A multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation

Abstract

The present invention discloses a multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation, which relates to the field of indoor positioning technology. The method comprises: using a Bluetooth positioning base station to obtain the in-phase orthogonal signal and received signal strength indicator value of the target to be positioned; inputting the in-phase orthogonal signal and received signal strength indicator value into a trained neural network model to determine the direction angle and pitch angle of the target to be positioned; using a two-dimensional DOA estimation algorithm to optimize and calibrate the direction angle and pitch angle to determine the optimized direction angle and pitch angle; and determining the current position of the target to be positioned based on the optimized direction angle and pitch angle. A two-dimensional DOA estimation algorithm is used to fuse the received signal strength indicator value. When the target to be positioned moves, its position is determined based on the received signal strength indicator value obtained in real time, thereby improving the real-time performance of positioning. The signal strength indicator value has a certain regional limitation effect, which greatly limits the position range of the target and enhances positioning accuracy.

Description

Technical Field

[0001] The present invention relates to the field of indoor positioning technology, and in particular to a multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation. Background Art

[0002] With the introduction of the Angle of Arrival (AOA) / Angle of Departure (AOD) direction-finding technology in Bluetooth Core Specification 5.1, Bluetooth has gradually become a research hotspot in the field of indoor positioning due to its standard specification protocols, rich application ecosystem, low cost, and low power consumption. Compared with other indoor positioning technologies, Bluetooth indoor positioning has more potential value and possibilities.

[0003] The core of Bluetooth AOA indoor positioning is data preprocessing, data optimization, parameter estimation, and coordinate calculation for the signal information received by the antenna array. Data preprocessing mainly adjusts the format of large amounts of received data and performs matrix conversion. Data optimization optimizes and compensates for noise, multipath effects, and hardware errors in the received designated signal data. Parameter estimation and coordinate calculation mainly estimate the target's spatial angle information and calculate the three-dimensional position coordinates.

[0004] In current research and applications, only a small number of Bluetooth AOA indoor positioning technologies have been implemented, such as Bluetooth AOA indoor positioning using an improved Sequential Simulated Annealing (SSA) algorithm, Bluetooth AOA indoor positioning that integrates received signal strength indication (RSSI) values, and high-precision positioning that integrates Bluetooth AOA with deep learning. Currently, indoor positioning based on Bluetooth AOA typically uses a single data source for research and implementation. While this positioning method is relatively easy to develop, the positioning results derived from using a single data source cannot achieve both high precision and real-time performance, especially for different application scenarios. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional Direction of Arrival (DOA) estimation to solve the problem in the prior art that the positioning results calculated using single data cannot achieve both high precision and high real-time performance.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation, comprising:

[0008] Use Bluetooth positioning base station to obtain the in-phase orthogonal signal and received signal strength indicator value of the target to be located;

[0009] Inputting the in-phase orthogonal signals and received signal strength indicator values ​​of the target to be located into a trained neural network model to determine the direction angle and pitch angle of the target to be located; the neural network model includes: a convolutional neural network, a recurrent neural network and a fully connected layer;

[0010] Using a two-dimensional DOA estimation algorithm to optimize and calibrate the direction angle and the pitch angle to determine optimized direction angle and pitch angle;

[0011] The current position of the target to be located is determined according to the optimized direction angle and pitch angle.

[0012] Optionally, determining the current position of the target to be located according to the optimized direction angle and pitch angle, and then further comprising:

[0013] A fusion sensor is used to obtain the known position of the target to be located; the fusion sensor includes: a gyroscope and an accelerometer;

[0014] A difference is made between the current position and the known position, and a result of the current position is determined according to the difference; the result of the current position is qualified or unqualified.

[0015] Optionally, subtracting the current position from the known position and determining the current position based on the difference specifically includes:

[0016] Calculating a difference between the current position and the known position;

[0017] If the difference is less than or equal to the preset value, the result of determining the current position is qualified, and the current position is used as the final positioning coordinate;

[0018] If the difference is greater than a preset value, the result of determining the current position is unqualified.

