Anomaly identification method of intelligent lock, intelligent lock and storage medium

Abstract

The application provides an abnormality identification method of a smart lock, a smart lock and a storage medium, the smart lock comprising a handle, a sensor group arranged on the handle and a master control module, the sensor group comprising at least a touch sensor, a pressure sensor and an inertial measurement unit, when the touch sensor detects that a user contacts the handle, the master control module starts the sensor group to synchronously collect multi-source time sequence data, the multi-source time sequence data comprising at least pressure distribution time sequence data collected by the pressure sensor in the process that the user holds the handle and motion time sequence data collected by the inertial measurement unit in the process that the user rotates the handle; the master control module extracts features from the multi-source time sequence data to obtain multi-dimensional behavior features; the master control module inputs the multi-dimensional behavior features into a pre-trained abnormality identification model to obtain a coercion probability value of the user in a coercion state, and determines an abnormality identification result of the user according to the coercion probability value, so that the identification trigger rate and reliability in a real coercion scenario can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of smart lock technology, and in particular to a method for identifying anomalies in a smart lock, a smart lock, and a storage medium. Background Technology

[0002] As smart locks gradually become the core entry point for home security, the need for personal safety protection in extreme scenarios such as being coerced into unlocking is becoming increasingly prominent. However, existing coercion alarm solutions mostly rely on user-initiated triggers, such as coercive fingerprints, coercive passwords, or emergency buttons on wearable devices. These require users to deliberately perform specific operations under high stress, which carries the risk of alarm failure due to forgetting due to stress, the coerced person recognizing the error, or restricting the use of the device. Furthermore, anti-tamper recognition can only identify external violent damage and cannot cover internal threat scenarios where the user is coerced into unlocking the door normally, resulting in insufficient reliability of security protection under coercion.

[0003] Among related technologies, while duress fingerprint and duress password schemes can trigger backend recognition, the triggering method is singular and easily bypassed by the duress; wearable device solutions rely on additional wearing, limiting their widespread adoption and potential for usage restrictions; anti-pry recognition only targets forced dismantling of the lock body and cannot identify situations where the user is duressed to unlock the lock normally. All of the above solutions struggle to achieve a seamless alarm that automatically detects duress without user intervention and is highly concealed, resulting in low trigger rates and poor reliability in real-world duress scenarios. Summary of the Invention

[0004] The main objective of this disclosure is to propose an anomaly identification method for smart locks, a smart lock, and a storage medium, which can improve the identification trigger rate and reliability in real-world coercive scenarios.

[0005] To achieve the above objectives, a first aspect of this disclosure provides an anomaly detection method for a smart lock, applied to a smart lock, the smart lock including a handle, a sensor group disposed on the handle, and a main control module, the sensor group including at least a touch sensor, a pressure sensor, and an inertial measurement unit, the anomaly detection method including: When the touch sensor detects that the user touches the handle, the main control module activates the sensor group to synchronously collect multi-source time-series data. The multi-source time-series data includes at least the pressure distribution time-series data collected by the pressure sensor during the user's grip on the handle, and the motion time-series data collected by the inertial measurement unit during the user's rotation of the handle. The main control module extracts features from the multi-source time-series data to obtain multi-dimensional behavioral features; The main control module inputs the multi-dimensional behavioral features into a pre-trained anomaly recognition model to obtain the coercion probability value of the user being in a coercive state, and determines the anomaly recognition result of the user based on the coercion probability value.

[0006] In some embodiments, the multi-source time-series data further includes physiological signal time-series data, which is collected and sent to the smart lock by a wearable device that is communicatively connected to the smart lock, or collected by a physiological signal sensor located on the handle during the user's grip of the handle.

[0007] In some embodiments, the main control module performs feature extraction on the multi-source time-series data to obtain multi-dimensional behavioral features, including: The pressure distribution time series data is subjected to feature extraction to obtain pressure distribution features, which include at least one of pressure distribution entropy, contact area change rate and pressure center of gravity movement trajectory. Feature extraction is performed on the motion time series data to obtain motion features, which include at least one of the following: high-frequency micro-tremor energy ratio, motion smoothness, and rotational angular velocity variation features. Multidimensional behavioral features are obtained based on the pressure distribution characteristics and the motion characteristics.

[0008] In some embodiments, the anomaly recognition model is a pre-trained temporal deep learning model; the main control module inputs the multi-dimensional behavioral features into the pre-trained anomaly recognition model, which is then executed locally on the smart lock.

[0009] In some embodiments, determining the user's anomaly identification result based on the coercion probability value includes: If the coercion probability value is greater than or equal to a preset first threshold, the abnormal identification result is determined to be that the user is in a coercive state; If the coercion probability value is less than a preset second threshold, the abnormal identification result is determined to be that the user is in a normal state, wherein the second threshold is less than the first threshold; If the coercion probability value is greater than or equal to the second threshold and less than the first threshold, the multi-dimensional behavioral features are compared with the user's personalized behavioral baseline. If the multi-dimensional behavioral features deviate from the personalized behavioral baseline by more than a preset amount, the abnormal identification result is determined to be that the user is in a coercive state.

[0010] In some embodiments, the anomaly identification method further includes: If the anomaly identification result indicates that the user is in a normal state, the user's personalized behavior baseline is updated based on the multi-dimensional behavioral features collected in this study.

[0011] In some embodiments, the smart lock further includes a communication module; the anomaly detection method further includes: If the abnormal identification result indicates that the user is in a coerced state, the main control module controls the smart lock to open, controls the output state of the smart lock to be consistent with that when the user is in a normal state, and silently sends the corresponding identification information to the preset identification object through the communication module.

