Feature matching method and system based on multi-source heterogeneous data of intelligent terminal

Abstract

The application provides a feature matching method and system based on multi-source heterogeneous data of intelligent terminals, and relates to the technical field of wireless communication. The method comprises the following steps: a target terminal acquires wireless network measurement records and motion state records, and respectively associates time markers; an environment fingerprint formed by at least two types of statistical quantities is generated based on the wireless network measurement records, which is used to indicate the wireless scene characteristics of the observation position; within a continuous time window, a continuity parameter is generated based on the motion state records to represent the motion continuity between records; data objects carrying time markers and associated with the environment fingerprint and the continuity parameter are combined in pairs to form candidate object pairs, and the candidate object pairs come from different modalities or different terminals; based on the environment fingerprint and the continuity parameter, a matching determination is made according to the time markers, and a target matching pair satisfying a predetermined determination rule is output; and the application can improve the accuracy and stability of cross-source data matching in a dynamic indoor scene.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and more specifically, to a feature matching method and system based on multi-source heterogeneous data from smart terminals. Background Technology

[0002] With the widespread adoption of smart terminals and the development of wireless communication technology, methods for location determination and data matching based on wireless signal characteristics have been extensively researched and applied in indoor environments. Existing technologies typically utilize single-standard wireless measurement data, such as the received signal strength fingerprint of Wi-Fi access points or the power distribution of Bluetooth Low Energy broadcast signals. By constructing a feature fingerprint database and comparing similarity during matching, cross-terminal data association and location matching can be achieved. Simultaneously, to avoid complete misalignment in the timing of cross-terminal data collection, existing solutions generally introduce timestamp filtering or window constraints to perform coarse-grained screening of the compared data. However, in complex and dynamic indoor scenarios, due to the random movement of people, frequent opening and closing of physical obstructions such as doors, and differences in the collection rhythm and accuracy of different terminals, existing technologies often rely on a single environmental similarity criterion. This can easily lead to mismatches or missed matches when environmental features are highly similar or when short-term disturbances are frequent.

[0003] Given the above situation, existing technologies still have limitations when processing cross-source heterogeneous data. On the one hand, the singular expression of environmental features cannot effectively distinguish data that are similar in time and space but belong to different entities, resulting in insufficient sufficiency in matching judgments. On the other hand, the time alignment problem caused by asynchronous collection across terminals has not been fully resolved, causing some data to be spatially consistent but temporally incomparable, thereby weakening the stability of matching results.

[0004] Therefore, in dynamic indoor scenarios, facing short-term environmental changes and asynchronous cross-source data acquisition, how to improve the sufficiency and temporal consistency of the criteria for cross-source record matching, so as to reduce mismatches caused by a single environmental similarity criterion and enhance matching stability, has become a technical problem to be solved. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a feature matching method and system based on multi-source heterogeneous data from smart terminals.

[0006] Firstly, this application provides a feature matching method based on multi-source heterogeneous data from smart terminals, including:

[0007] The target terminal acquires wireless network measurement records and motion status records, and associates them with time stamps.

[0008] Based on the wireless network measurement records, an environmental fingerprint is generated to indicate the wireless scene characteristics at the observation location. The environmental fingerprint is formed by combining at least two types of wireless measurement statistics.

[0009] Within a continuous time window defined by the time marker, a continuity parameter reflecting the continuity of motion between records is generated based on the motion state record;

[0010] Data objects carrying time stamps and associated with the environmental fingerprint and the continuity parameter, extracted from the wireless network measurement records and / or the motion state records, and whose time stamps satisfy the correspondence relationship within the same continuous time window, are combined in pairs to form candidate object pairs. The two data objects of the candidate object pair come from different modalities or different terminals.

[0011] For the candidate object pairs, based on the environmental fingerprint and the continuity parameter, a matching judgment is performed according to the time stamp, and the candidate object pairs that meet the predetermined judgment rules are determined as target matching pairs and output.

[0012] Optionally, generating an environmental fingerprint for indicating wireless scene characteristics at the observation location includes:

[0013] Within a continuous time window defined by the time stamp, measurements of different wireless standards are time-aligned and aggregated to form a slow variable summary that characterizes the resource configuration in the frequency and time domains, and the slow variable summary is incorporated into the environmental fingerprint.

[0014] The slow variable summary is used to characterize the resource reuse pattern, including at least one of the following: orthogonal frequency division multiple access resource unit occupancy distribution, basic service set color occurrence rate distribution, and time division duplex cycle pattern summary.

[0015] Optionally, generating an environmental fingerprint for indicating wireless scene characteristics at the observation location includes:

[0016] Within a continuous time window defined by the time marker, short-term multipath disturbance indication information is determined based on the wireless network measurement records according to a preset disturbance criterion. The short-term multipath disturbance indication information is used to identify samples or segments that belong to short-term multipath disturbances.

[0017] Based on short-term multipath disturbance indication information, robust statistical processing is performed to generate a robust statistical feature set from the wireless network measurement records, and the robust statistical feature set is recorded as part of the environmental fingerprint;

[0018] The robust statistical processing defines the processed sample or segment using the short-term multipath perturbation indication information and generates the robust statistical feature set according to robust statistical rules.

[0019] The robust statistical feature set is used to characterize the stable components within the continuous time window.

[0020] Optionally, the determination of short-time multipath disturbance indication information includes:

[0021] The continuous time window is scanned according to the segment generation rules to generate a segment set;

[0022] For each segment in the segment set, a wireless measurement sequence is extracted from the wireless network measurement record to form segment measurement data;

[0023] Based on the measurement data of the section, a general disturbance discrimination quantity is calculated and compared with a preset general disturbance criterion to obtain the disturbance discrimination result of the section;

[0024] For segments where the disturbance identification result indicates the presence of short-term multipath disturbance, short-term multipath disturbance indication information is generated. The short-term multipath disturbance indication information records at least the time location and segment identifier.

[0025] The short-term multipath disturbance indication information is provided to robust statistical processing to limit the sample or segment being processed.

[0026] Optionally, the method further includes:

[0027] Within the continuous time window defined by the time stamp, for each short-term multipath disturbance indication information, a wireless measurement sequence for type determination is extracted from the wireless network measurement record based on the segment identifier recorded in the short-term multipath disturbance indication information to form type determination data;

[0028] Based on the type determination data, the event discrimination value is calculated and compared with the preset event criteria to generate the disturbance type determination result;

[0029] Based on the disturbance type determination result, a disturbance type identifier is recorded in the short-term multipath disturbance indication information, wherein the disturbance type identifier is selected from pedestrian flow and / or door opening and closing;

[0030] Short-term multipath disturbance indication information with disturbance type identifiers is provided for robust statistical processing and environmental fingerprinting.

[0031] Optionally, the generation of continuity parameters reflecting the motion continuity between records includes:

[0032] Within a continuous time window defined by the time stamp, the disturbance type identifier recorded in the short-term multipath disturbance indication information is obtained, and event constraint information is generated accordingly. The event constraint information is used to indicate connection maintenance or hierarchical change during the continuity parameter generation process.

[0033] For the recorded motion state, attitude alignment and time normalization are performed to obtain a motion state sequence with equal time intervals;

[0034] Determine the set of stationary anchor points based on a preset stationary criterion and generate a set of motion segments including multiple segments;

[0035] Based on the event constraint information, the set of motion segments is corrected to obtain the corrected connection state information and hierarchical change information:

[0036] The corrected connection status information, hierarchy change information, and direction stability information are combined with the continuity mask to form the continuity parameter, and the continuity parameter is recorded for the matching determination.

[0037] Optionally, the modification includes:

[0038] When the disturbance type is identified as pedestrian flow, the connection between adjacent segments is maintained within the disturbance coverage area, and the impact of directional instability on continuity judgment is reduced;

[0039] When the disturbance type is identified as door opening or closing, hierarchical change information is generated within the disturbance coverage area, and connection interruption is allowed at the segment boundary.

[0040] Optionally, the modification includes:

[0041] Within a continuous time window defined by the time marker, Doppler reconstruction information is generated based on the wireless network measurement records. The Doppler reconstruction information is used to characterize the relative motion state and is obtained by aligning and de-drifting the phase evolution of the reference symbols and / or pilots within the resource unit over time.

[0042] Based on the Doppler reconstruction information, a Doppler continuity indicator is generated within the segment set. The Doppler continuity indicator is used to characterize the persistence, direction change, and amplitude level of Doppler.