[0019] Optionally, if the difference is greater than a preset value, determining that the result of the current position is unqualified, and then further comprising:

[0020] Debugging the parameters of the Bluetooth positioning base station until the difference is less than a preset value; the parameters include signal-to-noise ratio and sampling times.

[0021] Optionally, inputting the in-phase orthogonal signal and the received signal strength indicator value of the target to be located into a trained neural network model to determine the direction angle and pitch angle of the target to be located specifically includes:

[0022] The in-phase orthogonal signal and the received signal strength indicator value of the target to be located are transmitted to the convolutional neural network of the trained neural network model, and the time domain features of the in-phase orthogonal signal and the time domain features of the received signal strength indicator value are extracted; the convolutional neural network is a 7-layer convolutional neural network;

[0023] Transmitting the time domain features of the in-phase orthogonal signal and the time domain features of the received signal strength indicator value to the recurrent neural network of the trained neural network model, and extracting the frequency domain features of the in-phase orthogonal signal and the frequency domain features of the received signal strength indicator value;

[0024] The frequency domain features of the in-phase orthogonal signals and the frequency domain features of the received signal strength indication value are transmitted to the fully connected layer of the trained neural network model, and the nonlinear mapping of the time domain features of the in-phase orthogonal signals and the frequency domain features of the in-phase orthogonal signals, as well as the nonlinear mapping of the time domain features of the received signal strength indication value and the frequency domain features of the received signal strength indication value are extracted. The azimuth angle and the pitch angle of the target to be measured are determined based on the nonlinear mapping of the time domain features of the in-phase orthogonal signals and the frequency domain features of the in-phase orthogonal signals and the nonlinear mapping of the time domain features of the received signal strength indication value and the frequency domain features of the received signal strength indication value.

[0025] Optionally, optimizing and calibrating the azimuth angle and the elevation angle using a two-dimensional DOA estimation algorithm to determine the optimized azimuth angle and elevation angle specifically includes:

[0026] Determining a first matrix based on the two-dimensional DOA estimation algorithm and the in-phase orthogonal signals and the received signal strength indicator value of the target to be located;

[0027] determining a complex matrix according to the pseudo-inverse value of the first matrix;

[0028] Determine a row vector according to the complex matrix;

[0029] Determine, based on the row vector, an angle element corresponding to the real part of each eigenvalue and an angle element corresponding to the imaginary part of each eigenvalue;

[0030] The optimized azimuth angle and pitch angle are determined according to the angle element corresponding to the real part of each eigenvalue and the angle element corresponding to the imaginary part of each eigenvalue.

[0031] Optionally, based on the two-dimensional DOA estimation algorithm, determining a first matrix according to the in-phase orthogonal signals and the received signal strength indicator value of the target to be located specifically includes:

[0032] Based on the two-dimensional DOA estimation algorithm, the in-phase orthogonal signal and the received signal strength indicator value of the target to be located are obtained according to the Bluetooth positioning base station to determine the second matrix;

[0033] The first n columns are extracted from the eigenvector matrix of the second matrix to determine the first matrix.

[0034] Optionally, based on the two-dimensional DOA estimation algorithm, the in-phase orthogonal signals and received signal strength indicator values ​​of the target to be located are obtained according to the Bluetooth positioning base station to determine the first matrix, and the method further includes:

[0035] Create an identity matrix;

[0036] Based on the two-dimensional DOA estimation algorithm, a real matrix and an imaginary matrix are determined according to the unit matrix and the number of array elements of the Bluetooth positioning base station.

[0037] Optionally, the pseudo-inverse value of the first matrix is:

[0038] F1=((J×K1)×Ev)×((J×K2)×Ev);

[0039] F2=((K1×J)×Ev)×((K2×J)×Ev);

[0040] Wherein, F1 is the first pseudo-inverse value of the first matrix, F2 is the second pseudo-inverse value of the first matrix, J is the identity matrix, Ev is the first matrix, K1 is the real part matrix, and K2 is the imaginary part matrix.

