confidence index of each biometric feature, namely the first authentication mode, the second authentication mode, and the third authentication mode.

[0036] Specifically, the comprehensive risk index in step S3 is determined by the matching degree evaluation value between the first authentication mode and the second authentication mode, wherein, The matching accuracy evaluation value for fingerprint authentication is the ratio of the number of matching fingerprint points to the total number of preset template points; The matching degree evaluation value of facial features is the ratio of the similarity of facial feature vectors to the similarity threshold; The matching degree evaluation value of voiceprint features is the ratio of voiceprint log-likelihood to the likelihood threshold.

[0037] Specifically, the number of fingerprint matching points refers to the total number of feature point pairs that successfully establish a corresponding relationship during the fingerprint feature comparison process by geometrically matching and calculating the feature point set extracted from the real-time acquired fingerprint image with the pre-registered fingerprint template feature point set; the facial feature vector similarity is a measure of the spatial distance or angular similarity between the real-time facial feature vector extracted by the facial recognition model and the pre-stored registered template feature vector, usually calculated using cosine similarity or the normalized reciprocal of Euclidean distance; the voiceprint log-likelihood is the log-likelihood ratio calculated based on the degree of matching between the input real-time speech feature sequence and the target speaker registration model using an acoustic model (such as Gaussian mixture model, deep neural network). This value reflects the logarithmic advantage of the probability that the current speech feature sequence belongs to the target speaker model relative to the probability that it belongs to the general background model. The higher the value, the higher the confidence of the voiceprint matching.

[0038] In this embodiment of the invention, the total number of preset template points is the number of valid feature points contained in the fingerprint template entered by the user during the registration stage, and its value is 60 feature points; the similarity threshold is determined by statistical analysis based on a large-scale face recognition test set, and its value is 0.85; the likelihood threshold is determined by the balance point between the false acceptance rate and the false rejection rate of the voiceprint verification system, and its value is 5.0.

[0039] Specifically, the comprehensive risk index = first weight coefficient × first matching degree evaluation threshold / matching degree evaluation value of the first authentication mode + second weight coefficient × second matching degree evaluation threshold / matching degree evaluation value of the first authentication mode, where the first weight coefficient is 0.6, the first matching degree evaluation threshold is 0.8, the second weight coefficient is 0.4, and the second matching degree evaluation threshold is 0.7.

[0040] Understandably, dynamically determining the execution order of authentication modes based on the confidence index reflects the system's intelligent perception and adaptive scheduling capabilities regarding the quality of multimodal biometrics. Using the authentication mode with the highest confidence index as the first authentication mode prioritizes the use of the highest-quality biometrics collected under the current environmental conditions for identity verification, significantly improving the success rate and efficiency of the initial authentication attempt. Using the modes with the second and lowest confidence indices as the second and third authentication modes, respectively, forms a progressive backup verification channel. This design not only optimizes resource allocation in the authentication process but, more importantly, allows the system to immediately invoke the second authentication mode (with slightly lower quality) for auxiliary decision-making or fusion judgment when the first authentication mode fails due to minor interference or accidental factors. This effectively avoids unnecessary authentication failures caused by temporary quality fluctuations in a single modality while ensuring security, greatly enhancing the system's robustness and user experience.

[0041] The comprehensive risk index is constructed based on the matching evaluation values ​​of the first and second authentication modes, reflecting a deep consideration of the balance between authentication security levels and efficiency. The first authentication mode represents the most reliable biometrics currently available, and its matching results form the backbone of the risk assessment. The second authentication mode, as a secondary but still effective auxiliary verification method, provides important cross-validation information through its matching results. By assigning first and second weighting coefficients to each mode respectively, and performing normalized weighted fusion based on preset first and second matching evaluation thresholds, the system can quantitatively characterize the overall risk level of the current authentication attempt. This decision-making mechanism, which only fuses the first two high-quality modalities, effectively focuses on high-confidence information, avoiding noise interference that may be introduced by the low-quality third modality, thus improving the accuracy of risk assessment. Furthermore, by integrating the matching degrees of the two independent modalities, a dual verification barrier is constructed, significantly enhancing the system's ability to prevent spoofing attacks. The resulting risk index provides a precise and reliable quantitative basis for subsequent tiered response strategies.