[0012] To achieve the above objectives, a second aspect of the present disclosure provides a smart lock, including a handle, a sensor group disposed on the handle, and a main control module, wherein the sensor group includes at least a touch sensor, a pressure sensor, and an inertial measurement unit; The main control module is used to activate the sensor group to synchronously collect multi-source time-series data when the touch sensor detects that the user is touching the handle. The multi-source time-series data includes at least the pressure distribution time-series data collected by the pressure sensor during the user's grip on the handle, and the motion time-series data collected by the inertial measurement unit during the user's rotation of the handle. The main control module is also used to extract features from the multi-source time-series data to obtain multi-dimensional behavioral features; input the multi-dimensional behavioral features into a pre-trained anomaly recognition model to obtain the coercion probability value of the user being in a coercive state, and determine the user's anomaly recognition result based on the coercion probability value.

[0013] In some embodiments, the pressure sensor is an array of pressure sensors covering the grip area of ​​the handle; the touch sensor is distributed on the surface of the handle; and the inertial measurement unit is built into the handle.

[0014] To achieve the above objectives, a third aspect of this disclosure provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the anomaly identification method for the smart lock described in the first aspect embodiment.

[0015] To achieve the above objectives, a fourth aspect of this disclosure provides a storage medium, which is a computer-readable storage medium storing a computer program that, when executed by a processor, implements the anomaly identification method for the smart lock described in the first aspect embodiment.

[0016] The beneficial effects of the embodiments disclosed herein include: By incorporating a sensor array consisting of a touch sensor, a pressure sensor, and an inertial measurement unit into the handle, synchronous data acquisition automatically begins when the touch sensor detects user contact with the handle. During the normal unlocking process of the user naturally gripping and turning the handle, multi-source behavioral data, including pressure distribution time-series data and motion time-series data, are acquired. This process requires no conscious user intervention and does not rely on additional wearable devices, fundamentally avoiding the various failure risks of active triggering schemes and the limitations of wearable device adoption. Subsequently, feature extraction is performed on the multi-source time-series data to obtain multi-dimensional behavioral features, utilizing the user's muscle... The physiological and behavioral patterns of muscle tension leading to abnormal grip force distribution and stiff, trembling movements causing deviations from normal movement trajectories can be accurately identified by a pre-trained anomaly recognition model that outputs a coercion probability value and determines the anomaly recognition result. This can accurately identify internal threat scenarios where the user is coerced but appears to be unlocking normally. At the same time, the multi-dimensional features fused from multiple data sources make it difficult for the coercer to imitate all behavioral dimensions simultaneously, making it difficult to be forcibly bypassed. Furthermore, the entire recognition process is embedded in the normal unlocking process, is imperceptible to the user, and has strong concealment, thus achieving a seamless coercion alarm without the need for active user operation, improving the recognition trigger rate and reliability in real coercion scenarios. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the hardware structure module of the smart lock provided in the embodiments of this disclosure; Figure 2 This is a flowchart illustrating the anomaly detection method for smart locks provided in this embodiment of the disclosure; Figure 3 yes Figure 2 A flowchart further includes step S102; Figure 4 yes Figure 2 The flowchart of the process of determining the abnormal identification result of the user based on the coercion probability value in step S103 is further illustrated. Figure 5 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this disclosure. Detailed Implementation

[0018] The accompanying drawings in the embodiments clearly and completely describe the technical solutions in the embodiments of this disclosure. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0019] It is understood that in the specific embodiments of this disclosure, user behavior data, physiological signal data and related user data are involved. When the above embodiments of this disclosure are applied to specific products or technologies, user permission or consent can be obtained, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards.

[0020] Furthermore, when this embodiment of the disclosure needs to retrieve user behavior data, physiological signal data, and related user data, it can obtain separate permission or consent for the user behavior data, physiological signal data, and related user data through pop-up windows or redirection to a confirmation page. After clearly obtaining separate permission or consent for the user behavior data, physiological signal data, and related user data, it can then retrieve the necessary user behavior data, physiological signal data, and related user data for enabling the embodiments of this disclosure to operate normally.

[0021] In this disclosure, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0022] Before providing a further detailed description of the embodiments of this disclosure, the terms and concepts used in these embodiments are explained, and they are subject to the following interpretations: Smart lock: refers to a door lock device with electronic control unlocking function, including but not limited to home smart door locks, office access control locks, safe locks, hotel and apartment door locks, etc. The smart lock in this embodiment includes at least a handle, a sensor group disposed on the handle, and a main control module.

[0023] Coercive state: This refers to a situation where a user performs an unlocking operation against their will by being forced or threatened by others. When under coercion, the user will experience a series of involuntary physiological and behavioral changes due to fear and tension, such as abnormal grip strength and pressure distribution caused by muscle tension, physiological micro-tremors in the hands, stiff movements causing the movement trajectory to deviate from the normal pattern, and abnormal heart rate and heart rate variability.

[0024] Touch sensor: refers to a sensor used to detect whether a user's hand is in contact with the handle, such as a capacitive touch sensor. In this embodiment of the disclosure, it serves as a trigger source for data acquisition and an auxiliary verification dimension for gripping actions.

[0025] Pressure sensor: refers to a sensor used to collect the pressure at various positions during the user's grip on the handle. It is preferably an array-type pressure sensor that covers the grip area and can output the spatial distribution of pressure and its change over time.

[0026] An Inertial Measurement Unit (IMU) is an inertial sensor device that integrates an accelerometer and a gyroscope, such as a six-axis MEMS sensor (a three-axis accelerometer plus a three-axis gyroscope). It is used to collect motion timing data such as angular velocity and angular acceleration during the user's rotation of the handle. Its high sampling rate can capture minute hand tremors.

[0027] Personalized behavioral baseline: refers to a reference benchmark established based on the user's historical normal unlocking behavior data, which represents the user's normal behavior pattern. For example, it is composed of the mean and standard deviation of behavioral characteristics in various dimensions, and is used to measure the degree of deviation between the current behavior and the user's normal pattern.