[0041] Optionally, calculating the difference between the current position and the known position specifically includes:

[0042] Difference between the direction angle corresponding to the current position and the direction angle corresponding to the known position to determine the direction angle difference;

[0043] Difference between the pitch angle corresponding to the current position and the pitch angle corresponding to the known position to determine the pitch angle difference;

[0044] The direction angle difference is:

[0045] The pitch angle difference is:

[0046] Among them, MSE1 is the direction angle difference, MSE2 is the pitch angle difference, n1 is the number of collected in-phase orthogonal signals, n2 is the number of collected received signal strength indicator values, yi is the known direction angle, y j is a known pitch angle, is the angle value of the direction angle of the target to be measured, is the pitch angle value of the target to be measured.

[0047] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0048] The present invention provides a multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation. The method uses a Bluetooth positioning base station to obtain the in-phase orthogonal signals and received signal strength indicator values ​​of the target to be located. The in-phase orthogonal signals and received signal strength indicator values ​​of the target to be located are input into a trained neural network model to determine the azimuth and elevation angles of the target to be located. The method uses a two-dimensional DOA estimation algorithm to optimize and calibrate the azimuth and elevation angles to determine the optimized azimuth and elevation angles. The method then determines the current position of the target to be located based on the optimized azimuth and elevation angles. Since the two-dimensional DOA estimation algorithm is used to optimize and calibrate the azimuth and elevation angles, positioning accuracy is improved. Furthermore, the received signal strength indicator value is integrated. The position of the target to be located can be determined based on the received signal strength indicator value obtained in real time when the target to be located moves, thereby improving real-time performance. This method effectively overcomes the limitations of positioning results calculated using single data in terms of high precision and real-time performance, improving positioning accuracy while also taking into account high real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 This is a flow chart of the multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation of the present invention;

[0051] Figure 2 This is a schematic diagram of the multi-fusion structure corresponding to the multi-fusion Bluetooth AOA indoor positioning method based on DOA estimation provided by the present invention;

[0052] Figure 3 is a flow chart of the multi-fusion program of the present invention;

[0053] Figure 4 This is a flowchart of the multi-fusion program framework of the present invention.

[0054] Explanation of symbols:

[0055] 4x4 array antenna—1, antenna mapping—2, element mapping—3, input layer—4, hidden layer—5, output layer—6. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] The purpose of the present invention is to provide a multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation, which aims to improve the positioning accuracy by optimizing the calibration of the direction angle and the pitch angle by adopting a two-dimensional DOA estimation algorithm, and further integrate the received signal strength indication value. When the target to be positioned moves, its position can be determined according to the received signal strength indication value obtained in real time, thereby improving the real-time performance of positioning. Therefore, it effectively overcomes the limitations of the positioning results solved by using a single data in terms of high precision and real-time performance, and improves the positioning accuracy while taking into account high real-time performance.

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] Example 1

[0060] like Figure 1 As shown, the multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation in this embodiment includes:

[0061] Step 101: Use a Bluetooth positioning base station to obtain the in-phase and quadrature signals and received signal strength indicator values ​​of the target to be positioned.

[0062] In step 102, the in-phase orthogonal signals and the received signal strength indicator value of the target to be located are input into the trained neural network model to determine the direction angle and pitch angle of the target to be located; the neural network model includes: a convolutional neural network, a recurrent neural network and a fully connected layer; the neural network model is a multi-fusion neural network model.

[0063] Step 103 : Using a two-dimensional DOA estimation algorithm to optimize and calibrate the azimuth angle and the pitch angle, and determine the optimized azimuth angle and pitch angle.

[0064] Step 104: Determine the current position of the target to be located based on the optimized direction angle and pitch angle.

[0065] Furthermore, after step 104, the following steps are further included:

[0066] A fusion sensor is used to obtain the known position of the target to be located. The fusion sensor includes a gyroscope and an accelerometer, which are used to determine the direction and angle of the target to be located and correct positioning errors in real time.

[0067] The current position is subtracted from the known position and the result of the current position is determined based on the difference. The result of the current position is either qualified or unqualified; the difference is the mean square error value.