[0043] The connection state information and continuity mask are adjusted according to the Doppler continuity indication, and the correction result is recorded in the adjusted connection state information and continuity mask.

[0044] Optionally, the adjustment includes:

[0045] When the Doppler continuity indicator represents a reversal of the Doppler direction and is accompanied by a change in the frequency band usage status and the color of the basic service set, a connection interruption occurs at the segment boundary.

[0046] When the amplitude of the Doppler is not higher than a preset amplitude threshold and the slow variable summary representation has a timed silence arrangement, the connection relationship is maintained in the corresponding segment and the silence label is recorded in the continuity mask;

[0047] When the amplitude of the Doppler is not lower than another preset amplitude threshold and the motion state record indicates that the motion is stationary, the direction change component in the Doppler reconstruction information is smoothed to obtain a corrected direction change index, and the connection relationship between adjacent segments is maintained under the corrected direction change index.

[0048] Secondly, this application provides a feature matching system based on multi-source heterogeneous data from smart terminals, including:

[0049] The acquisition module is used to obtain wireless network measurement records and motion status records from the target terminal and associate them with time stamps.

[0050] The calculation module generates an environmental fingerprint based on the wireless network measurement records to indicate the characteristics of the wireless scene at the observation location. The environmental fingerprint is formed by combining at least two types of wireless measurement statistics. Within a continuous time window defined by the time stamp, a continuity parameter is generated based on the motion state records to reflect the continuity of motion between records.

[0051] The combination module is used to combine two data objects that carry time stamps and are associated with the environmental fingerprint and the continuity parameter, are extracted from the wireless network measurement record and / or the motion state record, and whose time stamps satisfy the corresponding relationship within the same continuous time window, to form candidate object pairs. The two data objects of the candidate object pair come from different modalities or different terminals.

[0052] The matching module is used to perform matching judgment on the candidate object pairs based on the environmental fingerprint and the continuity parameter, according to the time stamp, and to determine the candidate object pairs that meet the predetermined judgment rules as target matching pairs and output them.

[0053] Compared with existing technologies, this application establishes a unified time reference for cross-modal data by simultaneously acquiring wireless network measurement records and motion state records on the smart terminal side and associating them with time stamps, thus avoiding matching errors caused by asynchronous acquisition in existing technologies. Furthermore, this application does not rely solely on a single wireless environment feature as the matching basis, but constructs an environmental fingerprint based on wireless measurement data, consisting of at least two types of statistical quantities, to characterize the multidimensional characteristics of the wireless scene in which the terminal is located. Simultaneously, it combines continuous parameters generated from motion state records to constrain the trajectory maintenance of the same entity within a continuous time window. Through this multi-source, multi-dimensional matching mechanism with temporal constraints, mismatches caused by similar environmental features or short-term disturbances can be effectively reduced in dynamic indoor scenes, improving the sufficiency of matching criteria. At the same time, by using time stamps to uniformly associate environmental fingerprints, continuous parameters, and candidate object pairs, temporal consistency of cross-terminal data is achieved, thereby enhancing the stability and robustness of the matching results and more reliably supporting subsequent feature-based matching applications, such as cross-terminal identity recognition, indoor navigation assistance, or multi-terminal data fusion. Attached Figure Description

[0054] Figure 1 A flowchart illustrating the feature matching method based on multi-source heterogeneous data from smart terminals provided in this application embodiment;

[0055] Figure 2 A flowchart illustrating a method for generating an environmental fingerprint to indicate wireless scene features at an observation location, provided in an embodiment of this application;

[0056] Figure 3 A flowchart illustrating another method for generating an environmental fingerprint to indicate wireless scene characteristics at an observation location, provided in an embodiment of this application;

[0057] Figure 4 This is a schematic diagram of a feature matching system based on multi-source heterogeneous data from a smart terminal, provided in an embodiment of this application. Detailed Implementation

[0058] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0059] In this embodiment, the target terminal first acquires wireless network measurement records and motion state records generated during operation, and adds time stamps to both types of records to achieve unified temporal constraints in subsequent processing. Then, for the wireless network measurement records, an environmental fingerprint is constructed using at least two different combinations of wireless measurement statistics to indicate the wireless scene characteristics at the observation location, thereby forming a distinctive feature representation in the spatial domain.

[0060] Based on this, the motion state records are analyzed within a continuous time window defined by time markers to generate continuity parameters, which are used to characterize the motion continuity and trajectory stability between adjacent records.

[0061] Next, data objects carrying time stamps and associated with the environmental fingerprint and continuity parameters are extracted from wireless network measurement records and / or motion state records and combined in pairs to form candidate object pairs, wherein the two data objects of the candidate object pairs come from different modalities or different terminals.

[0062] Finally, for the candidate object pairs, based on the joint judgment of environmental fingerprint and continuity parameters, and combined with time stamp constraints, the object pairs that meet the predetermined judgment rules are selected as target matching pairs and output, thus realizing accurate matching of cross-source heterogeneous data.

[0063] This application is applicable to various dynamic indoor or semi-open scenarios, such as hospital wards, waiting halls, or rehabilitation training centers. In these environments, there is a large flow of people and short-term disturbances in the physical environment, and traditional matching methods based on single signal similarity are prone to failure.

[0064] This application integrates wireless network features with motion state features and performs candidate matching determination under temporal constraints. This helps to solve the problem of unstable matching of cross-terminal and cross-modal data in dynamic environments, thereby providing a reliable foundation for subsequent cross-terminal identity recognition, indoor navigation assistance, and medical behavior data fusion.

[0065] See Figure 1 The flowchart shown is a feature matching method based on multi-source heterogeneous data from a smart terminal provided in an embodiment of this application, including steps S101 to S105, wherein:

[0066] S101: The target terminal acquires wireless network measurement records and motion status records, and associates them with time stamps respectively;

[0067] S102: Based on the wireless network measurement records, generate an environmental fingerprint to indicate the wireless scene characteristics of the observation location, wherein the environmental fingerprint is formed by a combination of at least two types of wireless measurement statistics;

[0068] S103: Within a continuous time window defined by the time marker, based on the motion state record, generate a continuity parameter that reflects the continuity of motion between records;

[0069] S104: Combine two data objects that carry time stamps and are associated with the environmental fingerprint and the continuity parameter, are extracted from the wireless network measurement record and / or the motion state record, and whose time stamps satisfy the corresponding relationship within the same continuous time window, to form candidate object pairs. The two data objects of the candidate object pair come from different modalities or different terminals.

[0070] S105: For the candidate object pairs, based on the environmental fingerprint and the continuity parameter, a matching judgment is performed according to the time stamp, and the candidate object pairs that meet the predetermined judgment rules are determined as target matching pairs and output.

[0071] Regarding the above S101:

[0072] In this embodiment, the target terminal can be a smart terminal with wireless communication capabilities and motion sensing capabilities, such as a smartphone, tablet computer, or wearable device. The target terminal periodically collects wireless network measurement records and motion status records through the interface provided by its operating system. The wireless network measurement records may include, but are not limited to, the Received Signal Strength Indicator (RSSI), Channel State Information (CSI), and Access Point Identifier (BSSID) of wireless local area network signals, as well as the Reference Received Power (RSRP), Reference Received Quality (RSRQ), and Signal-to-Noise Ratio (SINR) of cellular communication networks.

[0073] Motion status recordings can be acquired by built-in sensor modules, such as a three-axis accelerometer, gyroscope, magnetometer, or barometer, to characterize the target terminal's attitude, acceleration, rotational angular velocity, and altitude changes.

[0074] During the data acquisition process, the target terminal assigns a time stamp to each wireless network measurement record and motion status record. The time stamp is preferably generated by the system clock or a unified external clock source to ensure that the data of different modes are on a unified time reference.

[0075] The time stamp resolution can be at the millisecond level to ensure alignment accuracy in rapidly changing scenarios. For cases with inconsistent sampling frequencies, temporal alignment of data can be achieved through interpolation, nearest neighbor matching, or discarding outliers, thereby establishing a temporal correlation between wireless network measurement records and motion state records.

[0076] Through the above process, multi-source heterogeneous data carrying a unified time stamp are obtained.

