Anti-lost backpack system with high-precision positioning and wireless communication function

Abstract

The present application relates to the technical field of anti-theft alarm and wearable article cross, disclose a kind of high-precision positioning and wireless communication function's anti-loss backpack system.Wearing state perception layer extracts wearing state vector by sensing array collection carrying characteristic signal;Multi-source positioning fusion engine obtains ultra-wide band and bluetooth low power consumption positioning signal and calculates relative position sequence;Dynamic anti-loss decision maker takes wearing state vector as front weighting factor, when representing carrying state, extract the phase difference rate of change of ultra-wide band signal and the signal attenuation gradient of bluetooth low power consumption signal, calculate the time correlation of both to construct space-time correlation feature matrix, input adaptive decision model and output anti-loss risk index, when exceeding dynamic threshold, trigger alarm instruction.The present application realizes the coupling of wearing state and positioning time sequence characteristics, overcomes the false alarm defect caused by signal fluctuation in single distance determination, improves the false negative detection rate of abnormal movement in stationary state, enhances system reliability.

Description

Technical Field

[0001] This invention relates to the field of cross-technology of anti-theft alarms and wearable items, and discloses an anti-loss backpack system with high-precision positioning and wireless communication functions. Background Technology

[0002] Existing anti-loss backpack systems with positioning and communication capabilities generally employ a hardware architecture combining a satellite positioning module and a short-range wireless communication module. During daily operation, the system periodically wakes the satellite positioning module to obtain the backpack's current geographical coordinates. Simultaneously, it uses the short-range wireless communication module to calculate the received signal strength indicator between the backpack and the associated terminal, converting this indicator into a spatial distance estimate. When the estimated spatial distance between the backpack and the associated terminal exceeds a preset fixed distance threshold, or when the backpack's geographical coordinates deviate from a preset electronic fence range, the system triggers an alarm mechanism, sending an alarm message to the associated terminal via the cellular mobile communication network. This architecture relies on a single-dimensional judgment logic based on distance or coordinates, enabling basic anti-loss monitoring functions in typical open environments.

[0003] The aforementioned conventional technical solutions suffer from rigid decision-making logic in practical applications. The received signal strength indication value of short-range wireless communication is highly susceptible to multipath effects, human obstruction, or co-channel electromagnetic interference in the environment, causing drastic fluctuations in the estimated spatial distance. In such cases, decision-making logic based on a fixed distance threshold is prone to false alarms due to signal fluctuations. This false alarm phenomenon is particularly frequent in complex electromagnetic environments such as crowded shopping malls or underground parking lots with severe signal obstruction. When both the backpack and the associated terminal are relatively stationary and the distance between them does not exceed the fixed threshold, if the backpack is moved slightly or moved a little by an unauthorized person, the system cannot detect this abnormal movement because the geographical coordinates and spatial distance do not change beyond the limits, resulting in a missed alarm.

[0004] Based on the specific implementation methods and operating mechanisms of the existing technologies, the existing anti-loss backpack systems have a core technical problem: the existing technologies rely on a single-dimensional spatial distance threshold or static coordinate comparison logic to determine loss, without coupling and analyzing the actual wearing status of the backpack with the dynamic evolution characteristics of multi-source positioning signals over time. This makes the system prone to false alarms due to instantaneous changes in ranging values ​​in complex environments with fluctuating wireless signals. At the same time, in relatively static scenarios where the backpack is moved abnormally within a small range by unauthorized personnel and the distance does not exceed the threshold, the lack of comprehensive analysis of signal time-series changes and wearing status makes it impossible to identify risks, resulting in missed alarms. Summary of the Invention

[0005] The purpose of this invention is to provide an anti-loss backpack system with high-precision positioning and wireless communication functions, which can solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A high-precision positioning and wireless communication anti-loss backpack system includes: a wear status sensing layer, which collects human body carrying feature signals through a capacitive sensing array and a pressure sensing array set in the shoulder strap area of ​​the backpack, and extracts the wear status vector of the human body carrying feature signals. The multi-source positioning fusion engine simultaneously acquires ultra-wideband positioning signals and Bluetooth Low Energy positioning signals, and calculates the relative position sequence between the backpack and the associated terminal. The dynamic anti-loss decision-maker uses the wearer state vector as a pre-weighting factor for positioning determination. When the wearer state vector represents a carrying state, it extracts the phase difference change rate of the ultra-wideband positioning signal and the signal attenuation gradient of the Bluetooth low power positioning signal, calculates the temporal correlation between the phase difference change rate and the signal attenuation gradient to construct a spatiotemporal correlation feature matrix, inputs the spatiotemporal correlation feature matrix into an adaptive decision-making model to output an anti-loss risk index, and triggers an alarm command when the anti-loss risk index exceeds a dynamically adjusted decision threshold.

[0007] Preferably, the wearable state perception layer extracts the wearable state vector by: performing frequency domain filtering on the capacitance sensing signal output by the capacitance sensing array to extract low-frequency human body dwelling feature components; performing time domain integration on the pressure sensing signal output by the pressure sensing array to extract continuous pressure distribution feature components; and performing feature-level fusion of the low-frequency human body dwelling feature components and the continuous pressure distribution feature components to generate a multi-dimensional wearable state vector, wherein each element in the wearable state vector corresponds to the spatial pressure distribution pattern and biocapacitance change law of the human body's shoulder contact area.

[0008] Preferably, the multi-source positioning fusion engine calculates the relative position sequence by: calculating the time difference of arrival of the ultra-wideband positioning signal to obtain a first spatial distance observation value; calculating the received signal strength indication of the Bluetooth low-power positioning signal to obtain a second spatial distance observation value; performing Kalman filtering fusion on the first spatial distance observation value and the second spatial distance observation value based on the signal quality indication parameters of the ultra-wideband positioning signal and the Bluetooth low-power positioning signal; and outputting the fused relative position sequence, wherein the relative position sequence includes spatial relative distance and relative angle that evolve continuously over time.