[0068] Furthermore, a difference is made between the current position and the known position, and a result of the current position is determined based on the difference, specifically including: calculating the difference between the current position and the known position; if the difference is less than or equal to a preset value, determining that the result of the current position is qualified, and taking the current position as the final positioning coordinate; if the difference is greater than the preset value, determining that the result of the current position is unqualified.

[0069] Furthermore, if the difference is greater than a preset value, the result of the current position is determined to be unqualified, and then it also includes: debugging the parameters of the Bluetooth positioning base station until the difference is less than the preset value; the parameters include signal-to-noise ratio and number of sampling times.

[0070] Furthermore, step 102 specifically includes:

[0071] The in-phase and quadrature signals and the received signal strength indicator values ​​of the target to be located are transmitted to the convolutional neural network of the trained neural network model to extract the time domain features of the in-phase and quadrature signals and the time domain features of the received signal strength indicator values; the convolutional neural network is a 7-layer convolutional neural network.

[0072] The time domain features of the in-phase and quadrature signals and the time domain features of the received signal strength indicator are transmitted to the recurrent neural network of the trained neural network model to extract the frequency domain features of the in-phase and quadrature signals and the frequency domain features of the received signal strength indicator. The recurrent neural network extracts the frequency domain features of the in-phase and quadrature signals and the frequency domain features of the received signal strength indicator by creating a long short-term memory (LSTM) network model.

[0073] The frequency domain features of the in-phase orthogonal signals and the frequency domain features of the received signal strength indication value are transmitted to the fully connected layer of the trained neural network model, and the nonlinear mapping between the time domain features of the in-phase orthogonal signals and the frequency domain features of the in-phase orthogonal signals, as well as the nonlinear mapping between the time domain features of the received signal strength indication value and the frequency domain features of the received signal strength indication value are extracted. The azimuth angle and the pitch angle of the target to be measured are determined based on the nonlinear mapping between the time domain features of the in-phase orthogonal signals and the frequency domain features of the in-phase orthogonal signals and the nonlinear mapping between the time domain features of the received signal strength indication value and the frequency domain features of the received signal strength indication value.

[0074] Specifically, a fully connected layer is used to process the nonlinear mapping of time domain features and frequency domain features for classification and regression.

[0075] Furthermore, a two-dimensional DOA estimation algorithm is used to optimize and calibrate the azimuth and elevation angles to determine the optimized azimuth and elevation angles, specifically including:

[0076] Based on a two-dimensional DOA estimation algorithm, a first matrix is ​​determined according to the in-phase orthogonal signals and the received signal strength indicator value of the target to be located.

[0077] A complex matrix is ​​determined according to the pseudo-inverse value of the first matrix.

[0078] According to the complex matrix, determine the row vector;

[0079] According to the row vector, determine the angle element corresponding to the real part of each eigenvalue and the angle element corresponding to the imaginary part of each eigenvalue.

[0080] The optimized azimuth angle and pitch angle are determined according to the angle element corresponding to the real part of each eigenvalue and the angle element corresponding to the imaginary part of each eigenvalue.

[0081] Furthermore, based on the two-dimensional DOA estimation algorithm, the first matrix is ​​determined according to the in-phase orthogonal signal and the received signal strength indicator value of the target to be located, specifically including:

[0082] Based on the two-dimensional DOA estimation algorithm, the in-phase orthogonal signals and the received signal strength indicator value of the target to be located are obtained according to the Bluetooth positioning base station to determine the second matrix.

[0083] Extract the first n columns from the eigenvector matrix of the second matrix to determine the first matrix.

[0084] Furthermore, based on a two-dimensional DOA estimation algorithm, the in-phase orthogonal signal and the received signal strength indicator value of the target to be located are obtained according to the Bluetooth positioning base station to determine the first matrix, which also includes:

[0085] Create an identity matrix.

[0086] Based on the two-dimensional DOA estimation algorithm, the real matrix and the imaginary matrix are determined according to the unit matrix and the number of array elements of the Bluetooth positioning base station.