[0077] Regarding S102 above:

[0078] In this embodiment, for the wireless network measurement records, the target terminal or the backend server can perform statistical processing on wireless measurement data of different standards within a preset continuous time window to generate an environmental fingerprint. Specifically, the environmental fingerprint is composed of at least two types of wireless measurement statistics to ensure sufficient distinguishability and robustness. The wireless measurement statistics may include, but are not limited to, the following: First, statistical features calculated based on Received Signal Strength Indicator (RSSI), such as mean, variance, quantiles, and stability indicators, to reflect the power distribution pattern at a spatial location; Second, frequency domain or subcarrier amplitude distribution histograms, phase difference sequences, or delay spread parameters extracted based on Channel State Information (CSI), to reflect the multipath characteristics of the channel; Third, statistics of macroscopic link quality indicators such as Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), and Signal-to-Noise Ratio (SINR), to reflect the overall link quality level; Fourth, slow variable indicators such as carrier occupancy rate, spectral energy distribution, or scheduling periodicity characteristics, to characterize the stability of frequency and time domain resource allocation.

[0079] When generating an environmental fingerprint, the statistical measures of different categories mentioned above can be normalized or standardized and then concatenated into a set of feature vectors. For example, a simplified environmental fingerprint vector F can be represented as: F=[RSSI_mean, RSSI_var,CSI_amp_hist, RSRP_mean, BSS_color_dist], where each term represents the RSSI mean, RSSI variance, CSI amplitude histogram, RSRP mean, and BSS color distribution, respectively. Furthermore, dimensionality reduction methods, such as Principal Component Analysis (PCA), hash mapping, or subspace projection, can be combined to obtain a low-dimensional fingerprint representation that retains the main features.

[0080] Optionally, during environmental fingerprint generation, an interference mask is used to suppress contaminated measurement components to prevent short-term anomalies from corrupting the overall fingerprint. The environmental fingerprint obtained through this process can provide consistent spatial feature representations across different time windows and terminals, thereby improving the robustness of cross-source data matching and temporal consistency.

[0081] For example, in a typical dynamic indoor scenario like a hospital waiting hall, a target terminal, such as a patient's smartphone, collects wireless network measurement records within a continuous time window of several minutes. Since the hall contains multiple Wi-Fi access points and cellular base station signal coverage, the wireless network measurement records can simultaneously include Wi-Fi RSSI sequences, CSI subcarrier amplitude and phase distributions, and cellular network RSRP and RSRQ information. When generating the environmental fingerprint, the measurement results of Wi-Fi and cellular signals are first time-aligned and normalized. Then, the mean and standard deviation are extracted from the RSSI distribution as features characterizing spatial power intensity. Next, the subcarrier amplitude histogram and phase difference between adjacent subcarriers are extracted from the CSI as features characterizing local multipath effects. Simultaneously, the mean RSRP and RSRQ fluctuation in the cellular measurements are calculated to reflect the overall link quality.

[0082] In this way, at least two types of statistics are combined to form a multi-dimensional feature vector, which serves as the environmental fingerprint within the time window, indicating the wireless scene characteristics of the waiting hall location. This environmental fingerprint ensures the stability of scene representation across terminals even in situations with frequent personnel movement and temporary obstruction of some access points, providing a reliable basis for subsequent candidate pair matching.

[0083] Regarding S103:

[0084] In this embodiment, within a continuous time window defined by a time marker, the motion state records collected by the target terminal are processed to generate continuity parameters reflecting the continuity of motion between records. The motion state records can originate from sensors built into the terminal, such as a three-axis accelerometer, gyroscope, magnetometer, or barometer, reflecting information such as acceleration, angular velocity, attitude direction, and altitude changes. By performing time normalization processing on these multi-dimensional sensor data, a motion state sequence under a unified time reference can be obtained.

[0085] Subsequently, on this motion state sequence, attitude alignment is first performed to eliminate directional deviations caused by differences in terminal grip and sensor mounting orientation.

[0086] For example, by selecting a reference attitude, the acquired acceleration or angular velocity vectors can be rotated and aligned to ensure that the motion states of different time slices are in a consistent reference coordinate system. Attitude alignment can avoid spurious discontinuities caused by inconsistent orientations.

[0087] After attitude alignment, the system constructs continuity parameters based on the time-normalized motion state sequence. These continuity parameters can consist of multiple sub-indicators, such as directional stability information, velocity change trend, acceleration smoothness, and a set of stationary anchor points. The set of stationary anchor points is generated using preset stationary criteria; for example, when both the acceleration magnitude and angular velocity are below a threshold and remain below it for a certain duration, the corresponding moment is marked as a stationary anchor point. Stationary anchor points are not only used to distinguish between continuous motion segments and stationary segments, but also serve as reference points for segment division.

[0088] Based on this, the time window can be divided into several motion segments, with data within each segment reflecting a continuous trajectory. When constructing continuity parameters, the system calculates the connectivity between segments, such as determining whether adjacent segments maintain consistent direction and stable speed, and whether there are abrupt trajectory changes caused by environmental disturbances. For detected directional changes, corrections can be made in conjunction with environmental disturbance indication information, thereby improving the reliability of the parameters.

[0089] The generated continuity parameters are typically stored in the form of structured data, including connection state information, directional stability indicators, hierarchical change flags, and continuity masks. These continuity parameters are ultimately used in conjunction with environmental fingerprints to participate in the matching of subsequent candidate object pairs. In dynamic indoor scenarios, these continuity parameters can effectively compensate for the uncertainties caused by short-term fluctuations in environmental features, ensuring that the matching determination considers both spatial features and the continuity of time and motion, thereby improving the accuracy and robustness of cross-source data matching.

[0090] Regarding S104 above:

[0091] In this embodiment, to establish comparable basic units between multi-source data from different modalities or different terminals, data objects are first extracted from the aforementioned dataset with added time stamps. The specific composition of the data object corresponds to its source: when the data object originates from wireless network measurement records, its core content is the associated environmental fingerprint; when the data object originates from motion state records, its core content is the associated continuity parameters. Each data object contains a time stamp and source information, such as terminal identifier and modality type, to indicate its state characteristics at a specific point in time.

[0092] When forming candidate object pairs, the system aggregates data objects within the same continuous time window into a candidate object set based on a unified time stamp benchmark. For cases with differences in sampling frequency or time offsets, corrections can be made through interpolation, normalization, or boundary truncation to ensure that candidate object pairs are established based on strict temporal consistency constraints. Subsequently, pairwise combinations are performed from the candidate object set to generate candidate object pairs, with each candidate object pair consisting of data objects from different modalities or different terminals. This approach allows for the formation of two typical scenarios:

[0093] Cross-modal matching pairs: These candidate pairs consist of different modal data from the same terminal, such as a combination of wireless network environment fingerprints and motion continuity parameters from the same smartphone. The aim is to enhance the spatiotemporal decision-making capabilities of a single terminal through multi-source fusion, for example, by using motion states to mitigate short-term fluctuations in the wireless environment, or by correcting motion trajectory drift through environmental fingerprints.

[0094] Cross-terminal matching pairs: These candidate pairs consist of data from different terminals, such as a combination of the environmental fingerprint of terminal A and the continuity parameters of terminal B. Their purpose is to determine the spatiotemporal correlation between multiple terminals, such as identifying whether two users have jointly traversed a certain area, or determining whether a specific scene event has been observed simultaneously by multiple terminals.

[0095] To avoid computational burden caused by an excessive number of candidate pairs, this embodiment can introduce a pre-screening mechanism before combination. For example, data objects with signal strength below a preset threshold can be removed, or records lacking key sensor information can be excluded; alternatively, only data objects with high feature saliency and confidence levels exceeding a predetermined standard can be retained to improve the efficiency and reliability of subsequent matching calculations.

[0096] Through the above process, candidate object pairs that meet the timing consistency constraints and have cross-modal or cross-terminal characteristics are finally obtained.

[0097] Regarding the above S105:

[0098] In this embodiment, the matching determination of candidate object pairs is not merely a simple comparison between data objects, but rather combines multi-dimensional feature calculations and temporal constraints. Specifically, the system first confirms the temporal consistency of candidate object pairs under a unified time reference, ensuring that both objects are within the same continuous time window. This step can be achieved by comparing whether the difference between their timestamps is less than a preset threshold, such as setting an allowable deviation range of 50ms or 100ms.

[0099] In calculating the similarity of environmental fingerprints, this embodiment preferably employs a combination of multiple methods. For example, based on the distribution of Received Signal Strength Indication (RSSI), cosine similarity or Euclidean distance can be used to measure the similarity between two environmental fingerprints in a candidate pair. When Channel State Information (CSI) is present, the correlation coefficient of subcarrier amplitudes or the relative offset of phase curves can be calculated to determine the consistency of the channel environment between the two. These similarity values ​​are uniformly normalized in the determination module to ensure the comparability between different feature quantities.

