A low-power bluetooth key control method and system based on multi-modal perception

Abstract

A low-power Bluetooth key control method and system based on multimodal perception, relating to the field of Bluetooth keys, is disclosed. The method includes: The vehicle end determines the parking environment type using multi-source sensor data such as GPS signal strength, ambient light intensity, and the number of nearby Bluetooth devices, encoding this data into an environment identification code. This code, along with vehicle status information such as door lock status and the last unlock timestamp, is encapsulated in a Bluetooth broadcast packet and broadcast. The key end receives and parses the environment identification code, retrieving the corresponding target weight vector (including RSSI, motion state, and ranging accuracy weight coefficients) from a pre-stored database. The key end collects multi-dimensional feature parameters such as the current RSSI value, motion state, and CS ranging distance value, calculating the RSSI matching score, motion matching score, and ranging reliability score respectively. These are then combined with the target weight vector to calculate the unlock confidence score. A tiered unlocking strategy is executed based on the score's range. Implementing this method can improve the accuracy of Bluetooth car key unlocking judgment.

Description

Technical Field

[0001] This application relates to the field of Bluetooth keys, and in particular to a low-power Bluetooth key control method and system based on multimodal sensing. Background Technology

[0002] With the development of automotive intelligence, Bluetooth Low Energy (BLE) technology has been widely used in keyless entry systems due to its advantages of low power consumption and fast connection. In actual use, vehicles are parked in various environments such as open parking lots, underground garages, and residential communities. The signal propagation characteristics and interference source distribution differ significantly in these environments, placing higher demands on the unlocking accuracy of Bluetooth key systems.

[0003] In related technologies, Bluetooth key control methods mainly employ a single-feature judgment mechanism based on RSSI. The vehicle sends a Bluetooth broadcast signal, the key measures the RSSI value, and when the RSSI value exceeds a preset fixed threshold, an unlock request is sent to the vehicle. The vehicle verifies the request and then unlocks the key. This scheme uses a unified RSSI threshold standard to determine the proximity of the vehicle owner and trigger the unlocking process.

[0004] Bluetooth signal propagation characteristics fluctuate when the parking environment changes. In underground parking garages, multipath reflections cause abnormally high RSSI values, which may lead to false unlocking even when the owner is far away. In open parking lots, the signal attenuates rapidly, and the RSSI value may still be below the threshold when the owner approaches, causing unlocking failure. Due to the use of a fixed threshold and a single feature for judgment, the false trigger rate and unlocking failure rate increase in complex environments. Summary of the Invention

[0005] This application provides a low-power Bluetooth key control method and system based on multimodal perception, which can improve the accuracy of Bluetooth car key unlocking judgment.

[0006] Firstly, this application provides a low-power Bluetooth key control method based on multimodal perception, applied to a Bluetooth key control system. This Bluetooth key control system includes a vehicle end and a key end. The method includes: the vehicle end acquiring multi-source sensor data and determining the parking environment type based on the combined characteristics of the multi-source sensor data, including GPS signal strength, ambient light intensity, and the number of nearby Bluetooth devices; the vehicle end encoding the parking environment type into an environmental identification code and encapsulating the environmental identification code and vehicle status information into a Bluetooth broadcast data packet for broadcasting, the vehicle status information including door lock status and the last unlock timestamp; and the key end receiving the Bluetooth broadcast data packet and parsing it to obtain the environmental identification code. The system retrieves the target weight vector corresponding to the environmental identification code from a pre-stored weight vector library. This target weight vector includes RSSI weight coefficients, motion state weight coefficients, and ranging accuracy weight coefficients. The key terminal acquires multi-dimensional feature parameters through the signal acquisition module and calculates these parameters as RSSI matching score, motion matching score, and ranging reliability score. These multi-dimensional feature parameters include the current RSSI value, motion state, and CS ranging distance value. The key terminal calculates the unlock confidence score based on the RSSI matching score, motion matching score, and ranging reliability score, respectively, and the corresponding weight coefficients in the target weight vector. The key terminal then executes a tiered unlocking strategy based on the numerical range of the unlock confidence score.

[0007] In the above embodiments, the vehicle broadcasts an environment identifier code based on the parking environment type. The key retrieves the corresponding weight vector based on this code, calculates the RSSI value, motion state, and CS ranging distance value as a matching score, and then fuses it with the weight coefficients to obtain an unlock confidence score. A tiered unlocking strategy is then executed based on the confidence score. The weight coefficients are dynamically adjusted according to the environment, multi-dimensional feature weighted fusion suppresses fluctuations of single features, and the tiered strategy avoids the limitations of fixed threshold judgments, thus improving the accuracy of unlocking judgments.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the vehicle terminal acquires multi-source sensor data and determines the parking environment type based on the combined characteristics of the multi-source sensor data. This multi-source sensor data includes GPS signal strength, ambient light intensity, and the number of nearby Bluetooth devices. Specifically, this includes: the vehicle terminal repeatedly acquiring GPS signal strength within a preset acquisition period, calculating the temporal stability characteristic parameter of the GPS signal strength, and determining the degree of GPS signal obstruction based on the temporal stability characteristic parameter. The degree of GPS signal obstruction includes complete obstruction, partial obstruction, and no obstruction. Simultaneously, the vehicle terminal records the acquisition time while acquiring the ambient light intensity value, and sets the ambient light intensity... The illumination deviation feature is calculated by comparing the value with the reference illuminance at the time of acquisition. Based on the illumination deviation feature, it is determined whether there is an artificial occlusion structure. The vehicle end identifies the device attributes of the detected surrounding Bluetooth devices, counts the number of static and dynamic devices among the surrounding Bluetooth devices, and calculates the ratio of the number of static devices to the number of dynamic devices as the device distribution feature value. The vehicle end inputs the GPS signal obstruction degree, illumination deviation feature, and device distribution feature value into a preset classification model for discrimination calculation, and outputs the parking environment type and the corresponding environment confidence level. When the environment confidence level is lower than the preset environment threshold, the vehicle end marks the parking environment type as an unknown environment.

[0009] In the above embodiments, the vehicle-side uses the temporal stability of GPS signals to determine the degree of occlusion, identifies artificial occlusion through illuminance deviation, and reflects device distribution through the ratio of Bluetooth devices. These combined features are input into a classification model to output the environment type. When the environment confidence level is below a threshold, it is marked as an unknown environment. The combined features from multiple sensor sources capture multi-dimensional environmental attributes, and the confidence assessment outputs a conservative judgment when features are unclear, thus improving the reliability of environment type determination.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the step of the key terminal calculating the unlocking confidence score based on the RSSI matching score, motion matching score, and ranging reliability score and the corresponding weight coefficients in the target weight vector specifically includes: the key terminal continuously acquiring the current RSSI value sequence within the time-series data acquisition window, filtering the current RSSI value sequence to obtain a stable RSSI value; the key terminal matching and calculating the stable RSSI value with a pre-stored distance mapping relationship to obtain an estimated distance value corresponding to the stable RSSI value, and using the closeness of the estimated distance value to a preset unlocking distance threshold as the RSSI matching score; the key terminal then calculates the unlocking confidence score from the target weight vector. The RSSI weight coefficient, motion state weight coefficient, and ranging accuracy weight coefficient are extracted from the revector, and adaptively adjusted according to the environment identification code. The key terminal then integrates the adjusted RSSI weight coefficient, motion state weight coefficient, and ranging accuracy weight coefficient with the RSSI matching score, motion matching score, and ranging reliability score to obtain the basic confidence score. The key terminal calculates the data consistency coefficient based on the consistency of the data from each sensor within the time-series data acquisition window, and performs a correction operation between the basic confidence score and the data consistency coefficient to obtain the unlocking confidence score.