[0042] Please see Figure 2 As shown, this is a flowchart illustrating how the certification process is determined based on a comprehensive risk index, according to an embodiment of the present invention. Specifically, the certification process is determined based on a comprehensive risk index, wherein... If the overall risk index is less than the first preset risk index, the authentication is deemed successful, and an unlocking command is generated to control the door lock to open. If the comprehensive risk index is greater than or equal to the first preset risk index and less than the second preset risk index, the certification is deemed to have failed, and the environmental interference parameters are used to determine whether the environment has an impact on the confidence index. If the comprehensive risk index is greater than or equal to the second preset risk index, the authentication is deemed unsuccessful, the unlocking is rejected, all context snapshots of this high-risk event are recorded, and an alarm message is sent to the preset security contact.

[0043] In this embodiment of the invention, the first preset risk index is 0.75 and the second preset risk index is 0.9. However, the above values ​​are not limited to these values, and those skilled in the art can adjust the above values ​​according to actual needs.

[0044] It is understood that the embodiments of the present invention construct a three-level intelligent response mechanism. Based on the first preset risk index, a safe passage range is defined, allowing reasonable fluctuations to improve the experience while ensuring basic security. Based on the second preset risk index, a high-risk warning line is defined to trigger active defense to deal with potential attacks. Environmental analysis and adaptive recovery are initiated in the two threshold ranges to realize the transformation from passive rejection to intelligent error correction, achieving the optimal balance between security and availability.

[0045] Specifically, the impact of the environment on the confidence index is determined based on environmental interference parameters. If the acoustic interference index is greater than or equal to the preset acoustic interference index, it is determined that the environment has an impact on the confidence index, and the priority of the voiceprint authentication mode is forcibly adjusted to the lowest level. At the same time, the likelihood threshold is reduced according to the difference between the acoustic interference index and the preset acoustic interference index. If the acoustic interference index is less than the preset acoustic interference index, it is determined that the environment has no impact on the confidence index, and the system allows the user to temporarily adjust the priority order of each authentication mode through voice commands. If the optical interference index is greater than or equal to the preset optical interference index, it is determined that the environment has an impact on the confidence index, and the door lock's infrared fill light is automatically activated or the camera's exposure parameters are adjusted.

[0046] Understandably, when the acoustic interference index is less than the preset acoustic interference index, it indicates that the current acoustic environment is good, and the system determines that environmental noise has no impact on authentication. At this point, while maintaining its autonomous decision-making framework, the system opens a high-priority interactive channel to the user, allowing the user to actively intervene and temporarily optimize the priority order of the authentication process via voice commands. This dynamically grants some process control to the user without compromising the overall system security, enabling the system to flexibly adapt to the user's immediate intentions and special needs in specific scenarios, thereby significantly improving the smoothness and flexibility of the authentication process.

[0047] Specifically, the decrease in the likelihood threshold is positively correlated with the acoustic interference index difference, which is the difference between the acoustic interference index and the preset acoustic interference index.

[0048] In this embodiment of the invention, the preset acoustic interference index is 40 dB, a threshold set based on the background noise level in a typical quiet indoor environment. When the environmental acoustic interference index reaches or exceeds this value, it indicates that environmental noise has significantly interfered with the clear acquisition of voiceprint features, and the system will determine that the environment has a negative impact on the confidence index of voiceprint features. The preset optical interference index is 60 lux, a threshold set based on the minimum illumination requirements for face recognition. When the environmental optical interference index reaches or exceeds this value, it indicates that the ambient lighting conditions are insufficient to support high-quality face image acquisition, and the system will determine that the environment has a negative impact on the confidence index of face features.

[0049] Please see Figure 3 The diagram shown is a schematic diagram of the module connection of the intelligent access control system according to an embodiment of the present invention; the intelligent access control system according to an embodiment of the present invention includes: The biometric acquisition module is used to collect the user's fingerprint image, facial image, facial video stream, and ambient background voiceprint to generate confidence indices for fingerprint features, facial features, and voiceprint features. An environmental perception module is used to perform real-time spectrum analysis on the background acoustic signature of the environment to generate an acoustic interference index, and to perform real-time frame analysis on the acquired facial video stream to generate an optical interference index. A voice interaction module, which is connected to the biometric data acquisition module, is used to receive and parse the user's voice commands and verify the source of the commands. The data processing module, which is connected to the biometric acquisition module, the environmental perception module and the voice interaction module respectively, is used to determine the priority of the authentication mode based on the confidence index, determine the comprehensive risk index of authentication recognition based on the matching degree evaluation value of the authentication mode, and determine whether the authentication is successful based on the comprehensive risk index. It is also used to dynamically adjust the authentication process according to the legal instructions verified by the voice interaction module. The door lock control module is connected to the environmental perception module and the data processing module respectively. It is used to execute targeted interference elimination and authentication recovery strategies based on the category of environmental interference parameters, under the condition that the environment has an impact on the confidence index based on the environmental interference parameters.