[0009] Preferably, the dynamic anti-loss decision device constructs the spatiotemporal correlation feature matrix by: extracting the phase difference change rate of the ultra-wideband positioning signal within a sliding time window to generate a first time-series sub-matrix; simultaneously extracting the signal attenuation gradient of the Bluetooth low-power positioning signal within the sliding time window to generate a second time-series sub-matrix; calculating the cross-correlation coefficient matrix between the first time-series sub-matrix and the second time-series sub-matrix; and aligning and splicing the timestamp intervals corresponding to the elements in the cross-correlation coefficient matrix that are greater than a preset correlation threshold to form the spatiotemporal correlation feature matrix.

[0010] Preferably, the dynamic anti-loss decision-maker outputs the anti-loss risk index by: inputting the spatiotemporal correlation feature matrix into the adaptive decision-making model containing a long short-term memory network structure, extracting the temporal dependency features and spatial mutation features in the spatiotemporal correlation feature matrix and performing a nonlinear mapping, and outputting the anti-loss risk index; The dynamically adjusted judgment threshold is calculated and updated in real time based on the historical positioning signal variance of the environment in which the backpack is located and the dispersion of the wearing state vector. When the historical positioning signal variance of the environment increases, the judgment threshold is decreased; when the dispersion of the wearing state vector increases, the judgment threshold is increased.

[0011] Preferably, the wearable state perception layer further performs the following operations: acquiring the capacitance change difference value of the capacitance sensing array in adjacent sampling periods; when the capacitance change difference value exceeds a preset transient threshold and the continuous pressure distribution feature component does not undergo synchronous abrupt change, it is identified as an interference signal that is not in contact with the human body; the low-frequency human body dwelling feature component and the continuous pressure distribution feature component in the sampling interval corresponding to the interference signal are set to zero; and interpolation compensation is performed using the previous historical valid feature components to generate the verified wearable state vector.

[0012] Preferably, the multi-source positioning fusion engine further performs the following operations: calculating the absolute value of the difference between the first spatial distance observation and the second spatial distance observation at the current time; when the absolute value of the difference exceeds a preset spatial consistency threshold, extracting the sum of squared residuals between the ultra-wideband positioning signal and the Bluetooth low-power positioning signal within a historical sliding window; comparing the sum of squared residuals between the ultra-wideband positioning signal and the Bluetooth low-power positioning signal; discarding spatial distance observations with large sums of squared residuals; and using the retained spatial distance observations as the observation input for Kalman filter fusion to perform state updates, outputting the corrected relative position sequence.

[0013] Preferably, after forming the spatiotemporal correlation feature matrix, the dynamic anti-loss decision device further introduces the continuous pressure distribution feature component in the wearable state vector as a third temporal sub-matrix, calculates the third-order tensor product of the third temporal sub-matrix with the first and second temporal sub-matrixes, expands the third-order tensor product in the time dimension to obtain the augmented spatiotemporal correlation feature matrix, and replaces the spatiotemporal correlation feature matrix with the augmented spatiotemporal correlation feature matrix and inputs it into the adaptive decision model.

[0014] Preferably, it also includes a communication wake-up scheduler, which monitors the rate of change of the anti-loss risk index output by the dynamic anti-loss decision device. When the rate of change of the anti-loss risk index is lower than a preset stable threshold, it controls the multi-source positioning fusion engine to enter a low-frequency sampling mode and shuts down the cellular long-distance communication link, while maintaining only the periodic broadcast listening of the Bluetooth low-power positioning signal. When the rate of change of the anti-loss risk index exceeds the stability threshold, the ultra-wideband positioning signal and the cellular long-distance communication link are activated, and the transmission power and retransmission count of the cellular long-distance communication link are dynamically adjusted according to the signal quality indication parameters of the current ultra-wideband positioning signal.

[0015] Preferably, it also includes a differentiated alarm linkage controller, which parses the wear state vector and the anti-loss risk index carried in the alarm trigger command. When the wear state vector represents an unworn state and the anti-loss risk index is in the first index range, the local high-frequency sound and light alarm device is activated and the physical reset path of the local high-frequency sound and light alarm device is cut off. When the wearable state vector represents a carrying state and the anti-loss risk index is in a second index range higher than the first index range, the local high-frequency sound and light alarm device is blocked, and a silent tracking command and real-time positioning trajectory are sent to the associated terminal through wireless communication.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The system extracts the wear state vector from the wearable state perception layer, which represents the characteristic signals carried by the wearer. This vector serves as a pre-weighting factor for positioning decisions. It combines the ultra-wideband positioning signal obtained from the multi-source positioning fusion engine with the Bluetooth Low Energy positioning signal to calculate the relative position sequence. A dynamic anti-loss decision-maker extracts the phase difference change rate of the ultra-wideband positioning signal and the signal attenuation gradient of the Bluetooth Low Energy positioning signal under the wearer's carrying status. The temporal correlation between these two signals is calculated to construct a spatiotemporal correlation feature matrix. This matrix is ​​input into an adaptive decision-making model, outputting an anti-loss risk index, which is compared with a dynamically adjusted decision threshold to trigger an alarm command. This mechanism deeply couples the wearer's state with the temporal change characteristics of multi-mode positioning signals. The calculation of temporal correlation filters out false triggers caused by instantaneous changes in a single signal, reducing the false alarm rate in complex electromagnetic environments. The introduction of the wear state vector enables the system to distinguish between carried and uncarried scenarios and adapt to different decision logics. In relatively static scenarios, abnormal movement is identified through the temporal correlation changes of the phase difference change rate and the signal attenuation gradient, overcoming the defect of missed detections.