[0011] In the above embodiment, the key end filters the RSSI value sequence to obtain a stable value and calculates the estimated distance. The weighting coefficients are adaptively adjusted based on the environmental identification code. The adjusted weights are then fused with the matching score to obtain the basic confidence score. Finally, a correction coefficient is calculated based on sensor data consistency to obtain the final confidence score. Temporal filtering eliminates measurement noise, environmental adaptive adjustment adjusts the propagation characteristics of the matching signal, and consistency correction is based on multi-sensor cross-validation, reducing confidence score calculation errors.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the key terminal executes a tiered unlocking strategy based on the numerical range of the unlock confidence score, specifically including: the key terminal obtaining a high confidence threshold and a low confidence threshold corresponding to the parking environment type from a pre-stored threshold configuration table based on the environment identifier code; when the unlock confidence score is greater than the high confidence threshold, the key terminal generates an automatic unlocking command within a preset response time and sends it to the vehicle terminal through an encrypted communication channel; the vehicle terminal directly executes the unlocking action after receiving the automatic unlocking command; when the unlock confidence score is not greater than the high confidence threshold and When the confidence score is not less than the low confidence threshold, the key terminal collects user behavior feature data through a motion sensor, calculates the similarity between the user behavior feature data and the standard behavior patterns in the pre-stored behavior pattern library, and when the similarity is greater than the preset behavior threshold, the key terminal generates a conditional unlock command and sends it to the vehicle. When the unlock confidence score is less than the low confidence threshold, the key terminal triggers a security protection process, suspends receiving Bluetooth broadcast data packets from the vehicle, and encapsulates the unlock confidence score, the current RSSI value, the CS ranging distance value, and the trigger time as an abnormal event record and stores them in the local storage unit.

[0013] In the above embodiments, the key acquires corresponding thresholds based on the environment. It automatically unlocks with high confidence, conditionally unlocks after verifying behavioral characteristics with medium confidence, and triggers security protection with low confidence. The environment-related thresholds adapt to different signal characteristics. Automatic unlocking with high confidence reduces latency, behavioral verification with medium confidence reduces false locking, and security protection with low confidence blocks abnormal requests, achieving a balance between accuracy and response speed.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, after the step of executing a graded unlocking strategy based on the numerical range of the unlock confidence score at the key end, the method further includes: the key end calculating a time interval feature value based on the door lock status in the vehicle status information and the last unlock timestamp, and retrieving a scene mode identifier matching the time interval feature value from a pre-stored scene mode library; the key end calculating a scene-corrected unlock confidence score based on the scene gain coefficient corresponding to the scene mode identifier and the unlock confidence score; and the key end recording the scene-corrected unlock confidence score, along with the environment identifier code and the vehicle status information, into a historical data cache.

[0015] In the above embodiments, the key terminal retrieves scene pattern identifiers based on time interval feature values, corrects the confidence score based on scene gain coefficients, and records historical data. The scene gain coefficient increases confidence for high-frequency scenes and decreases confidence for abnormal periods, incorporating time-dimensional behavioral patterns into the judgment. Historical data accumulates related information, enhancing the adaptability to user habits.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after the step of executing a graded unlocking strategy based on the numerical range of the unlock confidence score at the key end, the method further includes: the key end detecting the actual unlocking result after executing the unlocking command; increasing the high confidence threshold when unlocking fails; and decreasing the high confidence threshold when unlocking succeeds; and resetting the data in the time-series data acquisition window and clearing the current RSSI value sequence after each unlocking operation is completed.

[0017] In the above embodiment, the key adjusts the threshold based on the actual unlocking result, raising the threshold when there is a failure and lowering it when there is a success. After completion, the timing data is reset and the RSSI sequence is cleared. The result feedback forms a closed-loop adjustment, the dynamic threshold gradually approaches the actual requirement, and the data reset avoids historical interference and ensures independent evaluation, thus improving the adaptability of the unlocking control.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after the step of executing a graded unlocking strategy based on the numerical range of the unlock confidence score at the key end, the method further includes: the key end counting the number of consecutive unlocking failures; when the number of consecutive unlocking failures reaches a preset failure threshold, pausing the reception of Bluetooth broadcast data packets for a preset time period; after the preset time period ends, the key end initializing the high confidence threshold and the low confidence threshold to default values.

[0019] In the above embodiment, the key terminal counts the number of consecutive failures. When the number reaches a threshold, broadcast reception is paused for a preset period. After the period ends, the threshold is initialized to a default value. Failure accumulation identifies abnormal states, pausing reception to block repeated attempts reduces power consumption and risk, a cooling mechanism prevents continuous interference, and threshold initialization eliminates the impact of error adjustments, thus reducing the impact of consecutive failures on the system.

[0020] In a second aspect, embodiments of this application provide a Bluetooth key control system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the Bluetooth key control system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a Bluetooth key control system, cause the Bluetooth key control system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a Bluetooth key control system, cause the Bluetooth key control system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the Bluetooth key control system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. This application broadcasts an environmental identifier code based on the parking environment type at the vehicle end. The key end retrieves the corresponding weight vector based on this code, calculates the matching score by combining the RSSI value, motion state, and CS ranging distance value, and then fuses it with the weight coefficients to obtain an unlock confidence score. A tiered unlocking strategy is then executed based on the confidence score. The weight coefficients are dynamically adjusted according to the environment, and the multi-dimensional feature weighted fusion suppresses fluctuations of single features. The tiered strategy avoids the limitations of fixed threshold judgments, thus improving the accuracy of unlocking judgments.

[0026] 2. This application determines the degree of occlusion by using the temporal stability of GPS signals at the vehicle end, identifies artificial occlusion by using illuminance deviation, and reflects device distribution by using the ratio of Bluetooth devices. The combined features are input into a classification model to output the environment type. When the environment confidence level is below a threshold, it is marked as an unknown environment. The combined features from multiple sensor sources capture multi-dimensional environmental attributes, and the confidence level assessment outputs a conservative judgment when features are unclear, thus improving the reliability of environment type determination.

[0027] 3. This application obtains stable values ​​by filtering the RSSI value sequence at the key end and calculates the estimated distance. It adaptively adjusts the weighting coefficients based on the environmental identification code, then fuses the adjusted weights with the matching score to obtain the basic confidence score. Finally, it calculates a correction coefficient based on sensor data consistency to obtain the final confidence score. Time-series filtering eliminates measurement noise, environmental adaptive adjustment adjusts the propagation characteristics of the matching signal, and consistency correction is based on multi-sensor cross-validation, reducing confidence score calculation errors. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a low-power Bluetooth key control method based on multimodal sensing in an embodiment of this application.

[0029] Figure 2 This is another flowchart illustrating the low-power Bluetooth key control method based on multimodal sensing in this application embodiment;

[0030] Figure 3 This is a schematic diagram of the physical device structure of a Bluetooth key control system in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0034] The technical solutions in this application are mainly applied to keyless vehicle entry systems, particularly digital car key scenarios using Bluetooth Low Energy technology. In practical use, vehicles are parked in various environments, including open parking lots, underground garages, residential surface parking spaces, indoor parking lots, and underground parking garages in commercial complexes. The signal propagation characteristics differ significantly between these environments. In open parking lots, the Bluetooth signal propagation path is relatively simple, and signal attenuation is mainly affected by distance. Underground garages have numerous concrete walls and metal vehicles, resulting in severe multipath reflection and signal scattering, leading to drastic fluctuations in RSSI values. Residential parking spaces are surrounded by numerous fixed Bluetooth devices such as smart locks and smart home devices, creating a complex Bluetooth signal environment.

[0035] Traditional Bluetooth key control methods use a fixed RSSI threshold to determine the unlocking distance. In underground parking garages, multipath reflections cause abnormally strong RSSI values. When the driver is 5 meters away from the vehicle, the measured RSSI value may reach -60dBm, exceeding the set unlocking threshold of -65dBm and triggering a false unlock. In open parking lots, the signal attenuates rapidly, and when the driver approaches within 2 meters, the RSSI value is only -70dBm, below the threshold, causing unlocking failure. This single RSSI feature-based judgment mechanism cannot adapt to environmental changes, leading to increased false triggering and unlocking failure rates, affecting user experience and system security.

[0036] This application solves the problem of insufficient accuracy in Bluetooth car key unlocking judgment under different parking environments by sensing the parking environment type on the vehicle side and broadcasting the environment identification code on the key side, and dynamically adjusting the weight coefficients of RSSI, motion state and ranging accuracy according to the environment identification code, and calculating the unlocking confidence score by integrating multi-dimensional feature parameters. The application also implements a graded unlocking strategy based on the confidence score.

[0037] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a low-power Bluetooth key control method based on multimodal perception in an embodiment of this application.

[0038] S101. The vehicle acquires multi-source sensor data and determines the parking environment type based on the combined characteristics of the multi-source sensor data. The multi-source sensor data includes GPS signal strength, ambient light intensity, and the number of surrounding Bluetooth devices.