[0028] As smart locks gradually become the core entry point for home security, the need for personal safety protection for users in extreme scenarios such as being coerced into unlocking is becoming increasingly prominent. However, existing coercion alarm solutions mostly rely on user-initiated triggers, such as coercive fingerprints, coercive passwords, or emergency buttons on wearable devices. These require users to deliberately perform specific operations under high stress, which carries the risk of recognition failure due to forgetting due to stress, the coerced person recognizing the error, or restrictions on wearing the device. Furthermore, anti-tamper recognition can only detect external violent damage and cannot cover internal threat scenarios where users are coerced into unlocking the door normally.

[0029] Therefore, this disclosure provides an abnormal identification method for a smart lock, a smart lock, and a storage medium. The natural unlocking action of a user holding and turning the handle is used as a carrier for coercion detection. Through passive collection and intelligent analysis of multi-dimensional behavioral characteristics, the anti-coercion effect of non-sensory triggering, normal unlocking, and silent alarm is achieved.

[0030] Please see Figure 1 , Figure 1 This is a schematic diagram of the hardware structure module of the smart lock provided in this embodiment. The smart lock includes: a handle 10, a sensor group 20 disposed on the handle 10, and a main control module 30. The sensor group 20 includes at least a touch sensor 21, a pressure sensor 22, and an inertial measurement unit 23. Furthermore, the smart lock may also include a communication module 40.

[0031] The handle 10 is the component that the user inevitably contacts and operates when unlocking the door. It adopts an ergonomic grip design and is hollow inside to accommodate sensors and wiring. The handle was chosen as the data acquisition carrier because, regardless of whether the user is in a normal or coerced state, unlocking requires the natural action of gripping and turning the handle. By embedding the detection carrier into this inevitable action, the user does not need to perform any additional operations, and the coercor cannot bypass it. This is the structural basis for the non-intrusive triggering and anti-bypass mechanism of this disclosure.

[0032] In some embodiments, the pressure sensor 22 is an array-type pressure sensor covering the gripping area of ​​the handle 10. For example, it employs an 8-row × 16-column flexible piezoresistive thin-film sensor array with a total of 128 independent sensing units, directly attached to the user's finger contact side of the inner wall of the handle. The range is 0~50N, and the sampling rate can be configured to 50Hz. The pressure distribution and its temporal changes are collected by reflecting the pressure magnitude at each location through the resistance changes of each sensing unit. The touch sensor 21 is distributed on the surface of the handle 10. For example, it employs four independent sensing electrodes of a capacitive touch chip, respectively arranged in the upper, lower, left, and right areas of the handle. It is used to detect whether the hand is in contact with the handle and to roughly sense changes in the contact area and contact surface. The inertial measurement unit 23 is built into the handle 10, for example, fixed to a rigid bracket near the handle's pivot axis, with its sensitive axis aligned with the handle's rotation axis to accurately measure rotational angular velocity and angular acceleration. The sampling rate is not less than 200Hz to capture physiological micro-tremors of the hand, primarily in the 8~12Hz frequency band.

[0033] It should be noted that the above sensor selection is only an example. In some embodiments, the pressure sensor 22 may also be a capacitive pressure sensing array, a resistance strain gauge, or a discrete single-point thin-film pressure sensor arranged only in key locations such as the thumb area or palm area; the touch sensor 21 may also be an infrared proximity sensor, an ultrasonic sensor, or a mechanical micro switch; the inertial measurement unit 23 may also be a nine-axis IMU, an industrial-grade IMU with a higher sampling rate, or a separate combination of an independent gyroscope and an accelerometer. This disclosure does not impose specific limitations on these embodiments.

[0034] The main control module 30 is located within the lock body and includes a microprocessor and memory. For example, it may employ an ARM Cortex-M series microprocessor with a floating-point unit. The memory pre-stores the weight parameters of the anomaly recognition model and the user's personalized behavioral baseline data. The main control module 30 is used to run the anomaly recognition model, control the door lock's actions, and send alarm information. The communication module 40 may include one or more wireless communication methods such as Wi-Fi, BLE, 4G / 5G, Zigbee, and NB-IoT, used to send identification information to preset identification targets. The main control module 30 may also integrate an encryption chip for signing and encrypting the identification information, ensuring secure transmission.

[0035] Please see Figure 2 , Figure 2 This is a flowchart illustrating the anomaly detection method for a smart lock provided in this embodiment. This anomaly detection method is applied to the smart lock in the above embodiment and is executed by the main control module. The anomaly detection method includes steps S101 to S103: Step S101: When the touch sensor detects that the user is touching the handle, the main control module starts the sensor group to synchronously collect multi-source time-series data. The multi-source time-series data includes at least the pressure distribution time-series data collected by the pressure sensor during the user's grip on the handle, and the motion time-series data collected by the inertial measurement unit during the user's rotation of the handle. Step S102: The main control module extracts features from the multi-source time-series data to obtain multi-dimensional behavioral features; In step S103, the main control module inputs multi-dimensional behavioral features into the pre-trained anomaly recognition model to obtain the coercion probability value of the user being in a coercive state, and determines the anomaly recognition result of the user based on the coercion probability value.

[0036] Regarding step S101 above, in this embodiment of the disclosure, the smart lock is in a low-power standby state by default, with only the touch sensor operating at a low scanning frequency. When the touch sensor detects that a user is touching the handle, for example, when the capacitance change of any sensing channel exceeds a preset threshold and the contact area meets the gripping determination condition, a wake-up signal is generated, and the main control module then starts the sensor group to synchronously collect multi-source timing data. The reason for using the touch sensor as the collection trigger source is that touch detection has extremely low power consumption and requires no deliberate operation from the user. Once the user naturally grips the handle, the collection automatically begins, ensuring a seamless triggering mechanism and avoiding the problems of users forgetting their operations due to tension or being forcibly bypassed by the coerced in active triggering schemes such as coerced passwords and coerced fingerprints.