[0087] Furthermore, the pseudo-inverse value of the first matrix is:

[0088] F1=((J×K1)×Ev)×((J×K2)×Ev);

[0089] F2=((K1×J)×Ev)×((K2×J)×Ev);

[0090] Wherein, F1 is the first pseudo-inverse value of the first matrix, F2 is the second pseudo-inverse value of the first matrix, J is the identity matrix, Ev is the first matrix, K1 is the real part matrix, and K2 is the imaginary part matrix.

[0091] Furthermore, the difference between the current position and the known position is calculated, specifically including:

[0092] The direction angle corresponding to the current position is subtracted from the direction angle corresponding to the known position to determine the direction angle difference.

[0093] The pitch angle difference is determined by subtracting the pitch angle corresponding to the current position from the pitch angle corresponding to the known position.

[0094] The direction angle difference is:

[0095] The pitch angle difference is:

[0096] Among them, MSE1 is the direction angle difference, MSE2 is the pitch angle difference, n1 is the number of collected in-phase orthogonal signals, n2 is the number of collected received signal strength indicator values, yi is the known direction angle, y j is a known pitch angle, is the angle value of the direction angle of the target to be measured, is the pitch angle value of the target to be measured.

[0097] like Figure 2 As shown, the multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation provided by the present invention is applied to a multi-fusion structure.

[0098] The multi-fusion structure includes: a 4×4 array antenna 1, antenna mapping 2, array element mapping 3, input layer 4, hidden layer 5, and output layer 6. The input layer is used to input in-phase orthogonal signals and received signal strength indicators, the hidden layer is used to estimate the parameters of the azimuth and elevation angles, and the output layer is used to output the position coordinates of the target to be measured.

[0099] Specifically, the present invention integrates positioning based on RSSI values, positioning based on two-dimensional DOA estimation, and positioning based on a neural network model. The RSSI value is added to the neural network model to achieve positioning together with the I / Q signal. At the same time, the angle calculated in the neural network model is used to optimize the angle calculated. The RSSI value can limit the range of the target position to a certain extent, and can better determine the positioning accuracy to a certain extent. Among them, A, B, and C are three positioning base stations, and their intersection is the position of the target to be measured. P1, P2, and P3 are three positions obtained relative to the three base stations A, B, and C. P1, P2, and P3 respectively represent the rough position of the target to be measured, that is, the triangular range formed by P1, P2, and P3 is the position range of the target. In other words, the actual target position is within this range. Positioning only by RSSI values ​​can only obtain a position range of the target, but cannot accurately obtain the true position.

[0100] like Figure 3-Figure 4 As shown, based on the multi-fusion structure, the multi-fusion program execution process adopted by the multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation provided by the present invention includes:

[0101] Step 301: Acquire data: A Bluetooth positioning base station is used to acquire the in-phase and quadrature signals and received signal strength indicator values ​​of the target to be located from a server.

[0102] Step 302: Data parsing and conversion. Extract the key data components, remove symbols such as the device ID and IP address from the in-phase / quadrature (I / Q) signal and received signal strength indicator (RSSI) value of the target, and then perform format adjustment and conversion preprocessing on the I / Q signal and RSSI value. Specifically, the preprocessing involves converting the data containing the I / Q signal and RSSI value from character format to array format. This data is then stored in a 16×100 matrix by constructing the array format.

[0103] Step 303: Set debugging parameters. These include antenna type, antenna element spacing, element distribution, signal-to-noise ratio (SNR), and sampling times. Specifically, the antenna type is a two-dimensional dual-polarized antenna array, the antenna element spacing is 0.25 mm, and the element distribution is 4×4; the initial SNR is 20 dB; and the sampling times are set to 100. The debugging parameters will be used in the feedback of subsequent steps. These parameters primarily influence or modify the I / Q signals and RSSI values. For example, if the constructed matrix is ​​16×100, 16 represents the 4×4 element distribution parameter. The data interval in the matrix is ​​represented by the antenna element spacing parameter. Modifying the debugging parameters will also modify the matrix data.

[0104] Step 304: Custom data processing. Developers can encapsulate custom APIs to capture the acquired I / Q signals and data parsing functions, allowing them to be called from their own programs. This allows for rapid solution verification and testing, providing a data interface for obtaining the in-phase and quadrature signals and received signal strength indicator values ​​of the target to be located, extracting key data components, performing format adjustment and conversion preprocessing, and setting debugging parameters. This interface can be used to output data, allowing users to develop or replace their own subsequent functions.