[0100] For comparing continuous parameters, this embodiment can calculate them using a time-aligned sequence of motion states. For example, the dynamic time warping (DTW) method can be used to compare the morphological similarity of two motion state curves, or the root mean square value of the angle difference can be calculated based on the sequence of direction changes to quantify the consistency of the trajectories. If a long period of stillness is detected, anchor point matching can be used to verify whether the stillness intervals of the two objects overlap.

[0101] The judgment module weights and fuses the environmental fingerprint similarity and continuity parameter comparison results to generate a comprehensive discriminant. The weighting coefficients can be configured according to the needs of different scenarios. For example, in complex indoor scenarios, the weight of continuity parameters can be increased, while in more open spaces, the weight of environmental fingerprints can be increased. The comprehensive discriminant is compared with preset judgment rules. These rules can use a single threshold, such as a value greater than 0.8 indicating a match, or a hierarchical strategy, such as a two-level "coarse-fine" mode, where a more lenient threshold is used for rapid screening first, followed by stricter rules for fine-tuning.

[0102] Furthermore, to enhance robustness, this embodiment incorporates the aforementioned interference mask and perturbation type identifier during the determination process. For example, if the environmental fingerprint of a candidate pair contains a frequency band component marked as interfered, this component is automatically ignored when calculating similarity; if the motion state record shows a "people passing through" perturbation, the determination rule for the continuity parameter will relax its sensitivity to directional jitter. This determination method, combined with contextual features, enables the matching process to maintain higher accuracy and stability in dynamic environments.

[0103] When a candidate pair is confirmed as the target match, the system will output the matching result along with a timestamp, terminal identifier, and corresponding matching confidence index, so that it can be used in subsequent application scenarios such as cross-terminal identity recognition, indoor navigation assistance, or medical behavior data fusion.

[0104] For example, in a hospital waiting hall, there is a large flow of patients and their companions, and similar co-occurrence phenomena frequently occur between different devices in time and space. Taking a patient's smart bracelet and their smartphone as an example, both independently collect wireless network measurement records and motion status records. When the patient walks from a seat to the registration window in the hall, both the bracelet and the phone will generate corresponding data objects during that time period: the phone constructs an environmental fingerprint through the access point's RSSI and CSI, while the bracelet records and generates continuous parameters through the accelerometer and gyroscope.

[0105] During the matching process for candidate pairs, the system first detects that the difference between the two time stamps is less than a set threshold, such as 50ms, thus meeting the timing consistency requirement. Subsequently, by calculating the similarity between the environmental fingerprints of the phone and the bracelet (e.g., a cosine similarity of 0.85), and combining the continuity parameters generated from the bracelet's motion state and the phone's inertial sensor data, a high consistency in their trajectory patterns is obtained. When the overall discriminant value exceeds 0.8, the system confirms the pair as the target match.

[0106] This determination not only indicates that the wristband and mobile phone belong to the same user, but also demonstrates high stability even in complex situations involving the presence of other people or signal interference. Furthermore, this pairing information can be output to the hospital's navigation system, enabling precise patient location and route tracking, and assisting medical staff in providing timely services.

[0107] Optional, see Figure 2 The flowchart of a method for generating an environmental fingerprint for indicating wireless scene features at an observation location, provided in an embodiment of this application, includes steps S201 to S202, wherein:

[0108] S201: Within a continuous time window defined by the time stamp, time alignment and aggregation are performed on measurements of different wireless standards to form a slow variable summary for characterizing the frequency domain and time domain resource configuration, and the slow variable summary is incorporated into the environmental fingerprint.

[0109] S202: The slow variable summary is used to characterize the resource reuse pattern, including at least one of the following: orthogonal frequency division multiple access resource unit occupancy distribution, basic service set color occurrence rate distribution, and time division duplex cycle pattern summary.

[0110] In this embodiment, to enhance the stability and distinguishability of environmental fingerprints, the target terminal or backend server performs time alignment and aggregation on measurement data from different wireless standards within a continuous time window, generating a slow variable summary to characterize the frequency and time domain resource configuration. Slow variables refer to features that remain stable over a longer timescale and do not fluctuate drastically with instantaneous interference. They can effectively compensate for the randomness of rapidly changing features, such as RSSI and CSI, in dynamic scenarios.

[0111] Specifically, slow variable summaries can include the following types of information: First, the occupancy distribution of Orthogonal Frequency Division Multiple Access (OFDMA) resource units. Terminals can obtain the number and proportion of each subcarrier resource block scheduled within a preset time window by parsing the management frame or trigger frame of the wireless local area network, so as to reflect the resource reuse pattern of the area. Second, the color occurrence rate distribution of the Basic Service Set (BSS). For example, in Wi-Fi 6 or above, the frequency and proportion of different BSS color identifiers appearing over a period of time can be statistically analyzed to characterize the spatial reuse between access points. Third, the time division duplex (TDD) periodic pattern summary. Terminals can obtain the configuration rules of uplink and downlink time slots by parsing the system information block of the cellular communication system, such as SIB1 in 5G NR, and extract its periodic pattern as a summary to characterize the time domain configuration features of the macro link.

[0112] During the generation process, the system first performs time alignment on the measurement data of Wi-Fi and cellular networks to ensure that the measurement results of different standards fall within a unified time window. Subsequently, the above data is normalized and aggregated using statistical analysis methods.

[0113] For example, RSSI sequences can be used to help locate the switching boundary of TDD cycles, while the distribution of BSS color is statistically analyzed using histograms to form a probability vector. Ultimately, these statistical results are combined to form a slow variable summary.

[0114] The generated slow variable summary can be directly incorporated into the environmental fingerprint as an important component, participating in subsequent matching decisions along with other fast-changing features, such as the RSSI mean and CSI amplitude distribution. In this way, even under short-term disturbances such as pedestrian traffic or door opening and closing, the environmental fingerprint can maintain stable discriminative power, thereby significantly improving the robustness and accuracy of cross-source data matching.

[0115] For example, in a typical waiting room scenario, although patient movement may cause momentary fluctuations in RSSI, the TDD cycle pattern remains relatively constant within this time window, and the BSS Color occurrence rate remains relatively stable over tens of seconds. By incorporating these slow variables into the environmental fingerprint, misjudgments caused by momentary occlusion or channel fluctuations can be avoided, providing reliable support for cross-terminal matching.

[0116] Optional, see Figure 3 A flowchart of another method for generating an environmental fingerprint for indicating wireless scene features at an observation location, provided in an embodiment of this application, includes:

[0117] S301: Within the continuous time window defined by the time marker, short-time multipath disturbance indication information is determined based on the wireless network measurement records according to a preset disturbance criterion. The short-time multipath disturbance indication information is used to identify samples or segments belonging to short-time multipath disturbances.

[0118] S302: Based on short-time multipath disturbance indication information, perform robust statistical processing to generate a robust statistical feature set from the wireless network measurement records, and record the robust statistical feature set as part of the environmental fingerprint;

[0119] S303: The robust statistical processing defines the processed sample or segment with the short-time multipath perturbation indication information, and generates the robust statistical feature set according to robust statistical rules;

[0120] S304: The robust statistical feature set is used to characterize the stable components within the continuous time window.

[0121] In this embodiment, to avoid interference from short-term multipath disturbances in the environmental fingerprint generation process, the system performs disturbance detection on the wireless network measurement records within a continuous time window defined by time stamps. Specifically, the system first sets a set of preset disturbance criteria, which may include an instantaneous rate of change of the Received Signal Strength Indicator (RSSI) exceeding a set threshold, a sharp increase in the subcarrier phase variance in the Channel State Information (CSI), or an abnormal increase in the power delay curve (PDP) width within a short period of time. When the detection result meets any of the above criteria, a short-term multipath disturbance can be identified.

[0122] For the detected disturbance segments, the system generates short-term multipath disturbance indication information. This indication information includes at least the start and end times of the disturbance segment, the segment identifier, and the intensity level or category of the disturbance.

[0123] For example, if the RSSI curve experiences a sudden drop of more than 6dB within a 200ms time slice, the system will generate a short-term multipath disturbance indication message, mark the time slice as a disturbance segment, and record its start and end times.

[0124] After acquiring disturbance indication information, the system further performs robust statistical processing on the wireless network measurement records. Robust statistical processing refers to performing statistical calculations only on data samples that have not been marked for disturbance, in order to eliminate or mitigate the impact of short-term anomalies on the overall results. Robust statistical methods may include median filtering, quantile statistics, and weighted moving averages.