[0037] Multi-source time-series data refers to data sequences from multiple heterogeneous sensors arranged in chronological order. Pressure distribution time-series data is collected by pressure sensors during the user's grip on the handle, reflecting the pressure magnitude at various locations within the grip area and its distribution over time. Motion time-series data is collected by the inertial measurement unit (IMU) during the user's rotation of the handle, including sequences such as angular velocity and angular acceleration. The acquisition process begins when the user grips the handle and continues until the handle is rotated to unlock and released. For example, the acquisition termination condition is set when the touch sensor detects the hand leaving and the IMU's angular velocity drops below a preset value for multiple consecutive sampling points. The duration of a single acquisition can be between 1 and 3 seconds. Each sensor data frame carries a unified reference timestamp to achieve microsecond-level time alignment. The emphasis on synchronous acquisition and time alignment is crucial because subsequent analysis of data from different sources on the same timeline is necessary. If the time references of the data sources are inconsistent, the correspondence between pressure changes and motion changes will be disrupted, and the joint representation of multi-dimensional features will be lost.

[0038] Furthermore, the sampling frequency of each sensor can be configured separately, for example, 50Hz for the pressure sensor and 200Hz for the inertial measurement unit. The raw data collected can be preprocessed before being sent to feature extraction, including median filtering of the pressure data, low-pass filtering of the inertial measurement unit data to eliminate circuit noise, normalization according to their respective ranges, and interpolation alignment of all data resampled to a unified time grid based on timestamps. The specific preprocessing parameters of this embodiment are not limited.

[0039] In some embodiments, the multi-source time-series data also includes physiological signal time-series data, which is collected and sent to the smart lock by a wearable device communicating with the smart lock, or collected by a physiological signal sensor installed on the handle during the user's grip on the handle. The physiological signal time-series data includes, for example, physiological parameters that change over time, such as heart rate, heart rate variability, and skin conductivity. The wearable device approach refers to connecting to a user's smart bracelet or smartwatch via near-field communication such as Bluetooth, acquiring physiological data before and after unlocking upon gripping trigger. The handle integration approach involves embedding a photoplethysmography (PPG) sensor on the handle surface in the area where the user's fingers naturally rest, extracting heart rate and heart rate variability by detecting changes in light intensity on the skin of the fingers. The reason for using physiological signals as an optional enhancement dimension rather than a necessary dimension is that when people are fearful and nervous, the excitation of the sympathetic nervous system will cause physiological reactions that are beyond their control, such as an increase in heart rate and a decrease in heart rate variability. Cross-validation of this dimension with the behavioral dimension can further reduce false alarms and false negatives. However, the acquisition of the physiological dimension depends on additional hardware. Designing it as an optional dimension can ensure that the solution can be implemented independently without increasing hardware costs, while reserving enhancement space for high-precision scenarios.

[0040] Regarding step S102 above, in this embodiment of the disclosure, the main control module extracts features from the multi-source time-series data to obtain multi-dimensional behavioral features. These multi-dimensional behavioral features refer to a set of features extracted from the multi-source time-series data that characterize the user's unlocking behavior pattern from multiple perspectives. Feature extraction is performed instead of directly inputting the original time-series data into the model for two reasons: firstly, the original data has high dimensionality and noise, and direct input would increase the computational load of the model, which is not conducive to real-time operation on the smart lock's local microprocessor; secondly, statistical features and time-series features can make abnormal patterns under stress explicit, such as the degree of pressure distribution disorder and the proportion of micro-vibration energy, making it easier for the model to learn the discrimination boundary between normal and stress conditions.

[0041] Please see Figure 3 , Figure 3 yes Figure 2 The flowchart further includes step S102. In some embodiments, the process of the main control module extracting features from multi-source time-series data to obtain multi-dimensional behavioral features may also include steps S201 to S203: Step S201: Extract features from the time series data of pressure distribution to obtain pressure distribution features. The pressure distribution features include at least one of pressure distribution entropy, contact area change rate, and pressure center of gravity movement trajectory. Step S202: Extract features from motion time series data to obtain motion features, which include at least one of the following: high-frequency micro-tremor energy ratio, motion smoothness, and rotational angular velocity change features. Step S203: Obtain multi-dimensional behavioral features based on pressure distribution characteristics and motion characteristics.

[0042] In step S201 above, pressure distribution entropy refers to the information entropy calculated from the pressure distribution of a single frame. It is used to characterize the degree of disorder in the spatial distribution of grip pressure. Under normal conditions, the user's grip pressure distribution is uniform and regular, and the entropy value is relatively stable. Under stress conditions, muscle tension leads to abnormal grip strength and disordered pressure distribution, and the entropy value will deviate significantly. The contact area change rate refers to the rate of change of the contact area between the hand and the handle over time, reflecting the stability of the gripping action. The pressure center of gravity movement trajectory refers to the movement path of the pressure distribution center of gravity within the gripping area. For example, it can be quantified by the trajectory length. Under normal gripping conditions, the center of gravity movement is smooth and regular. Under stress conditions, the center of gravity shakes frequently, and the trajectory length increases abnormally. In addition, pressure distribution characteristics may also include average pressure, pressure variance, left-right and up-down pressure asymmetry, pressure change steepness, etc. This embodiment of the disclosure does not impose specific limitations on these.