[0039] Among them, multi-source sensor data refers to environmental perception data collected from different types of sensors on the vehicle side; GPS signal strength represents the power level of satellite signals received by the global positioning system; ambient light intensity value refers to the current ambient brightness value measured by a photosensitive sensor; the number of surrounding Bluetooth devices represents the total number of identifiable Bluetooth devices detected by the vehicle's Bluetooth module within the scanning range; and parking environment type refers to the vehicle parking scenario classification determined by a comprehensive judgment based on sensor data.

[0040] Specifically, after entering a parked state, the vehicle initiates an environmental perception process. The GPS module continuously receives satellite signals and records signal strength values. When the GPS signal strength consistently falls below a preset threshold, it is determined to be an obstructed environment. The illuminance sensor collects ambient brightness values ​​at fixed time intervals. The system compares the measured illuminance with the theoretical natural illuminance calculated based on the current time and geographical location. If the measured value is significantly lower than the theoretical value, it is inferred that there is building obstruction. The Bluetooth scanning module performs periodic scans at a set power level, identifying and counting the number of responding devices. The vehicle uses the GPS signal strength, illuminance deviation value, and number of Bluetooth devices to construct a feature vector, which is then input into a pre-trained environmental classification model. This model outputs a parking environment type label based on the numerical pattern of the feature vector.

[0041] In some embodiments, parking environment type determination can be achieved in several ways: Optionally, an environmental feature database is established on the vehicle side, which associates and stores sensor data from historical parking locations with manually labeled environmental types. A support vector machine (SVM) algorithm is used to train a classifier. In practical applications, the currently collected sensor data is input into the classifier, and the kernel function maps it to a high-dimensional space for linear segmentation, outputting the environmental type category. Optionally, a rule-matching method is used, where sensor data feature threshold combinations for various environments are predefined. The system matches rules one by one according to priority, logically judging the current data combination against the threshold conditions in the rule base. The environmental type corresponding to the first rule that satisfies all conditions is the determination result. It is understood that other machine learning or fuzzy logic methods can also be used to achieve environmental type recognition, which is not limited here.

[0042] S102. The vehicle end encodes the parking environment type into an environment identification code, and encapsulates the environment identification code and vehicle status information into a Bluetooth broadcast data packet for broadcasting. The vehicle status information includes the door lock status and the last unlock timestamp.

[0043] Among them, the environment identification code refers to the number or character identifier that converts the parking environment type into a number or character identifier according to a predetermined encoding rule; the Bluetooth broadcast data packet refers to the data frame structure that conforms to the Bluetooth Low Energy protocol format; the vehicle status information refers to the set of key parameters that reflect the current operating status of the vehicle; the door lock status indicates the current position status of the vehicle door lock mechanism; and the last unlock timestamp refers to the system time record of the most recent successful unlock operation.

[0044] Specifically, after determining the parking environment type, the vehicle looks up the corresponding code value in the environment code mapping table, which predefines the correspondence between various environments and unique identifiers. The system reads the current door lock status register value from non-volatile memory and simultaneously retrieves the timestamp data from the last unlock event record. The vehicle constructs a data packet according to the Bluetooth broadcast protocol specification, sequentially filling the manufacturer-defined data fields with the environment identifier, door lock status flag, and timestamp value, and adding a CRC checksum to the data packet to ensure transmission reliability. The Bluetooth controller loads the encapsulated data packet into the transmission buffer and transmits the wireless signal to the surrounding space according to the set broadcast interval and transmission power.

[0045] In some embodiments, environment identification code encoding and data encapsulation can be implemented in several ways: Optionally, a fixed-length encoding scheme is adopted, allocating an 8-bit binary code to each predefined parking environment type, mapping the door lock status to a 1-bit flag bit, converting the last unlock timestamp to a 32-bit Unix time format, and concatenating the three data items in a fixed order to form a data payload and filling it into a custom data area of ​​the Bluetooth broadcast packet; Optionally, a dynamic compression encoding method is adopted, allocating variable-length codes according to the historical frequency of the environment type, differentially encoding the relative values ​​of the door lock status and time interval to reduce the data volume, and storing the difference between the current time and the last unlock time in a compact format after logarithmic quantization at the vehicle end. It is understood that other data compression or encryption encoding methods can also be used to implement information encapsulation and broadcasting, which are not limited here.

[0046] S103. After receiving the Bluetooth broadcast data packet, the key terminal parses it to obtain the environment identification code, and retrieves the target weight vector corresponding to the environment identification code from the pre-stored weight vector library. The target weight vector includes RSSI weight coefficient, motion state weight coefficient and ranging accuracy weight coefficient.

[0047] Among them, the environment identification code refers to the environment type encoding value parsed from the Bluetooth broadcast data packet; the pre-stored weight vector library refers to the data structure that stores the mapping relationship between the environment identification code and the weight vector in the key end memory; the target weight vector refers to the weight coefficient combination that matches the currently received environment identification code; the RSSI weight coefficient represents the relative importance of the received signal strength indication in the unlocking judgment; the motion state weight coefficient refers to the weight ratio parameter of the user's motion characteristics in the unlocking decision; and the ranging accuracy weight coefficient represents the weight allocation value of the channel detection ranging result in the confidence calculation.

[0048] Specifically, the Bluetooth receiver module on the key continuously monitors the broadcast channel. When a valid Bluetooth broadcast signal is detected, a reception interrupt is triggered. The system first verifies the CRC checksum of the data packet to confirm that the transmission is error-free. Then, according to the Bluetooth protocol format, it locates the starting position of the manufacturer-defined data field and reads the environment identification code field sequentially according to the predetermined data structure specifications. The key extracts the value from the environment identification code and uses this value as an index key to perform a lookup operation in the locally stored weight vector library. The weight vector library is stored using a hash table or array structure, and each environment identification code corresponds to a vector containing three floating-point numbers. The retrieval module directly accesses the corresponding storage address according to the index, reads the three values ​​of RSSI weight coefficient, motion state weight coefficient, and ranging accuracy weight coefficient, and loads them into the calculation buffer.

[0049] In some embodiments, environment identifier code parsing and weight vector retrieval can be implemented in several ways: Optionally, the key-side uses a state machine parser, which, upon receiving a data packet, sequentially performs reception, verification, field extraction, and data storage operations, converts the extracted identifier code into an integer index value, and directly accesses the corresponding element in the pre-initialized weight array through the array index to read the three weight coefficients; Optionally, a descriptor matching method is used, where the key-side maintains an environment descriptor linked list, with each linked list node containing an environment identifier code and a weight vector pointer. The parsing module traverses the linked list to compare the identifier codes, and upon successful matching, accesses the storage address of the corresponding weight vector through the pointer in the node. It is understood that other data retrieval or caching optimization methods can also be used to achieve fast weight vector location, which is not limited here.

[0050] S104. The key end acquires multi-dimensional feature parameters through the signal acquisition module, and calculates the multi-dimensional feature parameters as RSSI matching score, motion matching score and ranging reliability score respectively. The multi-dimensional feature parameters include the current RSSI value, motion state and CS ranging distance value.

[0051] Among them, the signal acquisition module refers to the multi-type sensors and signal processing circuits integrated in the key terminal; the multi-dimensional feature parameters refer to the parameter vector composed of multiple independent measurement values ​​describing the relative relationship between the key and the vehicle; the current RSSI value represents the instantaneous power intensity index of the Bluetooth broadcast signal received by the key terminal from the vehicle terminal; the motion state refers to the motion mode classification of the key carrier detected by the accelerometer or gyroscope; the CS ranging distance value refers to the estimated spatial distance between the key and the vehicle measured by channel detection technology; and the RSSI matching score represents the quantitative value of the degree of conformity between the current RSSI value and the RSSI threshold corresponding to the expected unlocking distance.

[0052] Specifically, the Bluetooth RF front-end at the key measures the power level of the received signal and converts it into an RSSI value. The triaxial accelerometer in the inertial measurement unit collects acceleration vectors at a fixed sampling rate. After low-pass filtering, the motion recognition algorithm classifies the motion pattern into discrete states such as stationary, slow movement, or fast movement. The channel detection module initiates a ranging request to the vehicle and calculates the physical distance by measuring the round-trip time or phase difference of the signal. The system inputs the current RSSI value into the distance estimation model to obtain the corresponding distance prediction value, calculates the closeness of the predicted distance to the unlocking distance threshold and normalizes it into an RSSI matching score. The measured motion state is compared with the pre-stored standard behavior patterns of approaching vehicles using feature vector cosine similarity calculation to obtain a motion matching score. The reliability of the ranging result is evaluated based on the signal quality parameters of CS ranging and quantified into a ranging reliability score.