[0017] 2. By fusing the low-frequency human presence feature component of the capacitive sensing signal with the continuous pressure distribution feature component of the pressure sensing signal at the feature level, and performing zeroing and interpolation compensation when transient interference is detected, the anti-interference capability and accuracy of the wearable state vector are improved. By fusing the arrival time difference of the ultra-wideband positioning signal and the received signal strength indication of the Bluetooth Low Energy positioning signal using Kalman filtering, and removing observations with large residual sums of squares, the smoothness and accuracy of the relative position sequence are improved. By introducing the continuous pressure distribution feature component as the third temporal submatrix to calculate the third-order tensor product to obtain the augmented spatiotemporal correlation feature matrix, the input feature dimension of the adaptive judgment model is enriched, improving the discriminative power of the anti-loss risk index. The communication wake-up scheduler switches between low-frequency sampling mode and wake-up of the long-distance cellular communication link according to the rate of change of the anti-loss risk index, reducing the overall power consumption of the system. The differentiated alarm linkage controller selects to activate the local high-frequency audible and visual alarm device or send a silent tracking command based on the interval between the wearable state vector and the anti-loss risk index, realizing differentiated physical responses under different risk scenarios. Attached Figure Description

[0018] Figure 1 Flowchart for wearable state perception and wearable state vector generation; Figure 2 Flowchart of multi-source positioning signal acquisition and fusion positioning; Figure 3 Flowchart for dynamic anti-loss risk index calculation and alarm triggering; Figure 4 Flowchart for wearable status perception interference signal identification and compensation; Figure 5 Flowchart for multi-source location anomaly observation removal and filtering correction; Figure 6 The flowchart shows the system power consumption adaptive scheduling and differentiated alarm linkage process. Detailed Implementation

[0019] In one embodiment, a high-precision positioning and wireless communication-enabled anti-loss backpack system achieves its anti-loss monitoring function through a layered architecture. The various functional modules of the system are integrated within the internal compartments of the backpack, and data interaction with associated terminals is established via a wireless communication link. The wearability perception layer collects human carrying characteristic signals through capacitive sensing arrays and pressure sensing arrays located in the shoulder strap area of ​​the backpack. The capacitive sensing array is distributed on the inner side of the shoulder strap where it contacts the skin of the shoulder, while the pressure sensing array is distributed in the force-supporting area of ​​the shoulder strap. The sampling periods of both are synchronized. The wearability perception layer preprocesses and extracts features from the collected capacitive and pressure sensing signals to generate a multi-dimensional wearability state vector.

[0020] Specifically, the wearable state sensing layer performs frequency domain filtering on the capacitive sensing signal output from the capacitive sensing array to extract low-frequency human presence feature components. The frequency domain filtering uses a Butterworth low-pass filter with the following transfer function:

[0021] in, This is the cutoff angular frequency, corresponding to the highest effective frequency of the capacitance signal from human contact. This represents the filter order. After low-pass filtering, high-frequency electromagnetic interference components in the capacitive sensing signal are removed, and the retained low-frequency components reflect the biocapacitance changes in the contact area between the human body and the shoulder strap. The sampling frequency of the capacitive sensing signal is set to 100Hz, and the cutoff angular frequency is... For a cutoff frequency of 10Hz, the filter order is... Set it to 4 to avoid signal phase distortion while ensuring filtering effect.

[0022] The wearable state-sensing layer performs time-domain integration on the pressure sensing signal output from the pressure sensing array to extract the characteristic components of the continuous pressure distribution. The formula for calculating the time-domain integration is:

[0023] in, The first in the pressure sensor array Each sensing unit at time Output pressure signal value, The starting time of integration, For the first Each sensing unit at time The continuous pressure integral value. Integral start time. This refers to the moment the system powers on or the moment the wearer last switched to a non-wearing state. Through time-domain integration, transient impact interference in the pressure-sensing signal is smoothed, and the integration result reflects the continuous pressure distribution pattern of the human shoulder against the shoulder strap. The sampling frequency of the pressure-sensing signal is consistent with that of the capacitive sensing signal, at 100Hz, and the integration result is updated every 100 milliseconds.

[0024] The wearable state perception layer performs feature-level fusion of the extracted low-frequency human body dwelling feature components and continuous pressure distribution feature components to generate a multi-dimensional wearable state vector. Feature-level fusion employs a concatenation fusion method, sequentially concatenating each element of the low-frequency human body dwelling feature components and each element of the continuous pressure distribution feature components according to their spatial correspondence. Each element in the resulting wearable state vector corresponds to the spatial pressure distribution pattern and biocapacitance variation law of the human shoulder contact area. The dimension of the wearable state vector corresponds to the number of sensing units in the capacitance sensing array and pressure sensing array. The value range of each element is normalized to the [0,1] interval. The normalization process uses a min-max normalization method to map the original signal values ​​to the [0,1] interval, eliminating dimensional differences between different sensing units. The correspondence between the wearable state vector elements and the human body carrying features is shown in Table 1 below.

[0025] Table 1. Correspondence between elements of the wearing state vector and human carrying features.

[0026] Table 1 shows the correspondence between each element in the wear state vector and the human carrying characteristics. By comparing the values ​​of each element with typical ranges, the wearing state of the backpack can be preliminarily determined. When more than 80% of the elements in the wear state vector are within the typical value range of the carrying state, it is determined to be in a carrying state; when more than 80% of the elements are within the typical value range of the non-carrying state, it is determined to be in a non-carrying state. For wear state vectors in intermediate states, the system will extend the determination time and make a comprehensive judgment based on the vector change trend over multiple consecutive sampling periods.