[0043] In step S202 above, the high-frequency micro-tremor energy ratio refers to the proportion of energy in the 8-12Hz high-frequency tremor band to the total energy after spectral analysis of the angular velocity signal during rotation. The physical basis for this is that people in a state of fear will produce uncontrollable physiological micro-tremors, manifested as an increase in the spectral energy of this frequency band. Movement smoothness refers to the continuity of the rotational movement, which can be quantified, for example, by the autocorrelation decay time of angular velocity or the root mean square of jerk. Under stress, movements become stiff and rotations become discontinuous, resulting in decreased smoothness. The characteristics of angular velocity variation include statistical quantities such as mean angular velocity, variance, skewness, kurtosis, total rotation angle, and rotation duration. Again, the examples listed here are merely illustrations, and this disclosure does not impose specific limitations on these embodiments.

[0044] In step S203 above, the pressure distribution features and motion features are combined in chronological order, for example, a feature tensor formed by time step and feature dimensions, as multi-dimensional behavioral features. For example, multi-dimensional behavioral features may include 22-dimensional features from the pressure domain and 24-dimensional features from the motion domain, organized into an input tensor within a sliding window of 60 frames and a step size of 1 frame. If time-series physiological signal data is available, physiological domain features such as mean heart rate, heart rate growth rate, time-frequency domain indicators of heart rate variability, and skin conductance change rate can also be added. The reason for needing multi-dimensional behavioral features is that even if a coercor can observe and imitate some single behaviors of the user, such as deliberately loosening their grip, it is difficult to simultaneously imitate all behavioral dimensions, especially dimensions beyond conscious control such as microtremors. Multi-source data fusion makes coercion alarms less likely to be forcibly bypassed, and cross-validation of various dimensions significantly reduces false alarm and false negative rates.

[0045] It should be noted that the embodiments disclosed herein utilize the physiological and behavioral patterns of abnormal grip force distribution caused by muscle tension in users under duress, and the deviation of movement trajectory from the normal pattern caused by stiff and trembling movements. These patterns are physiological reactions that are not subject to the subjective will of the user and the duresser, so that the duress alarm has an objective physiological basis, rather than relying on any active operation that can be deliberately disguised or bypassed.

[0046] Regarding step S103 above, in this embodiment of the disclosure, the main control module inputs multi-dimensional behavioral features into a pre-trained anomaly recognition model to obtain a coercion probability value indicating that the user is in a coercive state, and determines the anomaly recognition result based on the coercion probability value. Here, the anomaly recognition model refers to a pre-trained model used to determine a coercive state from multi-dimensional behavioral features; the coercion probability value is a numerical value output by the model that characterizes the likelihood of the user being in a coercive state, for example, a value between 0 and 1, with a value closer to 1 indicating a higher probability of being in a coercive state.

[0047] In some embodiments, the anomaly detection model is a pre-trained temporal deep learning model, such as a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN). Temporal deep learning models are chosen because behavioral anomalies under stress manifest in the temporal evolution of the unlocking process, such as the persistent appearance of tremors or the gradual stiffening of movements. Temporal models can capture this temporal dependency, outperforming static models that only classify single-frame features. For example, taking a TCN, the model may include an input layer, three causal dilated convolutional layers, a global average pooling layer, a fully connected layer, and an output layer activated with a sigmoid function, outputting a stress probability value between 0 and 1. The input layer has a time step of 60 frames, with 56-dimensional features per frame. The three causal dilated convolutional layers have a kernel size of 3, with dilation coefficients of 1, 2, and 4, and each layer has 64 output channels. Each layer is followed by weight normalization and Dropout with a dropout rate of 0.2. In some embodiments, the anomaly detection model may also employ lightweight traditional models such as Transformer, CNN-LSTM hybrid model, graph neural network, or random forest, and this disclosure does not impose specific limitations on such models.

[0048] Furthermore, the process of inputting multi-dimensional behavioral features into the pre-trained anomaly recognition model is executed locally on the smart lock by the main control module. This means the model inference is completed on the smart lock's local microprocessor, without relying on external networks or cloud computing power. The emphasis on local execution is due to several factors: firstly, in coercive scenarios, the home network may be cut off or interfered with, and cloud-based inference carries the risk of failure; local inference ensures availability in extreme situations. Secondly, user behavior data is sensitive and private, and local processing avoids the risk of data leakage. Moreover, local inference has low latency; for example, after optimization with a neural network acceleration library, a single inference takes only tens of milliseconds, providing real-time recognition results within the few seconds it takes for the user to turn the handle, without slowing down the normal unlocking experience.

[0049] Furthermore, the anomaly detection model can be obtained using a two-stage training strategy of pre-training followed by fine-tuning. In the first stage, a general training dataset is constructed by collecting a large amount of unlocking behavior data from volunteers under normal and simulated coercive conditions. Simulated coercive conditions are collected by inducing tension and fear through watching horror videos or by having subjects unlock the door under pressure of perceived punishment. This training yields a general model that learns common behavioral characteristics under coercive conditions. In the second stage, during user interaction, the general model is fine-tuned using transfer learning based on the user's normal unlocking data, resulting in a personalized model adapted to that user. This strategy is adopted because large-scale acquisition of real coercive data is difficult, while the general model learns common coercive characteristics across population groups, and the personalized fine-tuning adapts to individual differences. The combination of these two approaches balances the model's generalization ability and individual discrimination accuracy.

[0050] Please see Figure 4 , Figure 4 yes Figure 2 The flowchart illustrating the further steps of determining the user's anomaly identification result based on the coercion probability value in step S103 is shown below. In some embodiments, the process of determining the user's anomaly identification result based on the coercion probability value may further include steps S301 to S303: Step S301: If the coercion probability value is greater than or equal to a preset first threshold, determine that the abnormal identification result is that the user is in a coercive state. Step S302: If the coercion probability value is less than a preset second threshold, determine that the abnormal identification result is that the user is in a normal state, wherein the second threshold is less than the first threshold. Step S303: When the coercion probability value is greater than or equal to the second threshold and less than the first threshold, the multi-dimensional behavioral features are compared with the user's personalized behavioral baseline. When the multi-dimensional behavioral features deviate from the personalized behavioral baseline by more than a preset range, the abnormal identification result is determined to be that the user is in a coercive state.