[0105] Step 305: Environmental Interference Optimization. In complex indoor environments, Kalman filtering is used to achieve overall optimization in the face of prominent multipath effects and noise interference. When multipath issues exist, the Kalman filter algorithm is optimized using the data from the parameter setting step to eliminate the impact of interference on the data. When there are no issues, the data remains unchanged. Specifically, the motion state of the target position to be located is continuously tracked. Every millisecond, the state is updated using the observed target coordinates (x, y) and the angle and speed collected by the sensor, continuously improving the accuracy of the position estimate.

[0106] Step 306: Hardware error optimization. Set the hardware type, configuration parameters, and sampling interval parameters to cope with hardware fluctuations in different scenarios. Perform hardware error compensation optimization on the data from the environmental interference optimization step to obtain the final data matrix.

[0107] Specifically, if the antenna hardware has phase shift and amplitude difference during sampling, the following formula can be used to calculate and compensate:

[0108] Where A represents the compensated amplitude difference, φ represents the compensated phase difference, I represents the in-phase component of the signal, and Q represents the quadrature component of the signal.

[0109] Step 307: Customize other optimizations. When estimating the parameters of the current position of the target and calculating the coordinates, the neural network model used is a multi-fusion neural network model. Similarly, the framework also extends the custom API interface and integrates RSSI and DOA estimation.

[0110] Step 308: Use a two-dimensional DOA estimation algorithm to perform parameter estimation and coordinate calculation.

[0111] Specifically include:

[0112] Create an identity matrix J, where J is a square matrix with all diagonal elements being 1 and the rest being 0; rotate matrix J by 90 degrees twice to obtain matrix J2.

[0113] Divide the number of antenna array elements by 2 and round it up to get matrix I0. Flip I0 horizontally, stack it, concatenate it, and divide it by 10 to get matrix Q. Then, subtract one from the number of antenna array elements and perform the same operation to get matrix Q2.

[0114] Multiply matrices Q, J2, and Q2, extract the real part of the result, and obtain the real matrix K1; extract the imaginary part of the result and obtain the imaginary matrix K2.

[0115] The I / Q signals and RSSI values ​​form the matrix X. The first n columns extracted from the eigenvector matrix of the matrix X form the matrix Ev. Then, the pseudo-inverse value of the matrix is ​​obtained to obtain F1 and F2. The formula is as follows:

[0116] F1=((J×K1)×Ev)×((J×K2)×Ev).

[0117] F2=((K1×J)×Ev)×((K2×J)×Ev).

[0118] Convert matrix F2 to a complex matrix and add it to matrix F1 to obtain a complex matrix P; obtain the eigenvalue matrix D and eigenvector matrix V of the complex matrix P; extract the diagonal elements from the eigenvalue matrix D to obtain column vectors, and then transpose them to row vectors to obtain E0.

[0119] Extract the real part of E0 to obtain a real row vector; then perform the inverse tangent calculation on each element; finally, multiply by 2 and divide by π to obtain the angle element G1 corresponding to each eigenvalue; extract the imaginary part of E0 to obtain an imaginary row vector; then perform the inverse tangent calculation on each element; finally, multiply by 2 and divide by π to obtain the angle element G2 corresponding to the imaginary part of each eigenvalue.

[0120] Square G1 and G2, take the square root, perform the inverse sine, and finally convert to an angular unit value to obtain the angle value theta1 of the direction angle of the target to be measured; perform element-wise division on each element in G2 and G1, perform the inverse tangent on the elements in the resulting row vector, and finally convert to an angular unit value to obtain the angle value theta2 of the pitch angle of the target to be measured.

[0121] Specifically, the fusion sensor can be used to obtain theta1 and theta2, and theta1 and theta2 can be used as the known positions of the target to be located.