[0125] For example, within a 1-second time window, if a 0.2-second time slice is marked as a disturbance, then only the median and interquartile range of the CSI amplitude data within the remaining 0.8 seconds are calculated to obtain robust statistical characteristics.

[0126] Ultimately, the robust statistical feature set generated by the system includes, but is not limited to, the RSSI median and variance, the quantile curves of the CSI subcarrier energy distribution, the robust mean of RSRP, and the range of SINR. Unlike conventional simple means or global variance, these robust features can better characterize stable components within a continuous time window, avoiding feature distortion caused by short-term anomalous disturbances. The generated robust statistical feature set will be recorded and stored as part of the environmental fingerprint to support subsequent matching decisions.

[0127] For example, in a hospital corridor scenario, when a nurse pushes open a ward door and passes by, the CSI signal collected by the patient's mobile phone may experience sudden multipath interference. If this interference is not addressed and the CSI phase or amplitude characteristics are directly analyzed, the environmental fingerprint may deviate from the actual location. Through the disturbance detection and robust statistical processing in this embodiment, the system can identify and remove this disturbed sample, generating an environmental fingerprint only based on the remaining stable samples, thereby ensuring that the fingerprint still accurately reflects the real wireless scene characteristics of the corridor.

[0128] Optionally, the determination of short-time multipath disturbance indication information includes:

[0129] The continuous time window is scanned according to the segment generation rules to generate a segment set;

[0130] For each segment in the segment set, a wireless measurement sequence is extracted from the wireless network measurement record to form segment measurement data;

[0131] Based on the measurement data of the section, a general disturbance discrimination quantity is calculated and compared with a preset general disturbance criterion to obtain the disturbance discrimination result of the section;

[0132] For segments where the disturbance identification result indicates the presence of short-term multipath disturbance, short-term multipath disturbance indication information is generated. The short-term multipath disturbance indication information records at least the time location and segment identifier.

[0133] The short-term multipath disturbance indication information is provided to robust statistical processing to limit the sample or segment being processed.

[0134] In this embodiment, in order to effectively identify short-term multipath disturbances within a continuous time window, the system first divides the time axis according to a preset segmentation rule to obtain several segment sets. The segmentation rule can be a fixed duration segmentation, such as every 100 milliseconds as a segment, or it can adopt a sliding window method, or trigger dynamic segmentation when a sudden change in signal energy is detected, thereby ensuring that different disturbance patterns can be covered.

[0135] For the aforementioned set of segments, the system extracts the corresponding wireless measurement sequence from each segment as segment measurement data. The wireless measurement sequence may include a sequence of Received Signal Strength Indication (RSSI) changing over time, a Channel State Information (CSI) subcarrier amplitude and phase sequence, or the Reference Received Power (RSRP) and Signal-to-Noise Ratio (SINR) curves in a cellular communication network. This ensures that the measurement data within each segment comprehensively reflects the wireless propagation characteristics of that time slot.

[0136] Subsequently, the system calculates a general disturbance discrimination quantity based on the measurement data of the aforementioned segment. The type of discrimination quantity can be selected according to the specific scenario. For example, it can be the variance or root mean square of the RSSI sequence to measure the power fluctuation amplitude; it can also be the standard deviation of the CSI phase curve or the attenuation rate of the autocorrelation function to characterize the intensity of multipath effects; or it can be the entropy value of the spectral energy distribution to identify the complexity of channel occupancy. The calculated discrimination quantity is compared with a preset general disturbance criterion. When the discrimination quantity exceeds a threshold, it is determined that a short-term multipath disturbance exists in the segment.

[0137] For segments identified as having short-term multipath disturbances, the system generates corresponding short-term multipath disturbance indication information. This indication information includes at least the segment's time location, segment identifier, and disturbance level flag, which serves as the basis for location and correction in subsequent processing. The generated short-term multipath disturbance indication information is input into the robust statistical processing module to limit samples or segments requiring special processing, such as weighted weakening, anomaly removal, or substitution interpolation.

[0138] For example, in a dynamic scenario in a hospital waiting hall, a patient's smartphone continuously collects CSI subcarrier phase data within a few-second time window. The system first divides the time window into multiple 200-millisecond segments and extracts the CSI phase sequence within each segment. For a given segment, the system calculates the standard deviation of the CSI phase and finds that it is significantly greater than a preset threshold of 0.5 radians, and the autocorrelation function decays sharply within a short delay, indicating that there is a short-term multipath disturbance in this segment.

[0139] At this point, the system generates corresponding short-term multipath disturbance indication information, recording the start and end times, identifier, and disturbance level of the segment. This indication information is used in subsequent robust statistical processing to suppress the weight of the disturbed segment in the environmental fingerprint, thereby avoiding misleading the overall matching results due to drastic signal fluctuations caused by instantaneous crowd movement.

[0140] Optional, also includes:

[0141] Within the continuous time window defined by the time stamp, for each short-term multipath disturbance indication information, a wireless measurement sequence for type determination is extracted from the wireless network measurement record based on the segment identifier recorded in the short-term multipath disturbance indication information to form type determination data;

[0142] Based on the type determination data, the event discrimination value is calculated and compared with the preset event criteria to generate the disturbance type determination result;

[0143] Based on the disturbance type determination result, a disturbance type identifier is recorded in the short-term multipath disturbance indication information, wherein the disturbance type identifier is selected from pedestrian flow and / or door opening and closing;

[0144] Short-term multipath disturbance indication information with disturbance type identifiers is provided for robust statistical processing and environmental fingerprinting.

[0145] In this embodiment, to further improve the precision of short-term multipath perturbation processing, after detecting the short-term multipath perturbation, the system also determines the perturbation type so that targeted corrections can be made in subsequent robust statistics and environmental fingerprint construction. Specifically, this process includes the following steps:

[0146] First, within a continuous time window defined by time stamps, the system extracts a complete measurement sequence from the corresponding wireless network measurement records for each short-term multipath disturbance indication, based on its recorded segment identifier, to form type determination data. This type determination data is typically represented as a continuous time series, such as RSSI curves, CSI amplitude and phase trajectories, or a signal-to-noise ratio (SNR) variation sequence over time.

[0147] Subsequently, the system calculates the event discriminant based on the type determination data. The selection of the discriminant can be designed according to the characteristics of different disturbance types:

[0148] When the candidate perturbation is "people passing through", indicators such as RSSI decrease magnitude and recovery rate, and CSI coherence loss interval length can be used as discriminants to characterize the speed and locality of short-term occlusion.

[0149] When the candidate perturbation is "gate opening and closing", indicators such as the amplitude of CSI subcarrier energy mutation and the step change of channel delay spread parameters can be selected as discriminants to characterize the change in the overall propagation path structure.

[0150] After the discriminant is calculated, the system compares it with preset event criteria. For example, if the RSSI drops below a set threshold within hundreds of milliseconds and recovers quickly, and the CSI subcarrier coherence is lost, the disturbance type is determined to be "people passing through"; if the CSI amplitude and delay spread parameters show a step change at a certain point in time, and this change can continue into the next time period, the disturbance type is determined to be "door opening and closing".

[0151] After the determination is completed, the system records the disturbance type identifier in the corresponding short-term multipath disturbance indication information, such as marked as "human_flow" or "door_event". This type identifier is then passed to the robust statistical processing module for dynamically adjusting statistical rules. For example, it relaxes the sensitivity to directional jitter in scenarios of pedestrian flow, or allows segment connection breaks in scenarios of door opening and closing. At the same time, this type identifier is also written into the environmental fingerprint for contextual annotation of dynamic environmental features, thereby enhancing the discriminative power of fingerprint matching across time and terminals.

[0152] For example, taking a hospital waiting hall as an example, during peak hours, patients and their companions move around frequently. The RSSI curve of the Wi-Fi signal collected by smartphones will show a short-term drop and recovery, and the CSI phase sequence will exhibit significant random fluctuations during this period. Based on this, the system determines the disturbance type as "human flow" and records "human_flow" in the indication information. In subsequent robust statistical processing, the environmental fingerprint generation rules will automatically ignore this short-term occlusion segment, thus ensuring the stability of the overall features.

[0153] In another scenario, when a patient enters the examination room, a consistent step change occurs in the CSI subcarrier energy distribution, and the channel delay spread parameter changes simultaneously. Based on this, the system determines the disturbance type as "door opening / closing" and records "door_event" in the indication information. In subsequent processing, this identifier prompts the system to allow a connection interruption at the segment boundary, thereby avoiding the erroneous interpretation of two physically isolated propagation states as consecutive states.