[0051] In the above steps, the first threshold and the second threshold constitute a dual-threshold hierarchical judgment mechanism. For example, the first threshold is 0.85 and the second threshold is 0.6. The specific values ​​can be set according to actual needs, and this embodiment does not impose specific restrictions on this. When the coercion probability value is greater than or equal to the first threshold, the model has given a high-confidence coercion judgment and directly determines that the user is in a coercive state. When the coercion probability value is less than the second threshold, it is determined that the user is in a normal state. When the coercion probability value falls in the suspicious interval between the two thresholds, the model confidence is insufficient. At this time, the user's personalized behavior baseline is introduced for secondary comparison. For example, the Mahalanobis distance between the current multi-dimensional behavioral features and the personalized behavior baseline is calculated. If the deviation exceeds a preset range, such as 3 times the standard deviation, i.e., 3σ, it is determined that the user is in a coercive state; otherwise, it is determined that the user is in a normal state.

[0052] The reason for adopting a tiered judgment mechanism that combines dual thresholds with personalized baseline comparison, rather than a single threshold, is that different users have significantly different behavioral habits. For example, some users are naturally vigorous and fast-moving, and a fixed threshold for these users is prone to false positives; while other users have gentler behaviors, and a fixed threshold is prone to missed detections. Samples within the suspicious interval are precisely those most affected by individual differences. Introducing the user's personalized behavioral baseline for comparison is equivalent to using the user's own historical normal patterns as a reference, rather than using a uniform population standard, thereby significantly reducing false positives while maintaining the recognition rate. It should be noted that the deviation judgment in this embodiment is based on the deviation between the current behavioral characteristics and the user's own baseline, rather than a fixed absolute value standard, so that the coercion judgment can be adapted to each user's behavioral habits.

[0053] In some embodiments, when the anomaly identification result indicates that the user is in a normal state, the main control module can also update the user's personalized behavior baseline based on the multi-dimensional behavioral features collected this time. Specifically, each time an unlocking is determined to be normal, the multi-dimensional behavioral features of this time are added as a new normal sample to the user's behavior database, and the baseline parameters, such as the mean and standard deviation of each feature dimension, are updated accordingly; for example, a new user can establish a stable personal baseline after about 20 normal unlockings. Furthermore, it can also support user-initiated feedback, such as users marking false alarms or missed alarms in historical unlocking records through a mobile application. The system corrects the sample labels accordingly and uses them for incremental learning of the model. For example, feedback samples are merged every preset period to fine-tune the model. During fine-tuning, the parameters of the previous layer of the model can be frozen, and only the parameters of the next fully connected layer can be updated to reduce computational overhead. The reason for setting up a baseline update mechanism is that users' behavioral habits will change slowly over time, such as with age or hand injuries. The static baseline will gradually become inaccurate. Continuous updates ensure that the baseline always fits the user's current behavioral pattern. The longer the usage time, the higher the recognition accuracy and the lower the false alarm rate for the user, achieving personalized adaptation.

[0054] In some embodiments, the smart lock also includes a communication module. When the anomaly detection result indicates the user is under duress, the main control module controls the smart lock to open and ensures the smart lock's output state is consistent with when the user is in a normal state. It also silently sends corresponding identification information to a preset target through the communication module. Maintaining the smart lock's output state consistent with the user's normal state means that the bolt retracts normally, the lock opens normally, and the smart lock does not emit any detectable sound, light, or vibration alerts. The display shows an interface consistent with normal unlocking, ensuring the duress cannot detect any anomalies. This design is because if the alarm is accompanied by any abnormal behavior, such as refusal to unlock, beeping, or flashing indicator lights, it may provoke the duress to take extreme actions, endangering the user's personal safety. By keeping the lock surface normal, the user can enter normally, while the identification process is conducted secretly in the background, prioritizing user safety.

[0055] The identification information can also be called prompt information, alarm information, or alert information. The preset identification object refers to the recipient of the identification information pre-set by the user, such as a preset emergency contact, property security system, or other security identification platform. The identification information includes, for example, user identifier, door lock location, timestamp, and a summary of multi-dimensional behavioral characteristics. If the smart lock integrates a camera, it can also include on-site captured images. The identification information is sent silently through an encrypted channel. Silent transmission means that no user-perceptible prompts are generated during the transmission process. Furthermore, to avoid identification failure due to communication failure, multiple redundancy mechanisms can be set up, such as encrypting and caching the identification information locally and retrying at preset intervals when the network is unavailable, automatically switching between multiple communication methods according to priority, and waking up the camera to record or starting environmental audio recording for evidence preservation. This disclosure embodiment does not impose specific limitations on these aspects.

[0056] In summary, this embodiment of the invention, through the anomaly identification method for smart locks implemented in steps S101 to S103, utilizes a sensor group comprised of a touch sensor, a pressure sensor, and an inertial measurement unit built into the handle. When the touch sensor detects user contact with the handle, synchronous data acquisition automatically begins. During the normal unlocking process of the user naturally gripping and turning the handle, multi-source behavioral data, such as pressure distribution timing data and motion timing data, are acquired. The entire process requires no conscious user intervention and does not rely on additional wearable devices. This fundamentally avoids the risks of active triggering schemes failing due to anxiety, forgetfulness, being detected, or being bypassed, as well as the limitations of wearable device adoption. Subsequently, the multi-source timing data is processed... Multi-dimensional behavioral features are extracted from the action features. By utilizing the physiological and behavioral patterns of users under duress, such as abnormal grip force distribution caused by muscle tension and deviation of movement trajectory from the normal pattern caused by stiff and trembling movements, the pre-trained anomaly recognition model outputs the duress probability value and determines the anomaly recognition result. It can accurately identify internal threat scenarios where users are duressed but open the lock normally on the surface. At the same time, the multi-dimensional features of multi-source data fusion make it difficult for duressers to imitate all behavioral dimensions at the same time, making it difficult to be forcibly bypassed. Moreover, the entire recognition process is embedded in the normal unlocking process, is imperceptible to users, and has strong concealment. Thus, it realizes a seamless duress alarm without the need for active user operation, improving the recognition trigger rate and reliability in real duress scenarios.