[0027] The multi-source positioning fusion engine simultaneously acquires ultra-wideband (UWB) positioning signals and Bluetooth Low Energy (BLE) positioning signals, and calculates the relative position sequence between the backpack and the associated terminal. The UWB positioning signal uses a pulse radio UWB system; the first spatial distance observation is obtained by measuring the time difference of arrival (TDOA) between the signal and the associated terminal. The distance calculation formula for the TDOA is:

[0028] in, The speed at which electromagnetic waves propagate in the air. This refers to the time difference between the transmission of an ultra-wideband signal from a backpack and its reception at the associated terminal. This is the first spatial distance observation calculated based on the ultra-wideband signal. The pulse repetition frequency of the ultra-wideband signal is set to 1 kHz, the time difference measurement accuracy reaches the nanosecond level, and the corresponding distance measurement accuracy is at the centimeter level.

[0029] Bluetooth Low Energy positioning signals obtain a second spatial distance observation value by measuring the received signal strength indicator. The conversion formula between received signal strength indicator and spatial distance is:

[0030] in, For reference distance, The received signal strength indication value at the reference distance. This is the path loss index. This is the strength indicator value of the Bluetooth Low Energy signal received by the associated terminal at the current moment. This is the second spatial distance observation calculated based on Bluetooth Low Energy signals. Reference distance. Set to 1 meter, the received signal strength indicator value at the reference distance. Path loss index obtained through factory calibration. Dynamically adjust according to environmental type, in open environments Set to 2.0, indoor environment Set to 3.0.

[0031] The multi-source positioning fusion engine uses signal quality indicator parameters from both ultra-wideband (UWB) and Bluetooth Low Energy (BLE) positioning signals to perform Kalman filtering fusion on first and second spatial distance observations. Signal quality indicator parameters include the signal-to-noise ratio (SNR) of the UWB signal and the bit error rate (BER) of the BLE signal. The observation noise covariance matrix of the Kalman filter is adjusted in real-time based on the signal quality indicator parameters; better signal quality results in a smaller observation noise covariance. The state vector of the Kalman filter includes relative distance, relative velocity, relative angle, and relative angular velocity. The state transition matrix is ​​constructed based on a uniform motion model, and the process noise covariance matrix is ​​determined based on the system's motion characteristics. The prediction step of the Kalman filter calculates the predicted state value and prediction covariance matrix for the current moment based on the state estimate from the previous moment. The update step corrects the predicted state value based on the observations at the current moment to obtain the optimal state estimate for the current moment. The Kalman filter fusion output is a fused relative position sequence, which includes spatial relative distance and relative angle that evolve continuously over time. The relative angle is obtained by measuring the angle of arrival (AOP) of the UWB signal.

[0032] The dynamic anti-loss decision-maker uses the wearer's state vector as a pre-weighting factor for positioning determination. When the wearer's state vector indicates a carrying status, it initiates the extraction of temporal features from multi-source positioning signals and the assessment of anti-loss risks. The dynamic anti-loss decision-maker extracts the phase difference change rate of the ultra-wideband positioning signal and the signal attenuation gradient of the Bluetooth Low Energy positioning signal. The phase difference change rate reflects the dynamic changes in the propagation path of the ultra-wideband signal, while the signal attenuation gradient reflects the rate at which the Bluetooth Low Energy signal changes with distance.

[0033] The formula for calculating the rate of change of phase difference is:

[0034] in, The phase difference of the ultra-wideband signal at the current moment. This represents the phase difference of the ultra-wideband signal at the previous moment. The sampling time interval, This represents the rate of change of the phase difference at the current moment. The phase difference of the ultra-wideband signal is obtained by measuring the phase difference between the received signal and the local reference signal, with a sampling time interval. Set to 100 milliseconds to match the output cycle of the multi-source localization fusion engine.

[0035] The formula for calculating the signal attenuation gradient is:

[0036] in, Current distance Bluetooth Low Energy signal strength indicator value at the location, For the previous distance Bluetooth Low Energy signal strength indicator value at the location, This represents the change in distance. This represents the signal attenuation gradient at the current moment. Distance change. The value is obtained by calculating the relative distance change between two adjacent sampling times through the fused relative position sequence.

[0037] The dynamic anti-loss decision-maker calculates the temporal correlation between the phase difference change rate and the signal attenuation gradient to construct a spatiotemporal correlation feature matrix. The sliding time window size is set to 10 seconds, and the sampling frequency is 10Hz, thus each sliding time window contains 100 sampling points. The sliding time window slides forward in 1-second steps, updating the contents of the first and second temporal sub-matrices after each slide. Within the sliding time window, the phase difference change rate of the ultra-wideband positioning signal is extracted to generate the first temporal sub-matrix. The rows of the first temporal sub-matrix correspond to the various sampling times within the sliding time window, and the columns correspond to different channels of the ultra-wideband signal. Simultaneously, within the sliding time window, the signal attenuation gradient of the Bluetooth Low Energy positioning signal is extracted to generate the second temporal sub-matrix. The rows of the second temporal sub-matrix correspond to the various sampling times within the sliding time window, and the columns correspond to different broadcast channels of the Bluetooth Low Energy signal.

[0038] The dynamic anti-loss decision-maker calculates the cross-correlation matrix between the first and second time-series sub-matrices. The formula for calculating the cross-correlation coefficient is as follows:

[0039] in, The first time series submatrix is ​​the first time series submatrix. Line number Column elements, The second time series submatrix is ​​the first... Line number Column elements, The first time-series submatrix is ​​the... The average value of the row elements. For the second time series submatrix The average value of the row elements. The number of sampling points within the sliding time window. The first time-series submatrix is ​​the... Row and the second time-series submatrix Cross-correlation coefficient between rows. The cross-correlation coefficient ranges from [-1, 1], and the closer the absolute value is to 1, the stronger the correlation between the two time series.