[0122] Then, combined with the azimuth and elevation angles output by the deep learning neural network model, the mean square error (MSE1) and MSE2 are calculated using the following formulas, combining theta1 and theta2, to optimize the calibration:

[0123] The direction angle difference is: The pitch angle difference is:

[0124]

[0125] Among them, MSE1 is the direction angle difference, MSE2 is the pitch angle difference, n1 is the number of collected in-phase orthogonal signals, n2 is the number of collected received signal strength indicator values, yi is the known direction angle, y j is a known pitch angle, is the angle value of the direction angle of the target to be measured, is the pitch angle value of the target to be measured.

[0126] Step 309: Dynamically optimize the data through debugging to estimate the azimuth and elevation angles and determine the final position. Set the ranges for MSE1 and MSE2. If MSE1 and MSE2 do not meet the range requirements, the model continues to calculate. If MSE1 or MSE2 still does not meet the range after the set number of iterations, theta1 or theta2 will replace the azimuth and elevation angles calculated by the model to further estimate the final position of the target.

[0127] This invention uses a multi-fusion technology implementation approach, integrating RSSI values ​​with deep learning technology based on two-dimensional DOA estimation technology. Building on the high-precision advantage of the two-dimensional DOA estimation algorithm, RSSI values ​​are combined for regional limitation, further calibrating positioning accuracy. Simultaneously, the high real-time performance brought by deep learning is integrated to compensate for the shortcomings of the former, ultimately forming a complementary relationship between the various technologies and maximizing positioning performance. This method also provides a multi-fusion program framework that implements functions such as noise optimization, multipath effect optimization, and hardware error optimization, fully addressing the most prominent key issues in various scenarios. It also features a personalized and integrated API interface that allows access to raw data, parameter adjustment, and function call and integration, allowing developers to flexibly customize and integrate according to their own needs to meet the personalized needs of users in different scenarios.

[0128] The multi-fusion Bluetooth AOA indoor positioning method based on DOA estimation provided by the present invention has the following advantages:

[0129] By leveraging the advantages of various technical algorithms, positioning accuracy, real-time performance, and robustness are significantly improved. Furthermore, it is highly universal and applicable to most scenarios, effectively solving positioning problems in different scenarios.

[0130] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0131] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the methods, systems, and core concepts of the present invention. At the same time, those skilled in the art will appreciate that variations in the specific implementation methods and scope of application are possible based on the concepts of the present invention. In summary, this specification should not be construed as limiting the present invention.

Claims

1. A multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation, characterized in that: include: Use Bluetooth positioning base station to obtain the in-phase orthogonal signal and received signal strength indicator value of the target to be located; Inputting the in-phase orthogonal signals and the received signal strength indicator value of the target to be located into the trained neural network model to determine the direction angle and pitch angle of the target to be located; The neural network model includes: a convolutional neural network, a recurrent neural network and a fully connected layer; Using a two-dimensional DOA estimation algorithm to optimize and calibrate the direction angle and the pitch angle to determine optimized direction angle and pitch angle; Determining the current position of the target to be located according to the optimized direction angle and pitch angle; The in-phase orthogonal signal and the received signal strength indicator value of the target to be located are input into the trained neural network model to determine the direction angle and pitch angle of the target to be located, specifically including: The in-phase orthogonal signal and the received signal strength indicator value of the target to be located are transmitted to the convolutional neural network of the trained neural network model, and the time domain features of the in-phase orthogonal signal and the time domain features of the received signal strength indicator value are extracted; the convolutional neural network is a 7-layer convolutional neural network; Transmitting the time domain features of the in-phase orthogonal signal and the time domain features of the received signal strength indicator value to the recurrent neural network of the trained neural network model, and extracting the frequency domain features of the in-phase orthogonal signal and the frequency domain features of the received signal strength indicator value; The frequency domain features of the in-phase orthogonal signal and the frequency domain features of the received signal strength indication value are transmitted to the fully connected layer of the trained neural network model, and the nonlinear mapping of the time domain features of the in-phase orthogonal signal and the frequency domain features of the in-phase orthogonal signal, as well as the nonlinear mapping of the time domain features of the received signal strength indication value and the frequency domain features of the received signal strength indication value are extracted. The direction angle and the pitch angle of the target to be located are determined based on the nonlinear mapping of the time domain features of the in-phase orthogonal signal and the frequency domain features of the in-phase orthogonal signal and the nonlinear mapping of the time domain features of the received signal strength indication value and the frequency domain features of the received signal strength indication value.

2. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 1 is characterized in that: Determining the current position of the target to be located according to the optimized direction angle and pitch angle, and then further comprising: A fusion sensor is used to obtain the known position of the target to be located; the fusion sensor includes: a gyroscope and an accelerometer; A difference is made between the current position and the known position, and a result of the current position is determined according to the difference; the result of the current position is qualified or unqualified.

3. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 2 is characterized in that: Subtracting the current position from the known position and determining the current position based on the difference specifically includes: Calculating a difference between the current position and the known position; If the difference is less than or equal to the preset value, the result of determining the current position is qualified, and the current position is used as the final positioning coordinate; If the difference is greater than a preset value, the result of determining the current position is unqualified.

4. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 3 is characterized in that: If the difference is greater than a preset value, determining that the result of the current position is unqualified, and then further comprising: Debugging the parameters of the Bluetooth positioning base station until the difference is less than a preset value; the parameters include signal-to-noise ratio and sampling times.

5. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 1 is characterized in that: The azimuth angle and the elevation angle are optimized and calibrated using a two-dimensional DOA estimation algorithm to determine the optimized azimuth angle and elevation angle, specifically including: Determining a first matrix based on the two-dimensional DOA estimation algorithm and the in-phase orthogonal signals and the received signal strength indicator value of the target to be located; determining a complex matrix according to the pseudo-inverse value of the first matrix; Determine a row vector according to the complex matrix; Determine, based on the row vector, an angle element corresponding to the real part of each eigenvalue and an angle element corresponding to the imaginary part of each eigenvalue; The optimized azimuth angle and pitch angle are determined according to the angle element corresponding to the real part of each eigenvalue and the angle element corresponding to the imaginary part of each eigenvalue.

6. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 5 is characterized in that: Determining a first matrix based on the two-dimensional DOA estimation algorithm and the in-phase orthogonal signals and the received signal strength indicator value of the target to be located specifically includes: Based on the two-dimensional DOA estimation algorithm, the in-phase orthogonal signal and the received signal strength indicator value of the target to be located are obtained according to the Bluetooth positioning base station to determine the second matrix; The first n columns are extracted from the eigenvector matrix of the second matrix to determine the first matrix.

7. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 6 is characterized in that: Based on the two-dimensional DOA estimation algorithm, the in-phase orthogonal signal and the received signal strength indicator value of the target to be located are obtained according to the Bluetooth positioning base station to determine the first matrix, and the method also includes: Create an identity matrix; Based on the two-dimensional DOA estimation algorithm, a real matrix and an imaginary matrix are determined according to the unit matrix and the number of array elements of the Bluetooth positioning base station.

8. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 7 is characterized in that: The pseudo-inverse value of the first matrix is: F1=((J×K1)×Ev)×((J×K2)×Ev); F2=((K1×J)×Ev)×((K2×J)×Ev); Wherein, F1 is the first pseudo-inverse value of the first matrix, F2 is the second pseudo-inverse value of the first matrix, J is the identity matrix, Ev is the first matrix, K1 is the real part matrix, and K2 is the imaginary part matrix.

9. The multi-fusion Bluetooth AOA indoor positioning method based on two-dimensional DOA estimation according to claim 3 is characterized in that: Calculating the difference between the current position and the known position specifically includes: Difference between the direction angle corresponding to the current position and the direction angle corresponding to the known position to determine the direction angle difference; Difference between the pitch angle corresponding to the current position and the pitch angle corresponding to the known position to determine the pitch angle difference; The direction angle difference is: The pitch angle difference is: Among them, MSE1 is the direction angle difference, MSE2 is the elevation angle difference, n1 is the number of collected in-phase orthogonal signals, n2 is the number of collected received signal strength indicator values, y i is the known direction angle, y j is a known pitch angle, is the angle value of the direction angle of the target to be measured, is the pitch angle value of the target to be measured.

Images