[0053] In some embodiments, multidimensional feature parameter acquisition and score calculation can be achieved in various ways: Optionally, the key end is configured with a timer to read the Bluetooth RSSI register value every 100 milliseconds, performs median filtering on the most recent 10 RSSI values ​​and takes the average, back-calculates the distance estimate based on the logarithmic distance path loss model, maps the distance difference to an RSSI matching score using a Gaussian function, continuously acquires triaxial data at a 50Hz sampling rate and calculates the standard deviation and zero-crossing rate of the synthetic acceleration, calculates the Euclidean distance between the feature vector and the reference vectors of walking and running modes respectively, and takes the mode corresponding to the minimum distance as the motion state; Optionally, a sliding window method is used, with the key end maintaining a 3-second data window, performing Kalman filtering on the RSSI sequence to predict the next time value and mapping it to a matching score using a Sigmoid function, extracting the time-domain peak and frequency-domain main frequency features from the acceleration sequence and inputting them into a pre-trained motion classification neural network to obtain the probability distribution of each motion state. It is understood that other signal processing or pattern recognition algorithms can also be used to achieve feature parameter extraction and score calculation, which is not limited here.

[0054] S105. The key end calculates the unlock confidence score based on the RSSI matching score, motion matching score, and ranging reliability score, respectively, and the corresponding weight coefficients in the target weight vector.

[0055] Among them, the unlock confidence score refers to the quantitative index of unlock feasibility calculated by combining the multi-dimensional feature evaluation results and the environmental weight coefficient. The weight coefficient refers to the normalized coefficient that reflects the relative importance of each feature parameter in the current environment. The fusion operation refers to the calculation operation that synthesizes multiple weighted scores into a single evaluation value through linear combination or nonlinear transformation.

[0056] Specifically, the key extracts three values ​​from the target weight vector: RSSI weight coefficient, motion state weight coefficient, and ranging accuracy weight coefficient. Simultaneously, it acquires the RSSI matching score, motion matching score, and ranging reliability score calculated in previous steps. The system performs a weighted summation operation: multiplying the RSSI matching score by the RSSI weight coefficient yields the RSSI weighted component; multiplying the motion matching score by the motion state weight coefficient yields the motion weighted component; and multiplying the ranging reliability score by the ranging accuracy weight coefficient yields the ranging weighted component. The sum of these three weighted components provides the base confidence score. The system can introduce a confidence correction mechanism to adjust the base confidence score based on historical unlocking success rates or current data stability. The corrected value is the final unlocking confidence score.

[0057] In some embodiments, the unlock confidence score can be calculated in several ways: Optionally, a linear weighted fusion method is used. The key reads the RSSI weight coefficient (0.4), motion state weight coefficient (0.3), and ranging accuracy weight coefficient (0.3), multiplies them with the corresponding matching scores, and sums them to obtain the basic confidence score. When the system detects that the standard deviation of the current RSSI value sequence is less than 3dBm, indicating signal stability, no correction operation is performed, and the result is directly output. Optionally, a nonlinear fusion model is used. The key inputs the three matching scores and corresponding weight coefficients into the fuzzy inference system. The fuzzy membership values ​​of each score are calculated through the membership function, and fuzzy inference is performed. After defuzzification, the results are converted into precise values ​​as the basic confidence score. The system further analyzes the consistency between the ranging result and the distance calculated by RSSI and makes positive or negative corrections based on the difference. It is understood that other fusion algorithms or adaptive correction strategies can also be used to achieve a comprehensive evaluation of the confidence score, which is not limited here.

[0058] S106. The key terminal executes a tiered unlocking strategy based on the numerical range of the unlock confidence score.

[0059] The numerical range refers to dividing the range of unlock confidence scores into several non-overlapping continuous segments. The tiered unlocking strategy refers to adopting different levels of strictness in unlocking judgment and execution mechanisms based on the range to which the confidence score belongs. The high confidence range indicates that the numerical range of confidence scores exceeding the preset high threshold corresponds to the automatic unlocking strategy. The medium confidence range indicates that the numerical range of confidence scores between the high and low thresholds corresponds to the conditional unlocking strategy. The low confidence range indicates that the numerical range of confidence scores below the preset low threshold corresponds to the rejection unlocking strategy.

[0060] Specifically, the key terminal queries the local threshold configuration table for the corresponding high-confidence and low-confidence thresholds based on the received environment identifier code. The system compares the calculated unlock confidence score with the two thresholds to determine the confidence range to which the score belongs. When the confidence score is greater than the high-confidence threshold, the key terminal immediately generates unlock request data containing a timestamp and a random number. The data is encrypted using the AES encryption algorithm and encapsulated into a Bluetooth command frame before being sent to the vehicle. After decryption and verification, the vehicle directly drives the door lock actuator to complete the unlocking action. When the confidence score is in the medium-confidence range, the system initiates an additional verification process. It collects behavioral characteristics such as the user's gait frequency and arm swing amplitude using an accelerometer and gyroscope. The feature data is then dynamically time-warped and matched with pre-stored typical owner behavior templates to calculate similarity. A conditional unlock command is only sent when the similarity exceeds the behavioral threshold. When the confidence score is lower than the low-confidence threshold, the system triggers a security protection mechanism, stops processing broadcast data packets from the vehicle, and packages and stores the current confidence score, RSSI value, ranging result, and system time in the abnormal event log.

[0061] In some embodiments, the tiered unlocking strategy can be implemented in multiple ways: Optionally, the key terminal configures a high-confidence threshold of 0.80 and a low-confidence threshold of 0.50 for the underground parking garage environment. When the calculated confidence score is 0.87, it is determined to belong to the high-confidence range. The system generates a data block containing the device ID and timestamp within a 200-millisecond response time and encrypts it using the AES-128 encryption function. The ciphertext is then encapsulated in a Bluetooth GATT feature value write request and sent to the vehicle terminal to execute unlocking. When the confidence score is 0.65, it is determined to belong to the medium-confidence range. The key terminal initiates a behavior recognition module to continuously collect 2 seconds of acceleration data extraction. The frequency and stride feature vectors are compared with five templates in a stored owner behavior template library to calculate cosine similarity. The highest similarity value is compared with a behavior threshold to determine whether to send a conditional unlock command. Optionally, a dynamic threshold adjustment mechanism is used. The key terminal counts the confidence score and actual results of the last 10 unlocking operations. If three consecutive unlocking attempts are successful at a confidence level of 0.82, the high confidence threshold is lowered from 0.85 to 0.80. If a false unlock occurs at a confidence level of 0.88, the high threshold is raised to 0.90. Simultaneously, a secondary confirmation mechanism is introduced in the medium confidence range, requiring the user to press a physical button on the key terminal or confirm the unlocking intention via voice command. It is understood that other adaptive strategies or multi-factor verification methods can also be used to achieve tiered unlocking control; this is not limited here.

[0062] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating the low-power Bluetooth key control method based on multimodal perception in this application embodiment.

[0063] S201. The vehicle end collects GPS signal strength multiple times within a preset collection period, calculates the time-domain stability characteristic parameters of the GPS signal strength, and determines the degree of GPS signal obstruction based on the time-domain stability characteristic parameters. The degree of GPS signal obstruction includes complete obstruction, partial obstruction, and no obstruction.

[0064] The preset acquisition period refers to the time interval pre-set by the vehicle for GPS signal strength sampling. GPS signal strength refers to the power of the satellite signal received by the GPS receiver module. The time-domain stability characteristic parameter refers to the statistical quantity describing the fluctuation characteristics of GPS signal strength over time. The GPS signal obstruction degree refers to the graded label of satellite signal obstruction based on the signal fluctuation pattern. The vehicle starts the GPS data acquisition task, reading the signal strength value output by the GPS module every 2 seconds within a 30-second acquisition period to obtain a signal strength sequence of 15 sampling points. The standard deviation and coefficient of variation of this sequence are calculated as the time-domain stability characteristic parameters. When the standard deviation is less than 2dBm and the coefficient of variation is less than 0.05, it is judged as no obstruction. When the standard deviation is between 2-5dBm and the signal strength drops by more than 10dBm at at least 3 consecutive sampling points, it is judged as partial obstruction. When the signal strength of more than 70% of the sampling points is below -130dBm and the standard deviation is greater than 8dBm, it is judged as complete obstruction. The vehicle stores the judgment results in the environmental judgment buffer for subsequent fusion analysis.