[0040] The dynamic anti-loss decision-maker aligns and concatenates the timestamp intervals corresponding to elements in the cross-correlation matrix that are greater than a preset correlation threshold, forming a spatiotemporal correlation feature matrix. The preset correlation threshold is set to 0.7 to filter signal segments with low temporal correlation and retain signal change features with strong correlation. During the alignment and concatenation process, the timestamp intervals with cross-correlation coefficients greater than the threshold are arranged in chronological order to form a continuous feature sequence, which serves as the row vector of the spatiotemporal correlation feature matrix.

[0041] The dynamic anti-loss decision-maker inputs the spatiotemporal correlation feature matrix into an adaptive decision-making model containing a long short-term memory (LSM) network structure and outputs an anti-loss risk index. The input layer of the adaptive decision-making model receives the spatiotemporal correlation feature matrix, and the number of neurons in the input layer is equal to the number of columns in the spatiotemporal correlation feature matrix. The hidden layer contains two layers of LSM network units: the first LSM network has 64 hidden units, and the second LSM network has 32 hidden units. Each LSM network unit includes an input gate, a forget gate, and an output gate. The input gate controls the input of new information, the forget gate controls the retention of old information, and the output gate controls the output of the hidden state. The output layer uses a sigmoid activation function for non-linear mapping, outputting an anti-loss risk index with values ​​ranging from [0,1].

[0042] The dynamic anti-loss decision device compares a dynamically adjusted judgment threshold with the anti-loss risk index. The dynamically adjusted judgment threshold is calculated in real time based on the variance of the historical positioning signals of the backpack's environment and the dispersion of the wearing state vector. The calculation formula is as follows:

[0043] in, The threshold is used as the benchmark for judgment. The historical location signal variance of the environment, The discreteness of the wearable state vector. and These are weighting coefficients. Benchmark threshold. Set to 0.5, weighting coefficient Set to 0.1, The setting is 0.2. The variance of the historical location signal of the environment is obtained by calculating the variance of the fused relative position sequence over the past 30 seconds, and the dispersion of the wearable state vector is obtained by calculating the standard deviation of each element in the wearable state vector. When the variance of the historical location signal of the environment increases, the decision threshold is adjusted. Reduce; as the dispersion of the wearable state vector increases, the decision threshold is lowered. Increased. When the anti-loss risk index exceeds the dynamically adjusted judgment threshold, the dynamic anti-loss decision device triggers an alarm command.

[0044] This embodiment generates a multi-dimensional vector representing the wearer's carrying status through the wearer status perception layer, which is used as a prerequisite for anti-loss determination. Combined with the high-precision relative position sequence output by the multi-source positioning fusion engine, the dynamic anti-loss decision-maker extracts the temporal change features of the multi-source positioning signals and constructs a spatiotemporal correlation feature matrix. The adaptive determination model outputs an anti-loss risk index and compares it with a dynamic threshold to trigger an alarm, thus realizing the coupled determination of wearer status and multi-source positioning signals.

[0045] In a preferred embodiment, the wearable state sensing layer also performs interference signal identification and compensation operations to improve the accuracy and anti-interference capability of the wearable state vector. The wearable state sensing layer acquires the capacitance change difference value of the capacitive sensing array in adjacent sampling periods, and the calculation formula for the capacitance change difference value is:

[0046] in, The first in the capacitive sensing array Each sensing unit at time The output capacitance signal value, For the first The capacitance signal value output by each sensing unit in the previous sampling period. The sampling time interval, For the first Each sensing unit at time The capacitance change difference value reflects the instantaneous rate of change of the capacitance signal. Interference signals that are not in contact with the human body usually cause rapid transient changes in the capacitance signal.

[0047] When the capacitance change difference exceeds a preset transient threshold and the continuous pressure distribution characteristic component does not undergo synchronous abrupt changes, the wearable state sensing layer identifies it as a non-human contact interference signal. The preset transient threshold is set to 0.5 (normalized capacitance change difference), and the preset pressure change threshold is set to 0.1 (normalized rate of change of the continuous pressure distribution characteristic component). When the capacitance change difference of any sensing unit in the capacitive sensing array exceeds the preset transient threshold, and the rate of change of the continuous pressure distribution characteristic component of all sensing units in the corresponding area of ​​the pressure sensing array does not exceed the preset pressure change threshold, it is determined to be a non-human contact interference signal. Such interference signals are usually caused by the proximity of a metal object, electrostatic discharge, or electromagnetic interference, and are not accompanied by synchronous changes in the pressure signal.

[0048] When interference signals are detected, the wearable state sensing layer sets the low-frequency human presence feature components and continuous pressure distribution feature components within the sampling interval corresponding to the interference signal to zero, and performs interpolation compensation using previous historical valid feature components. The interpolation compensation uses a linear interpolation method, and the calculation formula is as follows:

[0049] in, The effective sampling time preceding the start time of the interference signal. This is the next valid sampling time after the end of the interference signal. and They are respectively and Effective feature component values ​​at time t. For a moment The interpolation compensation feature component values ​​are calculated. When the interference signal lasts for more than 1 second, the wearable status sensing layer will trigger a wearable status re-determination process, confirming the current wearable status through signal changes over multiple consecutive sampling periods. Through interpolation compensation, the wearable status sensing layer generates a verified wearable status vector, eliminating the influence of non-human contact interference signals on wearable status determination.