[0154] Through the above embodiments, this application can achieve fine classification and context labeling of short-term multipath disturbances, effectively improving the robustness and accuracy of cross-source heterogeneous data matching in dynamic scenarios.

[0155] Optionally, the event criterion can be a pre-trained machine learning classification model, such as a support vector machine (SVM), decision tree, or convolutional neural network (CNN). During offline training, a large number of labeled CSI samples of pedestrian traffic and door opening / closing are collected, and the aforementioned event discriminant parameters are extracted as features to train the model. During online determination, the discriminant parameters extracted from unknown disturbance sections are input into the model, which directly outputs the disturbance type determination result. This approach can further improve classification accuracy and adaptability to complex scenarios.

[0156] Optionally, the generation of continuity parameters reflecting the motion continuity between records includes:

[0157] Within a continuous time window defined by the time stamp, the disturbance type identifier recorded in the short-term multipath disturbance indication information is obtained, and event constraint information is generated accordingly. The event constraint information is used to indicate connection maintenance or hierarchical change during the continuity parameter generation process.

[0158] For the recorded motion state, attitude alignment and time normalization are performed to obtain a motion state sequence with equal time intervals;

[0159] Determine the set of stationary anchor points based on a preset stationary criterion and generate a set of motion segments including multiple segments;

[0160] Based on the event constraint information, the set of motion segments is corrected to obtain the corrected connection state information and hierarchical change information:

[0161] The corrected connection status information, hierarchy change information, and direction stability information are combined with the continuity mask to form the continuity parameter, and the continuity parameter is recorded for the matching determination.

[0162] In this embodiment, within a continuous time window defined by time stamps, the system performs a series of processing steps on the acquired motion state records and short-term multipath disturbance indication information to generate continuity parameters that reflect the motion continuity between records. The process includes the following stages:

[0163] Step 1: Obtain the disturbance type and generate event constraint information:

[0164] The system first reads the disturbance type identifier recorded in the short-term multipath disturbance indication information, such as "pedestrian passage" or "door opening / closing". Based on different disturbance types, it generates corresponding event constraint information.

[0165] If marked as "people passing through", the event constraint information will indicate that the connection relationship should be maintained in this section to avoid false trajectory breaks caused by crowd obstruction;

[0166] If the identifier is "door opening / closing", the event constraint information will indicate that a hierarchy change may have occurred, such as switching from an outdoor corridor to an indoor room.

[0167] Step 2: Pose Alignment and Time Normalization:

[0168] The system performs attitude alignment and time normalization on the motion state recordings. By performing rotational correction on the acceleration and angular velocity signals, deviations caused by different terminal grip directions are eliminated; and records with different sampling frequencies are resampled to a uniform time step, such as 20ms, to form a motion state sequence with equal time intervals.

[0169] Step 3: Identify stationary anchor points and divide the movement sections:

[0170] In a time-normalized sequence of motion states, the system identifies stationary anchor points based on preset stationary criteria. For example, when both the acceleration magnitude and angular velocity are below a threshold and remain below it for a certain duration, it is determined to be a stationary point. Using the stationary anchor points as boundaries, the system divides the entire time window into multiple motion segments, each segment corresponding to a continuous motion behavior.

[0171] Step 4: Correct the set of motion segments based on event constraint information:

[0172] After obtaining the initial set of motion segments, the system corrects them based on event constraint information:

[0173] If the event constraint information indicates "people passing through", then the connection between adjacent segments will be maintained within the coverage area of ​​the disturbance, and will not be broken due to brief attitude fluctuations;

[0174] If the event constraint information indicates "door opening / closing", then hierarchical change information is allowed to be generated at the segment boundary to mark entry into or exit from a certain spatial area.

[0175] Step 5: Generate and record continuity parameters:

[0176] Finally, the system combines the corrected connection status information, hierarchy change information, directional stability information, and continuity mask to form a complete set of continuity parameters. This set can comprehensively reflect the motion continuity characteristics of the target terminal within the time window and serves as an important input for subsequent matching decisions.

[0177] In a hospital waiting hall scenario, patients' smartphones and wristbands collect wireless network measurement records and motion status records, respectively. As a patient walks from the hall to the examination room, the acceleration signal recorded by the wristband shows brief fluctuations, while the Wi-Fi signal fingerprint shows localized changes. The system determines that this fluctuation is due to "person movement," and therefore the generated event constraint information requires maintaining segment connectivity. Thus, even with brief instability in the acceleration signal, the system still identifies the patient's movement trajectory as a continuous movement process.

[0178] Conversely, when a patient enters the consultation room, the opening and closing of the door is recognized as "door opening and closing." The system generates hierarchical change information at the boundary of this segment, thereby marking the user's trajectory as switching from the "waiting hall level" to the "consultation room level." The resulting continuity parameters accurately reflect the patient's complete path and avoid erroneous breaks caused by interference or door movements.

[0179] Optionally, the modification includes:

[0180] When the disturbance type is identified as pedestrian flow, the connection between adjacent segments is maintained within the disturbance coverage area, and the impact of directional instability on continuity judgment is reduced;

[0181] When the disturbance type is identified as door opening or closing, hierarchical change information is generated within the disturbance coverage area, and connection interruption is allowed at the segment boundary.

[0182] In this embodiment, based on the aforementioned generated short-term multipath disturbance indication information and its recorded disturbance type identifier, the system further differentiates the connection status of motion segments. Specifically, it first traverses the set of motion segments within each continuous time window and reads the corresponding disturbance type identifier. If the determination result is "people passing through," the system will maintain the connection relationship between adjacent segments within the disturbance coverage area. To avoid erroneous breakage determinations caused by instantaneous directional jitter due to people blocking the path, the system reduces the weight of the directional instability index in the continuity judgment, so that the trajectory can maintain overall continuity even under short-term disturbances.

[0183] If the determination result is "door opening / closing", the system will generate hierarchical change information within the disturbance coverage area to reflect the transition of the spatial environment. For example, when the opening and closing of the door causes signal fading and trajectory abrupt change, the system will insert a hierarchical breakpoint at the boundary of the corresponding segment and record the breakpoint and its upper and lower hierarchical information in the continuity parameter, thereby allowing the motion trajectory to generate a reasonable connection interruption at this position.

[0184] For example, in a hospital waiting hall, when patients walk along a corridor wearing smartphones and smart bracelets, they may encounter short periods of crowds passing by. At this time, RSSI fluctuations and CSI phase disturbances appear in the environmental fingerprint, and the motion record also shows directional jitter. The system determines the type of "crowd passing" disturbance and reduces the weight of directional instability indicators in the corresponding sections, thereby maintaining the continuity of the walking trajectory and preventing the trajectory from being misjudged as interrupted due to crowd obstruction.

[0185] In another scenario, a patient enters a consultation room from a waiting hall; the opening and closing of the door causes significant signal fading and path interruption. After identifying the "door opening and closing" disturbance type, the system inserts hierarchical change information at the corresponding location, recording the hall and consultation room as different spatial hierarchies, and allowing trajectory breakpoints at segment boundaries. This enables the differentiation of trajectory segments from different spatial environments in subsequent data analysis, improving the rationality and accuracy of matching decisions in multi-room scenarios.

[0186] For example, if the determination result is "door opening / closing", the system will generate hierarchical change information within the disturbance coverage area to reflect the transition of the spatial environment. For instance, when the opening and closing of the door causes signal fading and trajectory abrupt change, the system will insert a hierarchical breakpoint at the corresponding segment boundary and record the breakpoint in the continuity parameter. For example, the hierarchical change information field can be updated from lobby_L1 to clinic_L1, or a Boolean flag for spatial crossing can be set to allow the movement trajectory to generate a reasonable connection interruption at this location.

[0187] Optionally, the modification includes:

[0188] Within a continuous time window defined by the time marker, Doppler reconstruction information is generated based on the wireless network measurement records. The Doppler reconstruction information is used to characterize the relative motion state and is obtained by aligning and de-drifting the phase evolution of the reference symbols and / or pilots within the resource unit over time.

[0189] Based on the Doppler reconstruction information, a Doppler continuity indicator is generated within the segment set. The Doppler continuity indicator is used to characterize the persistence, direction change, and amplitude level of Doppler.