[0065] S202. While collecting ambient light intensity values ​​at the vehicle end, the collection time is recorded. The ambient light intensity values ​​are compared with the reference light intensity corresponding to the collection time to calculate the light deviation characteristics. Based on the light deviation characteristics, it is determined whether there is an artificial shading structure.

[0066] Ambient illuminance refers to the physical quantity of ambient brightness measured by a photosensor. Acquisition time refers to the system time during which the illuminance sampling operation occurs. Reference illuminance refers to the theoretical natural illuminance benchmark calculated based on geographical location, time, and weather conditions. Illuminance deviation characteristic refers to a quantitative index of the difference between the measured illuminance and the reference illuminance. Artificial obstruction structures refer to buildings or obstructions that affect the arrival of natural light on the vehicle. The vehicle triggers the illuminance sensor to perform a measurement every 5 minutes. While recording the measured value of 12000 lux, it reads the real-time clock to obtain the current time as 14:30. Based on the vehicle's GPS coordinates (39.9°N, 116.4°E), it queries the astronomical algorithm library to calculate that the theoretical illuminance under clear weather conditions at this time should be 48000 lux. The calculated deviation rate is (48000-12000) / 48000 = 75%. When the deviation rate exceeds 60% and appears continuously in three consecutive measurements, an artificial obstruction structure is determined to exist. The vehicle then writes the illuminance deviation characteristic value of 0.75 into the corresponding position of the environmental feature array.

[0067] S203. The vehicle end identifies the device attributes of the detected surrounding Bluetooth devices, counts the number of static devices and the number of dynamic devices among the surrounding Bluetooth devices, and calculates the ratio of the number of static devices to the number of dynamic devices as the device distribution characteristic value.

[0068] Device attribute identification refers to the analysis process of extracting device type and mobility characteristics from Bluetooth scan response data. Static devices are those with fixed Bluetooth MAC addresses and unchanged locations, while dynamic devices are those with randomized MAC addresses or whose locations have changed. The device distribution characteristic value is the normalized value of the ratio of the number of static devices to the number of dynamic devices. The vehicle initiates a Bluetooth scanning task and scans continuously for 60 seconds. Among the 23 Bluetooth devices scanned, 15 fixed Bluetooth beacon devices are identified as static devices by parsing the device type field in the broadcast packets. The remaining 8 devices are classified as dynamic devices because their MAC addresses changed during the scan or their RSSI values ​​fluctuated by more than 15 dBm. The calculated device distribution characteristic value is 15 / 8 = 1.875. The vehicle performs a Sigmoid normalization transformation on this value, resulting in a characteristic value of 0.867. This value reflects the environmental characteristic of static devices dominating the current location. The vehicle stores the normalized device distribution characteristic value in a feature vector array.

[0069] S204. The vehicle terminal inputs the GPS signal obstruction degree, illumination deviation characteristics and device distribution characteristics into the preset classification model for discrimination calculation, and outputs the parking environment type and the corresponding environment confidence level. When the environment confidence level is lower than the preset environment threshold, the vehicle terminal marks the parking environment type as an unknown environment.

[0070] The pre-defined classification model refers to the environment type discrimination algorithm model that has been pre-trained and deployed on the vehicle. The discrimination operation refers to the process of inputting multi-dimensional features into the model and performing forward propagation to obtain the classification result. The environment confidence is the probability assessment value of the model on the reliability of the classification result. The pre-defined environment threshold is the confidence threshold used to determine whether the classification result is credible. The unknown environment refers to the conservative judgment result given when the classification confidence is insufficient. The vehicle reads the GPS signal occlusion level encoding value 2 (partial occlusion), the illumination deviation feature value 0.75, and the device distribution feature value 0.867 from the feature buffer. These three values ​​form an input vector [2, 0.75, 0.867], which is then fed into a pre-loaded random forest classification model. The model contains 50 decision trees that output the environmental type voting results. Among them, 32 trees output the "underground garage" category and 18 trees output the "indoor parking lot" category. The confidence score of the underground garage category is calculated to be 32 / 50 = 0.64. Since 0.64 is lower than the preset environmental threshold of 0.70, the vehicle marks the parking environment type as an unknown environment and sets a special value of 0xFF in the environmental identification code field to indicate that the environmental judgment is uncertain.

[0071] S205. The vehicle end encodes the parking environment type into an environment identification code, and encapsulates the environment identification code and vehicle status information into a Bluetooth broadcast data packet for broadcasting. The vehicle status information includes the door lock status and the last unlock timestamp.

[0072] Parking environment type encoding refers to the mapping operation that converts the environmental classification result into a digital identifier. The environment identifier code refers to a fixed-length digital code that represents a specific environment type. The Bluetooth broadcast data packet refers to the data frame structure that conforms to the Bluetooth Low Energy Broadcast protocol. The vehicle status information refers to the set of parameters that describe the key status of the vehicle. The door lock status refers to the current position status of the door locking mechanism. The last unlock timestamp refers to the Unix time value when the most recent unlocking operation was completed. The vehicle looks up the environmental identifier code 0x03 corresponding to "underground garage" in the environmental coding table, reads the door lock controller register to get the door lock status value 0x01 (locked), and extracts the last unlock timestamp 1735660800 (corresponding to 14:00 on December 31, 2025) from the event log. The environmental identifier code, door lock status and timestamp are encapsulated into the manufacturer data field of the Bluetooth broadcast packet according to the byte sequence [0x03, 0x01, 0x67, 0x82, 0x5E, 0x00]. The broadcast interval is set to 100 milliseconds and the transmission power is 0dBm. The Bluetooth controller sends this data packet once every 100 milliseconds on broadcast channels 37, 38 and 39.

[0073] S206. After receiving the Bluetooth broadcast data packet, the key terminal parses it to obtain the environment identification code, and retrieves the target weight vector corresponding to the environment identification code from the pre-stored weight vector library. The target weight vector includes RSSI weight coefficient, motion state weight coefficient and ranging accuracy weight coefficient.

[0074] Bluetooth broadcast data packet reception refers to the process of the key end listening to the broadcast channel and capturing valid data frames. Environment identification code parsing refers to the byte reading operation of extracting the environment type code from the data packet payload. The pre-stored weight vector library refers to the mapping relationship table between environment codes and weight coefficients stored in the key end's memory. The target weight vector refers to the combination of weight coefficients corresponding to the current environment identification code. The RSSI weight coefficient refers to the relative importance value of RSSI features in unlocking judgment. The motion state weight coefficient refers to the weight ratio parameter of motion features. The ranging accuracy weight coefficient refers to the confidence weight value of the ranging result. The Bluetooth module on the key end detects a valid broadcast packet on broadcast channel 37. After CRC verification, it reads the environmental identification code byte value 0x03 from offset 0 of the manufacturer data field. Using 0x03 as an index, it searches the hash table of the local weight vector library. It locates the storage area at address 0x2000+0x03*12 and reads 12 bytes of data. It parses the data into three values ​​in floating-point format: RSSI weight coefficient 0.35, motion state weight coefficient 0.40, and ranging accuracy weight coefficient 0.25. The key end loads these three weight coefficients into the register of the calculation unit for subsequent confidence score fusion calculation.

[0075] S207. The key end acquires multi-dimensional feature parameters through the signal acquisition module, and calculates the multi-dimensional feature parameters as RSSI matching score, motion matching score and ranging reliability score respectively. The multi-dimensional feature parameters include the current RSSI value, motion state and CS ranging distance value.