[0050] The multi-source positioning fusion engine also performs anomaly removal operations to improve the reliability of the relative position sequence. The engine calculates the absolute value of the difference between the first and second spatial distance observations at the current moment. When the absolute value of the difference exceeds a preset spatial consistency threshold, an anomaly detection process is initiated. The preset spatial consistency threshold is determined statistically based on the ranging accuracy of ultra-wideband positioning signals and Bluetooth Low Energy positioning signals; the spatial consistency threshold varies for different signal quality levels.

[0051] The multi-source positioning fusion engine extracts the sum of squared residuals between the UWB and Bluetooth Low Energy positioning signals within a historical sliding window. The historical sliding window is set to 5 seconds and contains 50 sampling points. The formula for calculating the sum of squared residuals is:

[0052] in, The first in the history sliding window The observation value at each moment, For the first The Kalman filter prediction at time 1. This represents the number of sampling points within the historical sliding window. This represents the sum of squared residuals. The sum of squared residuals reflects the overall deviation between the observed and predicted values; a larger sum of squared residuals indicates lower reliability of the observed values.

[0053] The multi-source positioning fusion engine compares the sum of squared residuals corresponding to the ultra-wideband (UWB) positioning signal and the Bluetooth Low Energy (BLE) positioning signal. It discards spatial distance observations with larger sums of squared residuals and uses the retained spatial distance observations as input for Kalman filtering fusion to update the state, outputting a corrected relative position sequence. When the sum of squared residuals corresponding to the UWB positioning signal is greater than that corresponding to the BLE positioning signal, the first spatial distance observation is discarded, and only the second spatial distance observation is used for Kalman filtering updates; conversely, the second spatial distance observation is discarded, and only the first spatial distance observation is used for Kalman filtering updates. When the sum of squared residuals of both observations exceeds a preset threshold, the multi-source positioning fusion engine uses only the predicted value from the Kalman filter as the current state estimate and marks the positioning result at that moment as low reliability. The parameters for determining anomalies in multi-source positioning observations are shown in the table below.

[0054] Table 2. Parameters for Anomaly Judgment of Multi-Source Localization Observations

[0055] Table 2 shows the anomaly detection parameters for multi-source localization observations under different signal quality levels. The spatial consistency threshold and the residual sum of squares threshold increase as signal quality decreases to adapt to signal fluctuation characteristics under different environments. The signal quality level is updated every second and is determined based on the current UWRS and Bluetooth Low Energy bit error rate.

[0056] After forming the spatiotemporal correlation feature matrix, the dynamic anti-loss decision-maker also introduces the continuous pressure distribution feature component from the wear state vector as a third temporal sub-matrix to enrich the feature dimensions. The rows of the third temporal sub-matrix correspond to each sampling moment within the sliding time window, and the columns correspond to different sensing units of the pressure sensing array. The elements of the third temporal sub-matrix are the continuous pressure distribution feature component values ​​of the corresponding sensing unit at the corresponding moment, reflecting the dynamic changes of the wear state within the sliding time window.

[0057] The dynamic anti-loss decision-maker calculates the third-order tensor product of the third time-series submatrix, the first time-series submatrix, and the second time-series submatrix. The formula for calculating the third-order tensor product is as follows:

[0058] in, The first time series submatrix is ​​the first time series submatrix. Line number Column elements, The second time series submatrix is ​​the first... Line number Column elements, The third time series submatrix is ​​the first... Line number Column elements, The number of sampling points within the sliding time window. For a third-order tensor The elements. The dimension of the third-order tensor is... ,in This represents the row number of the first time-series submatrix. This represents the row number of the second time-series submatrix. This represents the row number of the third time-series submatrix.

[0059] The dynamic anti-loss decision-maker expands a third-order tensor along the time dimension to obtain an augmented spatiotemporal correlation feature matrix. The expansion is achieved by concatenating each time slice of the third-order tensor column-wise to form a two-dimensional matrix. The dimension of the expanded augmented spatiotemporal correlation feature matrix is... ,in This represents the number of sampling points within the sliding time window. The dynamic anti-loss decision-maker replaces the original spatiotemporal correlation feature matrix with the augmented spatiotemporal correlation feature matrix as input to the adaptive decision-making model. The augmented feature matrix contains correlation information between changes in wearing status and changes in multi-source positioning signals, improving the discriminative power of the anti-loss risk index. The number of neurons in the input layer of the adaptive decision-making model is adjusted accordingly to be equal to the number of columns in the augmented spatiotemporal correlation feature matrix, while the structures of the hidden and output layers remain unchanged.

[0060] This embodiment improves the anti-interference capability of the wearable state vector by identifying and interpolating interference signals, enhances the reliability of the relative position sequence by eliminating abnormal observations, and enriches the input feature dimension by introducing continuous pressure distribution feature components to construct an augmented spatiotemporal correlation feature matrix, thereby further improving the accuracy and robustness of anti-loss determination.

[0061] In another preferred embodiment, the system further includes a communication wake-up scheduler. The communication wake-up scheduler establishes a data connection with the dynamic anti-loss decision unit and the multi-source positioning fusion engine. It dynamically adjusts the system's communication mode and sampling frequency based on the rate of change of the anti-loss risk index to reduce the overall power consumption of the system. The communication wake-up scheduler monitors the rate of change of the anti-loss risk index output by the dynamic anti-loss decision unit in real time, with a monitoring time interval set to 1 second. The formula for calculating the rate of change of the anti-loss risk index is:

[0062] in, The current risk index for preventing loss. The loss risk index at the previous monitoring time. For monitoring time intervals, This represents the rate of change of the anti-loss risk index. The rate of change of the anti-loss risk index reflects the changing trend of anti-loss risk; a larger rate of change indicates that the anti-loss risk is rising faster, requiring the system to improve its response speed.