[0050] Specifically, the biometric acquisition module includes a capacitive or optical fingerprint sensor for acquiring the user's fingerprint image and generating fingerprint feature data and its confidence index; a high-definition camera for acquiring the user's facial image or facial video stream and generating facial feature data and its confidence index; and a microphone array for acquiring the user's voiceprint signal when reciting the authentication password and the ambient background voiceprint, and generating voiceprint feature data and its confidence index.

[0051] The environmental perception module is electrically connected to the microphone array for real-time spectral analysis of the ambient background acoustic signature and calculation of the acoustic interference index, which reflects the intensity and stability of ambient noise. Simultaneously, it is electrically connected to the high-definition camera for real-time frame analysis of the facial video stream and calculation of the optical interference index, which reflects the ambient lighting conditions and imaging quality, by evaluating image illumination, sharpness, and contrast.

[0052] The voice interaction module receives the raw audio signal collected by the microphone array in the biometric acquisition module. While performing voice content recognition on the audio signal, it extracts the speaker's voiceprint features from the voice stream in real time. The voiceprint features are then quickly compared with the voiceprint templates of authorized users pre-stored in a secure area. Based on the comparison result where the voiceprint similarity is greater than a preset similarity, the command is determined to be legitimate.

[0053] In this embodiment of the invention, the preset similarity value is 95%, but this value is not limited to this. Those skilled in the art can adjust this value according to actual needs.

[0054] Specifically, embodiments of the present invention also provide a storage medium, such as an embedded flash memory, a secure digital card, or a read-only memory. The storage medium stores a computer program. When the controller in the smart lock device reads and executes the computer program, it controls the sensor group, the lock body mechanism, and various functional modules to implement the smart lock control method described above.

[0055] Please see Figure 4 As shown, it is a schematic diagram of the frame structure of the smart door lock device according to an embodiment of the present invention; the present invention provides a smart door lock device, including a lock body mechanism, a sensor group, a controller and a communication unit.

[0056] In this embodiment of the invention, the lock body mechanism is preferably an electromagnetic lock body or a motor-driven lock tongue, used to perform physical unlocking after authentication.

[0057] In this embodiment of the invention, the sensor group is integrated into the door lock panel or its interior, and includes: The fingerprint acquisition unit uses the aforementioned capacitive or optical fingerprint sensor; The image acquisition unit uses the aforementioned high-definition camera and may be equipped with an infrared fill light; The audio acquisition unit employs the aforementioned microphone array; In this embodiment of the invention, the controller is an embedded microprocessor or system-on-a-chip installed inside the door lock, which runs the computer program to specifically implement all the logical functions of the data processing module, the environmental perception module, and the door lock control module.

[0058] In this embodiment of the invention, the communication unit includes at least one of a Wi-Fi module, a Bluetooth module, and a Zigbee module, and is used to interact with a cloud server or a user's mobile terminal.

[0059] Optionally, the device may further include a power supply unit, such as a lithium battery pack, and a human-machine interface unit, such as status indicator lights, a touch screen, or buttons, for providing power support and local interaction functions.

[0060] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity.

[0061] This invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A control method for an AI-based smart door lock, characterized in that, include: Step S1: Collect the user's biometric information in real time, including fingerprint image, facial image, facial video stream and ambient background voiceprint. Based on the signal confidence and feature completeness of the biometric information, generate confidence indices for fingerprint features, facial features and voiceprint features respectively. Step S2: Obtain the real-time environmental interference parameters of the door lock. The environmental interference parameters include the acoustic interference index obtained by spectral analysis of the ambient background acoustic pattern and the optical interference index obtained by real-time frame analysis of the facial video stream. Step S3: Determine the priority of the authentication mode based on the confidence index, and obtain the matching degree evaluation value of each authentication mode. Determine the comprehensive risk index of authentication recognition based on the matching degree evaluation value of the first authentication mode and the second authentication mode. The authentication modes include fingerprint authentication, face authentication and voiceprint authentication. Step S4: Determine whether the certification is passed based on the comprehensive risk index. If the certification is not passed, determine whether the environment has an impact on the confidence index based on the environmental interference parameters. Step S5: Under the condition that the environment affects the confidence index, implement targeted interference elimination and authentication recovery strategies based on the category of environmental interference parameters, including improving the matching degree evaluation value of voiceprint features, or automatically activating the infrared fill light of the door lock or adjusting the camera exposure parameters.