[0190] The connection state information and continuity mask are adjusted according to the Doppler continuity indication, and the correction result is recorded in the adjusted connection state information and continuity mask.

[0191] In this embodiment, to further improve the accuracy of continuity parameters, the system generates Doppler reconstruction information based on wireless network measurement records within a continuous time window defined by time stamps. This information characterizes the relative motion state of the target terminal. Specifically, the system analyzes the phase evolution of reference symbols and / or pilot signals in the radio resource unit over time, employing drift removal and alignment methods to eliminate the influence of carrier frequency offset or hardware clock errors, thereby obtaining relatively stable Doppler reconstruction information. This information reflects the terminal's velocity trend and relative direction changes within the time window, effectively supplementing traditional calculations based on accelerometers or gyroscopes.

[0192] After acquiring Doppler reconstruction information, the system generates a Doppler continuity indicator within the segment set. This continuity indicator includes three aspects: first, the persistence of the Doppler effect, used to determine whether the motion state remains consistent over a long period; second, changes in direction, used to identify whether a turn or path adjustment has occurred; and third, the amplitude level, used to distinguish between different states such as stationary, low-speed motion, and fast motion. Through this continuity indicator, the system can grasp the user's motion trend on a more macroscopic temporal scale, avoiding misjudgments caused by short-term fluctuations.

[0193] After generating the Doppler continuity indication, the system corrects the existing connection state information and continuity mask accordingly. For example, when the continuity indication shows that the target terminal's movement trend remains stable, the connection markers for the corresponding segments can be strengthened in the continuity mask; when a significant change in direction is detected, a turning marker can be inserted into the connection state information so that subsequent matching decisions can distinguish different movement stages. The corrected results are recorded in the system storage as an important reference for subsequent target matching decisions.

[0194] For example, in a hospital corridor scenario, as a patient walks, the phase of the Wi-Fi pilot signal on their smartphone gradually accumulates a stable frequency shift over time. The system obtains Doppler reconstruction information through drift removal processing and identifies a constant positive frequency shift over several seconds, indicating that the patient is moving forward in a constant direction. During this process, the Doppler continuity indication generated by the system confirms the continuity of movement and reinforces the connection marker in the continuity mask. Subsequently, when the patient turns back at the end of the corridor, the pilot phase trend changes from a positive frequency shift to a negative frequency shift. Based on this, the system inserts a direction change marker into the connection status information, ensuring that subsequent matching decisions can identify the two different movement phases of "going back and forth." Thus, even in a dynamic crowd environment, the system can still accurately maintain the continuity and rationality of the user's trajectory.

[0195] Optionally, the adjustment includes:

[0196] When the Doppler continuity indicator represents a reversal of the Doppler direction and is accompanied by a change in the frequency band usage status and the color of the basic service set, a connection interruption occurs at the segment boundary.

[0197] When the amplitude of the Doppler is not higher than a preset amplitude threshold and the slow variable summary representation has a timed silence arrangement, the connection relationship is maintained in the corresponding segment and the silence label is recorded in the continuity mask;

[0198] When the amplitude of the Doppler is not lower than another preset amplitude threshold and the motion state record indicates that the motion is stationary, the direction change component in the Doppler reconstruction information is smoothed to obtain a corrected direction change index, and the connection relationship between adjacent segments is maintained under the corrected direction change index.

[0199] In practical implementation, within a continuous time window defined by time stamps, the system obtains three types of inputs from preceding steps: Doppler reconstruction information on phase-time evolution alignment and drift-free relative motion representation, Doppler continuity indicators for discretized descriptions of persistence, direction changes, and amplitude levels, and slow variable summaries and motion state records. The system first establishes a list of candidate adjustment points at the boundaries of the segment set and at inflection points of Doppler direction changes. Candidates include time location, segment identifier, direction sign, amplitude level, and corresponding slow variable segment summary.

[0200] For each candidate adjustment point in the list, the following processing is performed according to the triggering conditions and adjustment actions, and the results are synchronously written into the connection state information and continuity mask.

[0201] Disconnection rules for direction reversal accompanied by changes in resource reuse patterns:

[0202] The trigger condition is that the Doppler continuity indicator shows that the direction changes from "positive" to "negative" or from "negative" to "positive", and at the same time, the slow variable summary display in the neighborhood changes the band usage status and the color of the basic service set (BSS).

[0203] The adjustment action is to insert a breakpoint in the connection status table at the corresponding segment boundary, change the connection status from "hold" to "interrupted", and mark "connection interrupted" on the continuity mask. At the same time, the triggering reason is recorded as "direction reversal + resource / BSS change", including time location and boundary index, for subsequent review and tracing.

[0204] The results are recorded as updated connection status information, including state transitions before and after, trigger rule numbers, and evidence summaries (details of changes in direction symbols, resource unit usage distribution, and BSS color changes).

[0205] For maintenance rules with low amplitude and timed silent arrangements:

[0206] The triggering condition is that the amplitude of the Doppler continuity indication is not higher than the preset amplitude threshold, and the slow variable summary represents the existence of a timed silence arrangement within the same window, such as periodic gaps, sparse segments of pilot / reference symbols, or service scheduling silence.

[0207] The adjustment action is to maintain the connection relationship between adjacent segments and mark the time segment as "silent" on the continuity mask; in subsequent matching calculations, the weight of fast-changing components of environmental fingerprints, such as instantaneous RSSI / CSI fine-grained features, is reduced to avoid misjudgment caused by sparse samples during the silent period.

[0208] The results are recorded as "maintained" in the connection state information, with the start and end times of the silent period and a description of the silent source (silent indication from the slow variable summary) marked in the mask.

[0209] For smoothness maintenance rules involving high amplitude but static motion:

[0210] The triggering condition is that the amplitude of the Doppler continuity indication is not lower than another preset amplitude threshold, and the motion state record within the same time segment indicates that it is stationary, such as satisfying the stationary anchor point condition.

[0211] The adjustment process involves smoothing the directional change components in the Doppler reconstruction information. Preferred methods include sliding window midpoint, segmented robust smoothing, or jump suppression to obtain a corrected directional change index. When the corrected index does not reach the disconnection trigger threshold, the connection between adjacent segments is maintained, and "smoothing preservation" is marked on the continuity mask.

[0212] The result is recorded as the connection status remains unchanged; the mask records the time range of the smoothing process, the processing method identifier, and a summary of the corrected direction change indicators, which facilitates auditing and review.

[0213] Furthermore, if multiple triggering conditions are met simultaneously in the same neighborhood, the system executes them according to priority: disconnection rules with direction reversal and resource / BSS changes take precedence over smooth maintenance rules with high amplitude stillness, and smooth maintenance rules take precedence over silent maintenance rules. After all adjustments are completed, the system generates a "correction rule log," recording the triggering rule number, time position, segment identifier, input evidence summary, and changes in status and label before and after adjustment for each rule, and archiving it together with connection status information and continuity mask.

[0214] For example, when a patient turns back at the end of a corridor, the resource unit occupancy distribution and BSS color distribution of the Wi-Fi 6 network switch simultaneously. The system detects a Doppler direction reversal at the turning point, and the slow variable summary shows a change in resource and BSS colors, triggering a disconnection rule. The system writes "connection interrupted" and the triggering reason at the boundary of this segment, and subsequent matching determines that the two trajectories belong to different motion stages based on this.

[0215] For example, the lobby broadcasts a brief period of silence on the hour, with the slow variable summary exhibiting a periodic silence pattern, while the Doppler amplitude remains at a low level. The system maintains connectivity and marks "silent" in the continuity mask. Subsequent fingerprint matching reduces the weight of fast variable components to avoid similarity shifts caused by pilot sparsity.

[0216] For example, if a patient is seated and stationary, and nearby mobile devices or fans cause rapid phase perturbations, the Doppler amplitude may be high, but the motion state may be recorded as stationary. The system smooths the directional change component, resulting in a lower corrected directional change index, maintaining connectivity and labeling it "smoothed," thus avoiding misinterpreting a stationary state as a break.

[0217] In this way, the system constrains the consistency of Doppler evidence, slow variables of resource reuse, and motion state evidence within the same time frame, making the connection state information and continuity mask both sensitive and robust, providing reliable and traceable temporal semantic support for subsequent matching judgments based on environmental fingerprints and continuity parameters.