[0063] When the rate of change of the anti-loss risk index falls below a preset stability threshold, the communication wake-up scheduler controls the multi-source positioning fusion engine to enter low-frequency sampling mode and shuts down the cellular long-range communication link, maintaining only periodic broadcast listening of the Bluetooth Low Energy positioning signal. The preset stability threshold is set to 0.05 / second. In low-frequency sampling mode, the sampling period of the ultra-wideband positioning signal is extended from 100 milliseconds to 1 second, while the broadcast interval of the Bluetooth Low Energy positioning signal remains unchanged at 100 milliseconds. After the cellular long-range communication link is shut down, the system only interacts with associated terminals through the Bluetooth Low Energy link, transmitting necessary status information and low-priority positioning data. The system's power consumption is reduced to less than 30% of the normal operating mode, extending battery life.

[0064] When the rate of change of the anti-loss risk index exceeds a preset stability threshold, the communication wake-up scheduler activates the ultra-wideband positioning signal and the cellular long-distance communication link, and dynamically adjusts the transmission power and retransmission count of the cellular long-distance communication link based on the current signal quality indicator parameters of the ultra-wideband positioning signal. The communication wake-up scheduler completes the activation of the ultra-wideband positioning signal and the cellular long-distance communication link within 100 milliseconds, ensuring that the system can respond to the anti-loss risk in a timely manner. The better the signal quality of the ultra-wideband positioning signal, the lower the transmission power of the cellular long-distance communication link and the fewer the retransmissions; the worse the signal quality of the ultra-wideband positioning signal, the higher the transmission power of the cellular long-distance communication link and the more retransmissions. The transmission power adjustment range is 0dBm to 20dBm, and the retransmission count adjustment range is 0 to 3 times. By dynamically adjusting the transmission power and retransmission count, the system power consumption is reduced while ensuring communication reliability. The anomaly judgment parameters for multi-source positioning observations are shown in the table below.

[0065] Table 3. Parameters of Differentiated Alarm Linkage Strategy

[0066] Table 3 shows the differentiated alarm linkage strategies corresponding to different wearing states and anti-loss risk index ranges. The system selects the corresponding alarm action and communication method according to the actual scenario, realizing the matching of risk level and response measures. The response priority is from low to high as very low, low, medium, and high. High-priority alarm actions will interrupt low-priority operations.

[0067] The system also includes a differentiated alarm linkage controller, which establishes data connections with the dynamic anti-loss decision unit, local alarm device, and wireless communication module to execute differentiated alarm linkage operations based on the information in the alarm command. The differentiated alarm linkage controller parses the wearable state vector and anti-loss risk index carried in the alarm trigger command, and executes the corresponding alarm action according to the interval in which the wearable state and anti-loss risk index fall.

[0068] When the wearable state vector indicates an unworn state and the anti-loss risk index is within the first index range, the differentiated alarm linkage controller activates the local high-frequency audible and visual alarm device and disconnects its physical reset path. The first index range is set to [0.7, 1.0]. The local high-frequency audible and visual alarm device includes a high-brightness LED indicator and a high-decibel buzzer. The LED indicator flashes at 5Hz, the buzzer emits at 2kHz, and the sound pressure level is 85dB. The physical reset path is disconnected by controlling a relay to break the electrical connection between the local reset button and the alarm device. After disconnection, the alarm device can only be shut down via a remote reset command sent from the associated terminal, preventing unauthorized personnel from manually disabling the alarm.

[0069] When the wear status vector indicates a carrying state and the anti-loss risk index is in the second index range (higher than the first index range), the differentiated alarm linkage controller disables the local high-frequency audible and visual alarm device and sends a silent tracking command and real-time location trajectory to the associated terminal via wireless communication. The second index range is set to [0.6, 1.0]. The silent tracking command includes the backpack's unique identifier and the current anti-loss risk level. The real-time location trajectory is updated and transmitted to the associated terminal at 1-second intervals. The real-time location trajectory includes location information from the past 5 minutes, with each location point including relative distance, relative angle, and timestamp. The associated terminal can view the backpack's real-time location and movement trajectory through a companion application, achieving covert anti-loss tracking. When the anti-loss risk index drops below the second index range, the differentiated alarm linkage controller stops sending silent tracking commands and real-time location trajectories, and the system returns to normal monitoring status.

[0070] This embodiment uses a communication wake-up scheduler to dynamically adjust the system's communication mode and sampling frequency based on the rate of change of the anti-loss risk index, thereby reducing the overall power consumption of the system. The differentiated alarm linkage controller executes corresponding alarm actions based on the wearer's status and the anti-loss risk index, realizing differentiated responses under different risk scenarios.

Claims

1. A backpack system for preventing loss, featuring high-precision positioning and wireless communication, characterized in that: include: The wearable state perception layer collects human body carrying characteristic signals through a capacitive sensing array and a pressure sensing array set in the shoulder strap area of ​​the backpack, and extracts the wearable state vector of the human body carrying characteristic signals. The multi-source positioning fusion engine simultaneously acquires ultra-wideband positioning signals and Bluetooth Low Energy positioning signals, and calculates the relative position sequence between the backpack and the associated terminal. The dynamic anti-loss decision-maker uses the wearer state vector as a pre-weighting factor for positioning determination. When the wearer state vector represents a carrying state, it extracts the phase difference change rate of the ultra-wideband positioning signal and the signal attenuation gradient of the Bluetooth low power positioning signal, calculates the temporal correlation between the phase difference change rate and the signal attenuation gradient to construct a spatiotemporal correlation feature matrix, inputs the spatiotemporal correlation feature matrix into an adaptive decision-making model to output an anti-loss risk index, and triggers an alarm command when the anti-loss risk index exceeds a dynamically adjusted decision threshold.

2. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 1, characterized in that, The wearable state perception layer extracts the wearable state vector by: performing frequency domain filtering on the capacitance sensing signal output by the capacitance sensing array to extract low-frequency human body dwelling feature components; performing time domain integration on the pressure sensing signal output by the pressure sensing array to extract continuous pressure distribution feature components; and performing feature-level fusion of the low-frequency human body dwelling feature components and the continuous pressure distribution feature components to generate a multi-dimensional wearable state vector. Each element in the wearable state vector corresponds to the spatial pressure distribution pattern and biocapacitance change law of the human body's shoulder contact area.

3. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 1, characterized in that, The multi-source positioning fusion engine calculates the relative position sequence by: calculating the time difference of arrival of the ultra-wideband positioning signal to obtain a first spatial distance observation value; calculating the received signal strength indication of the Bluetooth low-power positioning signal to obtain a second spatial distance observation value; and performing Kalman filtering fusion on the first spatial distance observation value and the second spatial distance observation value based on the signal quality indication parameters of the ultra-wideband positioning signal and the Bluetooth low-power positioning signal, and outputting the fused relative position sequence, wherein the relative position sequence includes spatial relative distance and relative angle that evolve continuously over time.

4. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 1, characterized in that, The dynamic anti-loss decision device constructs the spatiotemporal correlation feature matrix by: extracting the phase difference change rate of the ultra-wideband positioning signal within a sliding time window to generate a first time-series sub-matrix; simultaneously extracting the signal attenuation gradient of the Bluetooth low-power positioning signal within the sliding time window to generate a second time-series sub-matrix; calculating the cross-correlation coefficient matrix between the first time-series sub-matrix and the second time-series sub-matrix; and aligning and splicing the timestamp intervals corresponding to the elements in the cross-correlation coefficient matrix that are greater than a preset correlation threshold to form the spatiotemporal correlation feature matrix.

5. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 1, characterized in that, The dynamic anti-loss decision-maker outputs the anti-loss risk index by: inputting the spatiotemporal correlation feature matrix into the adaptive decision-making model containing a long short-term memory network structure, extracting the temporal dependency features and spatial mutation features in the spatiotemporal correlation feature matrix and performing a nonlinear mapping, and outputting the anti-loss risk index. The dynamically adjusted judgment threshold is calculated and updated in real time based on the historical positioning signal variance of the environment in which the backpack is located and the dispersion of the wearing state vector. When the historical positioning signal variance of the environment increases, the judgment threshold is decreased; when the dispersion of the wearing state vector increases, the judgment threshold is increased.

6. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 2, characterized in that, The wearable state perception layer also performs the following operations: acquiring the capacitance change difference value of the capacitance sensing array in adjacent sampling periods; when the capacitance change difference value exceeds a preset transient threshold and the continuous pressure distribution feature component does not undergo synchronous abrupt change, it is identified as an interference signal that is not in human contact; the low-frequency human dwelling feature component and the continuous pressure distribution feature component in the sampling interval corresponding to the interference signal are set to zero; and interpolation compensation is performed using the previous historical valid feature components to generate the verified wearable state vector.

7. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 3, characterized in that, The multi-source positioning fusion engine further performs the following operations: calculates the absolute value of the difference between the first spatial distance observation and the second spatial distance observation at the current time; when the absolute value of the difference exceeds a preset spatial consistency threshold, extracts the sum of squared residuals between the ultra-wideband positioning signal and the Bluetooth low-power positioning signal within a historical sliding window; compares the sum of squared residuals corresponding to the ultra-wideband positioning signal and the Bluetooth low-power positioning signal; removes spatial distance observations with large sums of squared residuals; and uses the retained spatial distance observations as the observation input for Kalman filter fusion to perform state updates, outputting the corrected relative position sequence.

8. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 4, characterized in that, After forming the spatiotemporal correlation feature matrix, the dynamic anti-loss decision device also introduces the continuous pressure distribution feature component in the wearable state vector as a third temporal sub-matrix, calculates the third-order tensor product of the third temporal sub-matrix with the first and second temporal sub-matrix, expands the third-order tensor product in the time dimension to obtain the augmented spatiotemporal correlation feature matrix, and replaces the spatiotemporal correlation feature matrix with the augmented spatiotemporal correlation feature matrix and inputs it into the adaptive decision model.

9. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 1, characterized in that, It also includes a communication wake-up scheduler, which monitors the rate of change of the anti-loss risk index output by the dynamic anti-loss decision device. When the rate of change of the anti-loss risk index is lower than a preset stable threshold, it controls the multi-source positioning fusion engine to enter a low-frequency sampling mode and shuts down the cellular long-distance communication link, while maintaining only the periodic broadcast listening of the Bluetooth low-power positioning signal. When the rate of change of the anti-loss risk index exceeds the stability threshold, the ultra-wideband positioning signal and the cellular long-distance communication link are activated, and the transmission power and retransmission count of the cellular long-distance communication link are dynamically adjusted according to the signal quality indication parameters of the current ultra-wideband positioning signal.

10. The anti-loss backpack system with high-precision positioning and wireless communication functions according to claim 1, characterized in that, It also includes a differentiated alarm linkage controller, which parses the wear state vector and the anti-loss risk index carried in the alarm trigger command. When the wear state vector represents an unworn state and the anti-loss risk index is in the first index range, the local high-frequency sound and light alarm device is activated and the physical reset path of the local high-frequency sound and light alarm device is cut off. When the wearable state vector represents a carrying state and the anti-loss risk index is in a second index range higher than the first index range, the local high-frequency sound and light alarm device is blocked, and a silent tracking command and real-time positioning trajectory are sent to the associated terminal through wireless communication.

Images