[0076] The signal acquisition module refers to the hardware combination of Bluetooth RF front-end, inertial sensor and ranging unit integrated in the key terminal. Multidimensional feature parameters refer to multiple independent measurements describing the relative relationship between the key and the vehicle. The current RSSI value refers to the instantaneous power intensity of the vehicle broadcast signal received by the key terminal. Motion state refers to the motion pattern of the key carrier identified by the inertial sensor. CS ranging distance value refers to the physical distance measured by channel detection technology. RSSI matching score refers to the conformity quantification value of the current RSSI value and the RSSI corresponding to the target distance. Motion matching score refers to the similarity score between the measured motion state and the standard behavior of the approaching vehicle. Ranging reliability score refers to the confidence quantification index of the CS ranging result. The key reads the current RSSI value from the Bluetooth RF register as -62dBm. Substituting this value into the path loss model L=A+10n·log(d), the distance is calculated to be 2.8 meters. The difference of 0.2 meters between the calculated distance and the target unlocking distance of 3 meters is mapped to an RSSI matching score of 0.92 using the Gaussian function exp(-0.2² / 2σ²). The step frequency of 1.8Hz and the standard deviation of acceleration of 0.45m / s² are extracted from the data sequence collected by the triaxial accelerometer to form a feature vector. The cosine similarity is calculated with the stored "walking approach" pattern template vector to obtain a motion matching score of 0.88. The CS ranging module measures a flight time of 9.3 nanoseconds, corresponding to a distance of 2.79 meters. Based on the signal quality parameter SNR=18dB, the ranging reliability score of 0.85 is obtained by looking up a table.

[0077] S208. The key end continuously acquires the current RSSI value sequence within the time series data acquisition window, and performs filtering processing on the current RSSI value sequence to obtain stable RSSI values.

[0078] The timing data acquisition window refers to the fixed time period set by the key terminal for continuous acquisition of RSSI data. The current RSSI value sequence refers to a series of signal strength measurements acquired at sampling intervals within this time period. Filtering refers to the algorithm operation of removing random noise and outliers from the measurement data using digital signal processing methods. The stable RSSI value refers to a representative value that can accurately reflect the true signal strength level after filtering.

[0079] The key terminal sets the acquisition window duration to 3 seconds and the sampling interval to 100 milliseconds. Within the window time, 30 RSSI samples are performed to obtain the numerical sequence [-65, -63, -68, -64, -62, -67, -64, -66, -63, -65, -69, -64, -62, -63, -64, -66, -65, -64, -67, -63, -62, -64, -65, -66, -64, -63, -65, -64, -66, -65] dBm. First, outlier removal is performed on the sequence. The -69 dBm value deviates from the mean by more than 2 standard deviations and is removed from the sequence, resulting in 29 valid sampling points. Then, a 5-point median filter is applied to divide the sequence into 6 groups of 5 consecutive values. The 5 values ​​in each group are sorted by size and the median value is taken. For example, the first group [-65, -63, -68, -64, -62] is sorted into [-68, -65, -64, -63, -62] and the median value is taken as -64dBm. The filtered sequence [-64, -64, -63, -64, -65, -64] is obtained by processing in sequence. Finally, a weighted moving average is calculated for the filtered sequence. The most recent 10 filtered values ​​are assigned increasing weights [0.04, 0.06, 0.07, 0.08, 0.09, 0.10, 0.12, 0.13, 0.15, 0.16]. A weighted summation is performed: (-64×0.16-64×0.15-63×0.13-64×0.12-65×0.10-64×0.09-64×0.08-63×0.07-64×0.06-64×0.04) = -63.87dBm. This value is rounded to -64dBm and used as the stable RSSI value.

[0080] S209. The key end performs matching calculations between the stable RSSI value and the pre-stored distance mapping relationship to obtain the estimated distance value corresponding to the stable RSSI value. The degree of closeness between the estimated distance value and the preset unlocking distance threshold is used as the RSSI matching score.

[0081] The pre-stored distance mapping relationship refers to the mathematical model or lookup table of the correspondence between RSSI signal strength and physical distance stored in the key end memory. Matching calculation refers to the calculation process of inputting the measured RSSI value into the mapping relationship to solve for the corresponding distance. The estimated distance value refers to the spatial distance between the key and the vehicle calculated through the RSSI mapping relationship. The preset unlocking distance threshold refers to the maximum effective distance limit that the system allows to perform unlocking operations. Proximity refers to the quantitative evaluation index of the difference between the estimated distance and the unlocking threshold.

[0082] The key loads the pre-stored logarithmic distance path loss model RSSI(d) = RSSI0-10n·log from memory. 10(d / d0), where RSSI0 is the signal strength at a reference distance d0 = 1 meter (-50dBm), and the path loss index n = 2.8. For calibration in a garage environment, the stable RSSI value of -64dBm is substituted into the formula for inverse calculation to solve for the distance: -64 = -50 - 10 × 2.8 × log 10 (d / 1), calculated to be 10 × 2.8 × log 10 (d)=14, log 10 (d)=0.5, d=10^0.5=3.16 meters, this value is determined as the estimated distance value. The preset unlocking distance threshold in the system configuration is read as 3.5 meters, and the distance difference Δd=3.5-3.16=0.34 meters is calculated. The proximity score function Score=exp(-((Threshold-Distance) / σ)²) is constructed, where σ=0.5 meters is the distance tolerance parameter. Substituting the value, Score=exp(-((3.5-3.16) / 0.5)²)=exp(-(0.68)²)=exp(-0.46)=0.631, and this score is used as the RSSI matching score.

[0083] S210: The key end extracts the RSSI weight coefficient, motion state weight coefficient, and ranging accuracy weight coefficient from the target weight vector, and adaptively adjusts the RSSI weight coefficient, motion state weight coefficient, and ranging accuracy weight coefficient according to the environment identification code.

[0084] Target weight vector extraction refers to the operation of reading three weight coefficients corresponding to the current environment from the local weight library. Adaptive adjustment refers to the calculation process of dynamically correcting the extracted weight coefficients according to the environmental characteristics to adapt to specific environmental conditions.

[0085] The key terminal reads the initial weight vector corresponding to environment identifier code 0x03 from the weight vector library, obtaining RSSI weight coefficient w1=0.35, motion state weight coefficient w2=0.40, and ranging accuracy weight coefficient w3=0.25. The system recognizes that 0x03 represents an underground parking garage environment and queries the environment adjustment parameter table for the three correction factors of this environment: RSSI signal attenuation is large, so the weight needs to be increased, therefore RSSI correction factor α1=1.20; motion sensor performs stably in enclosed space, therefore motion correction factor α2=0.95; multipath effect affects ranging accuracy, so the weight needs to be reduced, therefore ranging correction factor α3=0.85. To perform weight adjustment calculations, first multiply each weight by its corresponding correction factor: w1'=0.35×1.20=0.42, w2'=0.40×0.95=0.38, w3'=0.25×0.85=0.2125. The adjusted total weight S is then calculated as: S=0.42+0.38+0.2125=1.0125. To ensure the weight sum is equal to 1, normalization is performed. The final RSSI weight coefficients are calculated as: w1''=0.42 / 1.0125=0.415, motion state weight coefficient w2''=0.38 / 1.0125=0.375, and ranging accuracy weight coefficient w3''=0.2125 / 1.0125=0.210. The normalization result is verified as: 0.415+0.375+0.210=1.000.

[0086] S211. The key end integrates the adjusted RSSI weighting coefficient, motion state weighting coefficient, and ranging accuracy weighting coefficient with the RSSI matching score, motion matching score, and ranging reliability score to obtain the basic confidence score.

[0087] Fusion calculation refers to the mathematical operation of combining multiple sub-scores into a comprehensive score by weighting and summing them according to their respective weight coefficients. The basic confidence score refers to the preliminary confidence assessment value that only considers feature weight fusion and has not yet undergone data consistency correction.

[0088] The key reads the adjusted three weighting coefficients from the calculation buffer: RSSI weight 0.415, motion state weight 0.375, and ranging accuracy weight 0.210. Simultaneously, it reads three matching scores: RSSI matching score 0.631, motion matching score 0.88, and ranging reliability score 0.85. A weighted fusion operation is performed, calculating the RSSI weighted component P1 = 0.415 × 0.631 = 0.262, the motion state weighted component P2 = 0.375 × 0.88 = 0.330, and the ranging accuracy weighted component P3 = 0.210 × 0.85 = 0.179. The base confidence score C_base = P1 + P2 + P3 = 0.262 + 0.330 + 0.179 = 0.771 is obtained by summing the three weighted components. This score reflects the comprehensive satisfaction of the unlocking conditions by the multi-dimensional features under the current environmental weight configuration. The value ranges from 0 to 1, with a value closer to 1 indicating a more sufficient unlocking condition. The calculated result of 0.771 is stored in the confidence evaluation register for consistency correction processing.