[0057] The following is a complete explanation of the above-mentioned anomaly detection method for smart locks, using a specific embodiment as an example. Suppose a user returns home late at night and is followed and coerced by a stranger outside their door, who demands they open the door: When a user reaches out and grasps the smart lock handle, the capacitive touch sensors distributed on the handle's surface detect the contact and immediately activate the main control module. The 8×16 array pressure sensors inside the handle begin synchronous data acquisition at 50Hz, and the six-axis IMU at 200Hz, all without requiring any additional user intervention. Due to fear, the user's muscles tense, resulting in a grip force significantly greater than usual and a disordered pressure distribution. Their hand involuntarily trembles at a high frequency, and their handling motion becomes stiff and disjointed. The data acquisition ends once the user turns the handle to unlock and releases it.

[0058] After filtering, normalizing, and aligning the collected data with timestamps, the main control module extracts pressure distribution features such as pressure distribution entropy, contact area change rate, and pressure center of gravity movement trajectory, as well as motion features such as the proportion of high-frequency micro-vibration energy in the 8~12Hz range, motion smoothness, and rotational angular velocity changes. These features are combined to form a multi-dimensional behavioral feature tensor, which is input into the locally preset TCN anomaly recognition model and outputs a stress probability value of 0.91.

[0059] Since 0.91 is greater than or equal to the first threshold of 0.85, the main control module directly determines that the user is under duress. It then controls the bolt to retract normally, the door lock to open normally, and the display shows the same successful unlocking screen as usual, without any audio or visual prompts. Simultaneously, the main control module silently sends identification information, including the user's identifier, door lock location, timestamp, and on-site captured data, to the user's pre-set emergency contacts and the property security system via an encrypted channel. The duress remains unaware of any abnormalities, and the user's personal safety is guaranteed.

[0060] In contrast, if the same user unlocks the door normally after returning home from get off work, the model outputs a coercion probability value of 0.3, which is less than the second threshold of 0.6. This is considered a normal state and no other alarms are triggered. The behavioral feature of this instance is used to update the user's personalized behavioral baseline. If the user's hand fatigue after carrying heavy objects causes the unlocking action to be slightly abnormal, the model outputs 0.7, which falls into the suspicious range. The system compares this feature with the user's personalized behavioral baseline. If the deviation does not exceed 3σ, it is still considered normal, thus avoiding false alarms.

[0061] Please see Figure 1 The smart lock of this disclosure embodiment can realize the above-mentioned abnormal identification method of smart lock. The smart lock includes: handle 10, sensor group 20 disposed on handle 10 and main control module 30. The sensor group 20 includes at least touch sensor 21, pressure sensor 22 and inertial measurement unit 23.

[0062] The main control module 30 is used to activate the sensor group 20 to synchronously collect multi-source time-series data when the touch sensor 21 detects that the user is touching the handle 10. The multi-source time-series data includes at least the pressure distribution time-series data collected by the pressure sensor 22 during the user's grip on the handle 10, and the motion time-series data collected by the inertial measurement unit 23 during the user's rotation of the handle 10. The main control module 30 is also used to extract features from the multi-source time-series data to obtain multi-dimensional behavioral features. The multi-dimensional behavioral features are input into a pre-trained anomaly recognition model to obtain the stress probability value of the user being in a stress state, and the anomaly recognition result of the user is determined based on the stress probability value.

[0063] In some embodiments, the pressure sensor 22 is an array of pressure sensors covering the gripping area of ​​the handle 10; the touch sensor 21 is distributed on the surface of the handle 10; and the inertial measurement unit 23 is built into the handle 10. This arrangement is because the array of pressure sensors covering the gripping area can fully characterize the spatial distribution of pressure; the touch sensors distributed on the handle surface can promptly and reliably sense contact under any gripping posture; and the inertial measurement unit, built into the handle and installed close to the pivot, can simultaneously sense the overall rotation of the handle and the micro-vibrations of the hand. The three are integrated with the handle, without changing the user's usage habits or requiring the user to wear any additional devices, facilitating direct integration and mass production in existing smart lock products.

[0064] In summary, the smart lock provided in this embodiment of the present disclosure, by executing the abnormal identification method of the smart lock in the above embodiment, fully embeds the duress detection into the user's natural action of holding and turning the handle to unlock the door. It completes the collection of multi-source behavioral data and duress alarm without the user's awareness, and keeps the door lock open normally and silently alarms in the background when a duress state is identified. This achieves a duress alarm that does not require active user operation, is highly concealed, and is not easily bypassed, thereby improving the identification trigger rate and reliability in real duress scenarios and effectively protecting the user's personal safety.

[0065] The specific implementation method of this smart lock is basically the same as the specific embodiment of the anomaly recognition method of the smart lock described above, and will not be repeated here. Under the premise of meeting the requirements of the embodiments of this disclosure, the smart lock can also be equipped with other functional modules to realize the anomaly recognition method of the smart lock in the above embodiments. Furthermore, in addition to home smart door locks, the solution of the embodiments of this disclosure can also be applied to various scenarios requiring duress unlocking recognition, such as office access control, safes, hotel door locks, apartment door locks, and bank self-service equipment security cabins; the embodiments of this disclosure do not impose specific limitations on these applications.