2. The control method for an AI-based smart door lock according to claim 1, characterized in that, The confidence index of each biometric feature in step S1 is obtained in the following way. in, The confidence index of fingerprint features is determined by the density of effective feature points and the sharpness of ridges in the fingerprint. The confidence index of facial features is determined by the accuracy of key feature point localization and the uniformity of image illumination. The confidence index of voiceprint features is determined by the signal-to-noise ratio and spectral stability index within the audio segment.

3. The control method for an AI-based smart door lock according to claim 2, characterized in that, In step S3, the corresponding authentication modes are determined by sorting the confidence indices of each biometric feature from largest to smallest, namely the first authentication mode, the second authentication mode, and the third authentication mode.

4. The control method for an AI-based smart door lock according to claim 3, characterized in that, The comprehensive risk index in step S3 is determined by the matching degree evaluation value between the first certification mode and the second certification mode, wherein, The matching accuracy evaluation value for fingerprint authentication is the ratio of the number of matching fingerprint points to the total number of preset template points; The matching degree evaluation value of facial features is the ratio of the similarity of facial feature vectors to the similarity threshold; The matching degree evaluation value of voiceprint features is the ratio of voiceprint log-likelihood to the likelihood threshold.

5. The control method for an AI-based smart door lock according to claim 4, characterized in that, In step S4, the certification is determined based on a comprehensive risk index. If the overall risk index is less than the first preset risk index, the authentication is deemed successful, and an unlocking command is generated to control the door lock to open. If the comprehensive risk index is greater than or equal to the first preset risk index and less than the second preset risk index, the certification is deemed to have failed, and the environmental interference parameters are used to determine whether the environment has an impact on the confidence index. If the comprehensive risk index is greater than or equal to the second preset risk index, the authentication is deemed unsuccessful, the unlocking is rejected, all context snapshots of this high-risk event are recorded, and an alarm message is sent to the preset security contact.

6. The control method for an AI-based smart door lock according to claim 5, characterized in that, The determination of whether the environment affects the confidence index is based on environmental interference parameters. If the acoustic interference index is greater than or equal to the preset acoustic interference index, it is determined that the environment has an impact on the confidence index, and the priority of the voiceprint authentication mode is forcibly adjusted to the lowest level. At the same time, the likelihood threshold is reduced according to the difference between the acoustic interference index and the preset acoustic interference index. If the acoustic interference index is less than the preset acoustic interference index, it is determined that the environment has no impact on the confidence index, and the system allows the user to temporarily adjust the priority order of each authentication mode through voice commands. If the optical interference index is greater than or equal to the preset optical interference index, it is determined that the environment has an impact on the confidence index, and the door lock's infrared fill light is automatically activated or the camera's exposure parameters are adjusted.

7. The control method for an AI-based smart door lock according to claim 6, characterized in that, The decrease in the likelihood threshold is positively correlated with the acoustic interference index difference, which is the difference between the acoustic interference index and the preset acoustic interference index.

8. An intelligent access control system, applied to the control method of the intelligent door lock according to any one of claims 1-7, characterized in that, include: The biometric acquisition module is used to collect the user's fingerprint image, facial image, facial video stream, and ambient background voiceprint to generate confidence indices for fingerprint features, facial features, and voiceprint features. An environmental perception module is used to perform real-time spectrum analysis on the background acoustic signature to generate an acoustic interference index, and to perform real-time frame analysis on the acquired facial video stream to generate an optical interference index. A voice interaction module, which is connected to the biometric data acquisition module, is used to receive and parse the user's voice commands and verify the source of the commands. The data processing module, which is connected to the biometric acquisition module, the environmental perception module and the voice interaction module respectively, is used to determine the priority of the authentication mode based on the confidence index, determine the comprehensive risk index of authentication recognition based on the matching degree evaluation value of the authentication mode, and determine whether the authentication is successful based on the comprehensive risk index. It is also used to dynamically adjust the authentication process according to the legal instructions verified by the voice interaction module. The door lock control module, which is connected to the environmental perception module and the data processing module respectively, is used to execute targeted interference elimination and authentication recovery strategies based on the category of environmental interference parameters, under the condition that the environment has an impact on the confidence index based on the environmental interference parameters.

9. A storage medium, characterized in that, It stores a computer program, and when the computer runs the computer program, it executes the control method of the smart door lock according to any one of claims 1-7.

10. A smart door lock device, applied to the control method of the smart door lock according to any one of claims 1-7, characterized in that, It includes a lock mechanism, sensor group, controller and communication unit.

Images