[0089] S212. The key end calculates the data consistency coefficient based on the consistency of the data from each sensor in the time-series data acquisition window, and performs a correction operation on the basic confidence score and the data consistency coefficient to obtain the unlock confidence score.

[0090] Data consistency refers to the degree of agreement between measurement results from different sensors within a time window. The data consistency coefficient is a normalized index that quantifies the consistency degree into a value within the range of 0 to 1. Correction operation refers to the calculation operation that uses the consistency coefficient to adjust the reliability of the basic confidence score.

[0091] The key extracts three distance measurement results from the time-series data buffer for cross-validation: RSSI estimates the distance to be 3.16 meters, the CS ranging module directly measures the distance to be 3.08 meters, and the motion sensor calculates the current distance based on gait recognition, which is calculated as the initial distance of 6.8 meters minus the moving distance of 3.6 meters, equaling 3.20 meters, based on the user's approach to the vehicle at a speed of 1.2 meters per second in the last 3 seconds. Calculate the mean of the three distances, d_mean = (3.16 + 3.08 + 3.20) / 3 = 3.147 meters. Calculate the sum of squares of the deviations of each distance from the mean, Σ(d - d_mean)² = (3.16 - 3.147)² + (3.08 - 3.147)² + (3.20 - 3.147)² = 0.000169 + 0.004489 + 0.002809 = 0.007467. Calculate the standard deviation, σ = √(0.007467 / 3) = 0.050 meters. The standard deviation is mapped to a consistency coefficient using a Gaussian decay function K=exp(-σ² / (2θ²)), where the threshold parameter θ=0.10 m. Substituting this into the calculation, K=exp(-(0.050)² / (2×0.10²))=exp(-0.0025 / 0.02)=exp(-0.125)=0.883. A correction operation is then performed using the weighted correction formula C_final=C_base×(λ+(1-λ)×K), where λ=0.6 is the base weight retention coefficient. Substituting this into the value, C_final=0.771×(0.6+(1-0.6)×0.883)=0.771×(0.6+0.353)=0.771×0.953=0.735, yielding a final unlock confidence score of 0.735.

[0092] S213. The key terminal executes a tiered unlocking strategy based on the numerical range of the unlock confidence score.

[0093] A numerical interval refers to several non-overlapping continuous segments divided according to the range of confidence scores. A tiered unlocking strategy refers to differentiated unlocking judgment rules and execution processes designed for different confidence intervals.

[0094] The key terminal reads the corresponding threshold ranges for the underground parking garage environment from the environment configuration table: high confidence threshold T_high = 0.80, low confidence threshold T_low = 0.55. The calculated unlock confidence score of 0.735 is compared with the thresholds. Since 0.735 < 0.80 and 0.735 ≥ 0.55, the score is determined to fall within the medium confidence range [0.55, 0.80]. The system calls the conditional unlocking strategy module corresponding to the medium confidence range, which requires additional authentication. The key terminal activates the triaxial accelerometer to continuously collect data for 2 seconds at a sampling rate of 100Hz, obtaining 600 triaxial acceleration sampling points. Peak detection is performed on the Z-axis acceleration sequence to identify 3 complete gait cycles, and the step frequency is calculated as 3 steps / 2 seconds = 1.5Hz. For each gait cycle, the peak, trough, and zero-crossing time interval of acceleration are extracted to form a feature vector [9.2, 0.8, 0.65]. Five historical proximity behavior template vectors of the driver are loaded from the local behavior template library. The Euclidean distance between the current feature vector and each template vector is calculated, resulting in distance values ​​[0.12, 0.31, 0.08, 0.19, 0.25]. The third template, corresponding to the smallest distance of 0.08, is selected, and the behavior similarity S_behavior = 1 / (1+0.08) = 0.926 is calculated. The behavior verification threshold T_behavior = 0.85 is used for judgment. Since 0.926 > 0.85, the verification passes. The key generates an unlock request data packet containing the device's unique identifier 0x4A3C2B1D, the current timestamp 1735664400, and a 32-bit random number 0x8F2E7A5C. The data packet is encrypted using a pre-shared key using AES-128-CBC encryption. The encrypted 128-bit ciphertext is then encapsulated in a write request containing the Bluetooth GATT signature value 0x2A00, and a conditional unlock command is sent to the vehicle.

[0095] In some embodiments, this step specifically includes:

[0096] The key terminal retrieves the high-confidence threshold and low-confidence threshold corresponding to the parking environment type from the pre-stored threshold configuration table based on the environment identification code. When the unlock confidence score is greater than the high-confidence threshold, the key terminal generates an automatic unlock command within a preset response time and sends it to the vehicle terminal through an encrypted communication channel. After receiving the automatic unlock command, the vehicle terminal directly executes the unlocking action. When the unlock confidence score is not greater than the high-confidence threshold and not less than the low-confidence threshold, the key terminal collects user behavior feature data through motion sensors and calculates the similarity between the user behavior feature data and the standard behavior patterns in the pre-stored behavior pattern library. When the similarity is greater than the preset behavior threshold, the key terminal generates a conditional unlock command and sends it to the vehicle terminal. When the unlock confidence score is less than the low-confidence threshold, the key terminal triggers a security protection process, suspends the reception of Bluetooth broadcast data packets broadcast by the vehicle terminal, and encapsulates the unlock confidence score, the current RSSI value, the CS ranging distance value, and the trigger time as an abnormal event record and stores it in the local storage unit.

[0097] In some embodiments, after the key terminal executes the tiered unlocking strategy based on the numerical range of the unlock confidence score, the method further includes: the key terminal calculates a time interval feature value based on the door lock status and the last unlock timestamp in the vehicle status information, and retrieves a scene mode identifier matching the time interval feature value from a pre-stored scene mode library; the key terminal calculates a scene-corrected unlock confidence score based on the scene gain coefficient corresponding to the scene mode identifier and the unlock confidence score; the key terminal records the scene-corrected unlock confidence score, along with the environment identifier code and vehicle status information, into a historical data cache; the key terminal detects the actual unlocking result after executing the unlocking command, increasing the high confidence threshold when unlocking fails and decreasing the high confidence threshold when unlocking succeeds; after each unlocking operation, the key terminal resets the data in the timing data acquisition window and clears the current RSSI value sequence; the key terminal counts the number of consecutive unlocking failures, and when the number of consecutive unlocking failures reaches a preset failure threshold, it pauses receiving Bluetooth broadcast data packets for a preset time period; after the preset time period ends, the key terminal initializes the high confidence threshold and low confidence threshold to default values.

[0098] In some embodiments, after the key terminal executes the tiered unlocking strategy based on the numerical range of the unlock confidence score, the method further includes: the key terminal calculates a time interval feature value based on the door lock status and the last unlock timestamp in the vehicle status information, and retrieves a scene mode identifier matching the time interval feature value from a pre-stored scene mode library; the key terminal calculates a scene-corrected unlock confidence score based on the scene gain coefficient corresponding to the scene mode identifier and the unlock confidence score; the key terminal records the scene-corrected unlock confidence score, along with the environment identifier code and vehicle status information, into a historical data cache; the key terminal detects the actual unlocking result after executing the unlocking command, increasing the high confidence threshold when unlocking fails and decreasing the high confidence threshold when unlocking succeeds; after each unlocking operation, the key terminal resets the data in the timing data acquisition window and clears the current RSSI value sequence; the key terminal counts the number of consecutive unlocking failures, and when the number of consecutive unlocking failures reaches a preset failure threshold, it pauses receiving Bluetooth broadcast data packets for a preset time period; after the preset time period ends, the key terminal initializes the high confidence threshold and low confidence threshold to default values.

[0099] In some embodiments, after the key terminal executes the tiered unlocking strategy based on the numerical range of the unlock confidence score, the method further includes: the key terminal calculates a time interval feature value based on the door lock status and the last unlock timestamp in the vehicle status information, and retrieves a scene mode identifier matching the time interval feature value from a pre-stored scene mode library; the key terminal calculates a scene-corrected unlock confidence score based on the scene gain coefficient corresponding to the scene mode identifier and the unlock confidence score; the key terminal records the scene-corrected unlock confidence score, along with the environment identifier code and vehicle status information, into a historical data cache; the key terminal detects the actual unlocking result after executing the unlocking command, increasing the high confidence threshold when unlocking fails and decreasing the high confidence threshold when unlocking succeeds; after each unlocking operation, the key terminal resets the data in the timing data acquisition window and clears the current RSSI value sequence; the key terminal counts the number of consecutive unlocking failures, and when the number of consecutive unlocking failures reaches a preset failure threshold, it pauses receiving Bluetooth broadcast data packets for a preset time period; after the preset time period ends, the key terminal initializes the high confidence threshold and low confidence threshold to default values.