[0066] This disclosure also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method for identifying abnormalities in a smart lock. This electronic device can be integrated into the main control module of the smart lock, or it can be any other smart terminal.

[0067] Please see Figure 5 , Figure 5 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 501 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this disclosure. The memory 502 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 502 and is called and executed by the processor 501 to execute the anomaly identification method of the smart lock according to the embodiments of this disclosure. The input / output interface 503 is used to implement information input and output; The communication interface 504 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 505 transmits information between various components of the device (e.g., processor 501, memory 502, input / output interface 503, and communication interface 504); The processor 501, memory 502, input / output interface 503, and communication interface 504 are connected to each other within the device via bus 505.

[0068] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for identifying anomalies in a smart lock.

[0069] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0070] The embodiments described in this disclosure are for the purpose of more clearly illustrating the technical solutions of this disclosure and do not constitute a limitation on the technical solutions provided by this disclosure. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by this disclosure are also applicable to similar technical problems.

[0071] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this disclosure, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0072] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0073] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0074] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in this disclosure and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such orders can be interchanged where appropriate so that embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0075] It should be understood that in this disclosure, "at least one item" means one or more, and "more than one" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0076] In the embodiments provided in this disclosure, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of the units described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0077] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0078] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0079] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0080] The preferred embodiments of the present disclosure have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present disclosure. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and spirit of the present disclosure shall be within the scope of the claims of the present disclosure.

Claims

1. A method for anomaly detection in a smart lock, characterized in that, Applied to smart locks, the smart lock includes a handle, a sensor group disposed on the handle, and a main control module. The sensor group includes at least a touch sensor, a pressure sensor, and an inertial measurement unit. The anomaly detection method includes: When the touch sensor detects that the user touches the handle, the main control module activates the sensor group to synchronously collect multi-source time-series data. The multi-source time-series data includes at least the pressure distribution time-series data collected by the pressure sensor during the user's grip on the handle, and the motion time-series data collected by the inertial measurement unit during the user's rotation of the handle. The main control module extracts features from the multi-source time-series data to obtain multi-dimensional behavioral features; The main control module inputs the multi-dimensional behavioral features into a pre-trained anomaly recognition model to obtain the coercion probability value of the user being in a coercive state, and determines the anomaly recognition result of the user based on the coercion probability value.

2. The anomaly identification method for smart locks according to claim 1, characterized in that, The multi-source time-series data also includes physiological signal time-series data, which is collected and sent to the smart lock by a wearable device that is communicatively connected to the smart lock, or collected by a physiological signal sensor located on the handle during the user's grip of the handle.

3. The anomaly identification method for smart locks according to claim 1, characterized in that, The main control module extracts features from the multi-source time-series data to obtain multi-dimensional behavioral features, including: The pressure distribution time series data is subjected to feature extraction to obtain pressure distribution features, which include at least one of pressure distribution entropy, contact area change rate and pressure center of gravity movement trajectory. Feature extraction is performed on the motion time series data to obtain motion features, which include at least one of the following: high-frequency micro-tremor energy ratio, motion smoothness, and rotational angular velocity variation features. Multidimensional behavioral features are obtained based on the pressure distribution characteristics and the motion characteristics.

4. The anomaly identification method for smart locks according to claim 1, characterized in that, The anomaly detection model is a pre-trained temporal deep learning model; the main control module inputs the multi-dimensional behavioral features into the pre-trained anomaly detection model, which is then executed locally on the smart lock.

5. The anomaly identification method for smart locks according to claim 1, characterized in that, The step of determining the user's anomaly identification result based on the coercion probability value includes: If the coercion probability value is greater than or equal to a preset first threshold, the abnormal identification result is determined to be that the user is in a coercive state; If the coercion probability value is less than a preset second threshold, the abnormal identification result is determined to be that the user is in a normal state, wherein the second threshold is less than the first threshold; If the coercion probability value is greater than or equal to the second threshold and less than the first threshold, the multi-dimensional behavioral features are compared with the user's personalized behavioral baseline. If the multi-dimensional behavioral features deviate from the personalized behavioral baseline by more than a preset amount, the abnormal identification result is determined to be that the user is in a coercive state.

6. The anomaly identification method for a smart lock according to claim 5, characterized in that, The anomaly detection method further includes: If the anomaly identification result indicates that the user is in a normal state, the user's personalized behavior baseline is updated based on the multi-dimensional behavioral features collected in this study.

7. The anomaly identification method for smart locks according to claim 1, characterized in that, The smart lock also includes a communication module; the anomaly detection method further includes: If the abnormal identification result indicates that the user is in a coerced state, the main control module controls the smart lock to open, controls the output state of the smart lock to be consistent with that when the user is in a normal state, and silently sends the corresponding identification information to the preset identification object through the communication module.

8. A smart lock, characterized in that, It includes a handle, a sensor group disposed on the handle, and a main control module. The sensor group includes at least a touch sensor, a pressure sensor, and an inertial measurement unit. The main control module is used to activate the sensor group to synchronously collect multi-source time-series data when the touch sensor detects that the user is touching the handle. The multi-source time-series data includes at least the pressure distribution time-series data collected by the pressure sensor during the user's grip on the handle, and the motion time-series data collected by the inertial measurement unit during the user's rotation of the handle. The main control module is also used to extract features from the multi-source time-series data to obtain multi-dimensional behavioral features; input the multi-dimensional behavioral features into a pre-trained anomaly recognition model to obtain the coercion probability value of the user being in a coercive state, and determine the user's anomaly recognition result based on the coercion probability value.

9. The smart lock according to claim 8, characterized in that, The pressure sensor is an array-type pressure sensor that covers the gripping area of ​​the handle; the touch sensor is distributed on the surface of the handle; and the inertial measurement unit is built into the handle.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the anomaly identification method of the smart lock according to any one of claims 1 to 7.

Images