[0218] Based on the same inventive concept, this application also provides a feature matching system based on multi-source heterogeneous data of smart terminals, which corresponds to the feature matching method based on multi-source heterogeneous data of smart terminals. Since the principle of solving the problem by the system in this application is similar to the feature matching method based on multi-source heterogeneous data of smart terminals described above in this application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0219] Reference Figure 4 The diagram shown is a schematic of a feature matching system based on multi-source heterogeneous data from a smart terminal provided in an embodiment of this application. The system includes:

[0220] The acquisition module 10 is used to acquire wireless network measurement records and motion status records from the target terminal and associate them with time markers respectively;

[0221] The calculation module 20 generates an environmental fingerprint based on the wireless network measurement records to indicate the characteristics of the wireless scene at the observation location. The environmental fingerprint is formed by combining at least two types of wireless measurement statistics. Within a continuous time window defined by the time stamp, it generates a continuity parameter based on the motion state records to reflect the continuity of motion between records.

[0222] The combination module 30 is used to combine two data objects that carry time stamps and are associated with the environmental fingerprint and the continuity parameter, are extracted from the wireless network measurement record and / or the motion state record, and whose time stamps satisfy the corresponding relationship within the same continuous time window, to form candidate object pairs. The two data objects of the candidate object pair come from different modalities or different terminals.

[0223] The matching module 40 is used to perform matching judgment on the candidate object pairs based on the environmental fingerprint and the continuity parameter, according to the time stamp, and to determine the candidate object pairs that meet the predetermined judgment rules as target matching pairs and output them.

[0224] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. 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 or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A feature matching method based on multi-source heterogeneous data of a smart terminal, characterized in that, The method comprises: acquiring, by a target terminal, wireless network measurement records and motion state records, and associating time markers respectively; generating, based on the wireless network measurement records, an environment fingerprint for indicating wireless scene features at an observation position, the environment fingerprint being formed by at least two types of wireless measurement statistics; generating, based on the motion state records, a continuity parameter for reflecting motion continuity between records within a continuous time window defined by the time markers; combining, two by two, data objects carrying time markers and associated with the environment fingerprint and the continuity parameter, which are extracted from the wireless network measurement records and / or the motion state records and whose time markers satisfy a correspondence within the same continuous time window, to form candidate object pairs, the two data objects of each candidate object pair being derived from different modalities or different terminals respectively; for the candidate object pairs, performing matching determination based on the environment fingerprint and the continuity parameter according to the time markers, and determining the candidate object pairs satisfying a predetermined determination rule as target matching pairs and outputting the target matching pairs. 2.The method of claim 1, wherein, The generating of the environment fingerprint for indicating wireless scene features at the observation position comprises: within the continuous time window defined by the time markers, performing time alignment and aggregation on measurements of different wireless standards to form a slow variable summary for characterizing resource configuration conditions in the frequency domain and the time domain, and incorporating the slow variable summary into the environment fingerprint; the slow variable summary is used to characterize resource multiplexing patterns, including at least one of the following: orthogonal frequency division multiple access resource unit occupation distribution, basic service set color occurrence rate distribution, and time division duplex cycle mode summary. 3.The method of claim 2, wherein, The generating of the environment fingerprint for indicating wireless scene features at the observation position further comprises: within the continuous time window defined by the time markers, determining, based on the wireless network measurement records, short-term multipath disturbance indication information according to a preset disturbance criterion, the short-term multipath disturbance indication information being used to identify samples or sections belonging to short-term multipath disturbance; based on the short-term multipath disturbance indication information, performing robust statistical processing to generate a robust statistical feature set from the wireless network measurement records, and recording the robust statistical feature set as a component of the environment fingerprint; the robust statistical processing limits the samples or sections to be processed according to the short-term multipath disturbance indication information, and generates the robust statistical feature set according to a robust statistical rule; the robust statistical feature set is used to characterize stable components within the continuous time window. 4.The method of claim 3, wherein, The determining of the short-term multipath disturbance indication information comprises: performing time scanning on the continuous time window according to a section generation rule to generate a section set; extracting wireless measurement sequences from the wireless network measurement records for each section in the section set to form section measurement data; based on the section measurement data, calculating a general disturbance discriminant and comparing it with a preset general disturbance criterion to obtain a disturbance discrimination result of the section; for the section with the disturbance discrimination result indicating the existence of short-term multipath disturbance, generating short-term multipath disturbance indication information, the short-term multipath disturbance indication information at least recording a time position and a section identifier. The short-time multipath disturbance indication information is provided to robust statistical processing to limit the samples or segments processed. 5.The method of claim 4, wherein, Further comprising: Within the continuous time window defined by the time markers, for each short-time multipath disturbance indication information, a wireless measurement sequence for type determination is extracted from the wireless network measurement records according to the segment identification recorded by the short-time multipath disturbance indication information to form type determination data; Based on the type determination data, an event discriminant is calculated and compared with a preset event criterion to generate a disturbance type determination result; According to the disturbance type determination result, a disturbance type identification is recorded in the short-time multipath disturbance indication information, and the disturbance type identification is selected from people flow passing and / or door opening and closing; The short-time multipath disturbance indication information recorded with the disturbance type identification is provided to robust statistical processing and environmental fingerprint record usage. 6.The method of claim 5, wherein, The generation of the continuity parameter reflecting the continuity of the motion between records includes: Within the continuous time window defined by the time markers, the disturbance type identification recorded in the short-time multipath disturbance indication information is obtained, and event constraint information is generated therefrom, which is used to indicate connection maintenance or level change during the continuity parameter generation process; For the motion state record, posture alignment and time normalization are performed to obtain a motion state sequence with equal time intervals; According to a preset stationary criterion, a stationary anchor point set is determined and a motion segment set including multiple segments is generated; According to the event constraint information, the motion segment set is modified to obtain modified connection state information and level change information: The modified connection state information, level change information, direction stability information, and continuity mask are combined to form the continuity parameter, and the continuity parameter is recorded for the matching determination. 7.The method of claim 6, wherein, The modification includes: When the disturbance type identification is people flow passing, the connection relationship between adjacent segments within the disturbance coverage range is maintained, and the influence of direction instability on continuity judgment is reduced; When the disturbance type identification is door opening and closing, level change information is generated within the disturbance coverage range, and connection interruption is allowed at the segment boundary. 8.The method of claim 6, wherein, The modification includes: Within the continuous time window defined by the time markers, Doppler reconstruction information is generated based on the wireless network measurement records, the Doppler reconstruction information is used to represent the relative motion state, and is obtained by aligning and de-drifting the phase evolution of reference symbols and / or pilots within a resource unit over time; According to the Doppler reconstruction information, Doppler continuity indications are generated within the segment set, the Doppler continuity indications are used to represent the persistence, direction change, and amplitude level of the Doppler; According to the Doppler continuity indications, the connection state information and the continuity mask are adjusted, and the modification results are recorded in the adjusted connection state information and the continuity mask. 9.The method of claim 8, wherein, The adjustment includes: When the Doppler continuity indication represents that the Doppler direction is reversed and is accompanied by changes in frequency band usage state and basic service set color, connection interruption occurs at the segment boundary. When the amplitude of the Doppler is not higher than a preset amplitude threshold and the slowly varying quantity abstract representation exists a timing silence arrangement, the connection relationship of the corresponding section is maintained and a silence label is recorded in the continuity mask; When the amplitude of the Doppler is not lower than another preset amplitude threshold and the motion state record indicates stillness, a smoothing process is performed on the direction change component in the Doppler reconstruction information to obtain a modified direction change index, and the connection relationship of adjacent sections is maintained under the modified direction change index.

10. A feature matching system based on multi-source heterogeneous data of a smart terminal, characterized in that, Comprise: A collection module configured to acquire wireless network measurement records and motion state records by a target terminal, and associate time markers respectively; A calculation module configured to generate, based on the wireless network measurement records, an environment fingerprint for indicating a wireless scene feature of an observation position, the environment fingerprint being formed by combination of at least two types of wireless measurement statistics; and generate, based on the motion state records, a continuity parameter for reflecting motion continuity between records within a continuous time window defined by the time markers; A combination module configured to combine, two by two, data objects carrying time markers and associated with the environment fingerprint and the continuity parameter, which are extracted from the wireless network measurement records and / or the motion state records and whose time markers satisfy a corresponding relationship within the same continuous time window, to form candidate object pairs, the two data objects of the candidate object pairs being respectively derived from different modalities or different terminals; A matching module configured to perform, for the candidate object pairs, matching determination based on the environment fingerprint and the continuity parameter according to the time markers, and determine the candidate object pairs satisfying a predetermined determination rule as target matching pairs and output.

Images