[0100] The Bluetooth key control system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the physical device structure of a Bluetooth key control system in an embodiment of this application.

[0101] It should be noted that, Figure 3 The structure of the Bluetooth key control system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0102] like Figure 3 As shown, the Bluetooth key control system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 302 or programs loaded from storage section 308 into Random Access Memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0103] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including hard disks, etc.; and communication section 309 including network interface cards such as LAN (Local Area Network) cards, modems, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0104] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0105] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0106] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0107] Specifically, the Bluetooth key control system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the low-power Bluetooth key control method based on multimodal perception provided in the above embodiment.

[0108] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the Bluetooth key control system described in the above embodiments; or it may exist independently and not assembled into the Bluetooth key control system. The storage medium carries one or more computer programs that, when executed by a processor of the Bluetooth key control system, cause the Bluetooth key control system to implement the low-power Bluetooth key control method based on multimodal perception provided in the above embodiments.

[0109] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0110] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0111] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A low-power Bluetooth key control method based on multimodal sensing, characterized in that, Applied to a Bluetooth key control system, the Bluetooth key control system comprising a vehicle end and a key end, the method includes: The vehicle terminal acquires multi-source sensor data and determines the parking environment type based on the combined characteristics of the multi-source sensor data. The multi-source sensor data includes GPS signal strength, ambient light intensity value, and the number of surrounding Bluetooth devices. The vehicle terminal encodes the parking environment type into an environment identification code, and encapsulates the environment identification code and vehicle status information into a Bluetooth broadcast data packet for broadcasting. The vehicle status information includes the door lock status and the last unlock timestamp. After receiving the Bluetooth broadcast data packet, the key terminal parses the environment identification code and retrieves the target weight vector corresponding to the environment identification code from the pre-stored weight vector library. The target weight vector includes RSSI weight coefficient, motion state weight coefficient and ranging accuracy weight coefficient. The key terminal acquires multi-dimensional feature parameters through the signal acquisition module, and calculates the multi-dimensional feature parameters as RSSI matching score, motion matching score and ranging reliability score respectively. The multi-dimensional feature parameters include the current RSSI value, motion state and CS ranging distance value. The key terminal calculates the unlock confidence score based on the RSSI matching score, the motion matching score, and the ranging reliability score, respectively, and the corresponding weight coefficients in the target weight vector. The key terminal executes a tiered unlocking strategy based on the numerical range of the unlock confidence score.

2. The method according to claim 1, characterized in that, The vehicle-side acquires multi-source sensor data and determines the parking environment type based on the combined characteristics of the multi-source sensor data. The multi-source sensor data includes steps such as GPS signal strength, ambient light intensity, and the number of nearby Bluetooth devices. Specifically, this includes: The vehicle terminal collects GPS signal strength multiple times within a preset collection period, calculates the time-domain stability characteristic parameter of the GPS signal strength, and determines the degree of GPS signal obstruction based on the time-domain stability characteristic parameter. The degree of GPS signal obstruction includes complete obstruction, partial obstruction, and no obstruction. The vehicle terminal collects ambient light illuminance values ​​and records the collection time. The ambient light illuminance values ​​are compared with the reference illuminance corresponding to the collection time to obtain the light deviation characteristics. Based on the light deviation characteristics, it is determined whether there is an artificial shading structure. The vehicle terminal identifies the device attributes of the detected surrounding Bluetooth devices, counts the number of static devices and the number of dynamic devices among the surrounding Bluetooth devices, and calculates the ratio of the number of static devices to the number of dynamic devices as a device distribution feature value. The vehicle terminal inputs the GPS signal obstruction degree, the illumination deviation characteristics, and the device distribution characteristics into a preset classification model for discrimination calculation, and outputs the parking environment type and the corresponding environment confidence level. When the environment confidence level is lower than a preset environment threshold, the vehicle terminal marks the parking environment type as an unknown environment.

3. The method according to claim 1, characterized in that, The step of calculating the unlock confidence score based on the RSSI matching score, the motion matching score, and the ranging reliability score, respectively, and the corresponding weight coefficients in the target weight vector, specifically includes: The key end continuously acquires the current RSSI value sequence within the time-series data acquisition window, and performs filtering processing on the current RSSI value sequence to obtain a stable RSSI value; The key end performs matching calculations between the stable RSSI value and the pre-stored distance mapping relationship to obtain the estimated distance value corresponding to the stable RSSI value, and uses the closeness of the estimated distance value to the preset unlocking distance threshold as the RSSI matching score; The key end extracts the RSSI weight coefficient, the motion state weight coefficient, and the ranging accuracy weight coefficient from the target weight vector, and adaptively adjusts the RSSI weight coefficient, the motion state weight coefficient, and the ranging accuracy weight coefficient according to the environment identification code; The key terminal integrates the adjusted RSSI weight coefficient, the motion state weight coefficient, and the ranging accuracy weight coefficient with the RSSI matching score, the motion matching score, and the ranging reliability score to obtain the basic confidence score. The key terminal calculates a data consistency coefficient based on the consistency of data from each sensor within the time-series data acquisition window, and performs a correction operation between the basic confidence score and the data consistency coefficient to obtain an unlock confidence score.

4. The method according to claim 1, characterized in that, The step of the key terminal executing a tiered unlocking strategy based on the numerical range of the unlocking confidence score specifically includes: The key terminal obtains the high confidence threshold and low confidence threshold corresponding to the parking environment type from the pre-stored threshold configuration table based on the environment identification code; When the unlock confidence score is greater than the high confidence threshold, the key terminal generates an automatic unlock command within a preset response time and sends it to the vehicle terminal through an encrypted communication channel. After receiving the automatic unlock command, the vehicle terminal directly executes the unlocking action. When the unlock confidence score is not greater than the high confidence threshold and not less than the low confidence threshold, the key terminal collects user behavior feature data through a motion sensor, calculates the similarity between the user behavior feature data and the standard behavior patterns in the pre-stored behavior pattern library, and when the similarity is greater than the preset behavior threshold, the key terminal generates a conditional unlock command and sends it to the vehicle terminal. When the unlock confidence score is less than the low confidence threshold, the key terminal triggers a security protection process, suspends receiving Bluetooth broadcast data packets broadcast by the vehicle terminal, and encapsulates the unlock confidence score, the current RSSI value, the CS ranging distance value, and the trigger time as an abnormal event record and stores them in the local storage unit.

5. The method according to claim 1, characterized in that, After the step of executing a tiered unlocking strategy based on the numerical range of the unlocking confidence score at the key end, the method further includes: The key terminal calculates a time interval feature value based on the door lock status and the last unlock timestamp in the vehicle status information, and retrieves a scene mode identifier that matches the time interval feature value from the pre-stored scene mode library; The key terminal calculates the scene-corrected unlock confidence score based on the scene gain coefficient corresponding to the scene mode identifier and the unlock confidence score; The key terminal records the scene-corrected unlock confidence score, the environment identification code, and the vehicle status information together in the historical data cache area.

6. The method according to claim 1, characterized in that, After the step of executing a tiered unlocking strategy based on the numerical range of the unlocking confidence score at the key end, the method further includes: The key end detects the actual unlocking result after executing the unlocking command. When unlocking fails, the high confidence threshold is increased; when unlocking succeeds, the high confidence threshold is decreased. After each unlocking operation, the key resets the data in the time-series data acquisition window and clears the current RSSI value sequence.

7. The method according to claim 6, characterized in that, After the step of executing a tiered unlocking strategy based on the numerical range of the unlocking confidence score at the key end, the method further includes: The key terminal counts the number of consecutive unlocking failures. When the number of consecutive unlocking failures reaches a preset failure threshold, the reception of the Bluetooth broadcast data packets is paused for a preset time period. After the preset time period ends, the key terminal initializes the high confidence threshold and the low confidence threshold to default values.

8. A Bluetooth key control system, characterized in that, The Bluetooth key control system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the Bluetooth key control system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the Bluetooth key control system, the Bluetooth key control system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the Bluetooth key control system, it causes the Bluetooth key control system to perform the method as described in any one of claims 1-7.

Images