A star glint phase ranging method based on orthogonal matching pursuit and interpolation refinement

Abstract

This invention proposes a phase ranging method based on orthogonal matched pursuit (OMP) and three-point interpolation refinement, applicable to BLE / SLE narrowband wireless communication systems. The method first constructs a delay-frequency dictionary, obtains multi-channel phase vectors through bidirectional phase measurement, and then expands and normalizes them. Subsequently, the OMP algorithm selects the atom that best matches the measured phase from a dynamically generated predefined dictionary, obtaining a coarse delay estimate. Then, based on the matching amplitudes of the selected atom and its two adjacent points, a three-point parabolic interpolation formula is used to refine the delay using a sub-step size, obtaining a more accurate delay estimate. Finally, the refined delay is converted into a distance value. This method effectively suppresses multipath interference through sparse modeling and overcomes the dictionary step size limit through the interpolation step, significantly improving ranging accuracy and possessing good embedded deployment value.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication and positioning technology, and particularly relates to a phase ranging method based on a multi-carrier phase difference (MCPD) ranging framework, combined with orthogonal matching pursuit (OMP) sparse reconstruction and three-point parabolic interpolation refinement. BACKGROUND

[0002] NearLink is a new type of short-range wireless communication technology, which has the characteristics of low power consumption, low delay, high reliability, etc., and is particularly suitable for high-precision ranging and positioning tasks in intelligent terminal, vehicle-mounted device, Internet of Things and other scenarios. The low-power version (SLE) supports multi-carrier modulation and two-way communication, providing a good physical layer basis for multi-carrier phase difference (MCPD, Multi-Carrier Phase Difference) ranging method.

[0003] The current mainstream MCPD ranging method mostly uses linear least squares fitting (LS) or FFT spectrum peak detection to estimate the propagation delay. These methods have certain accuracy in line-of-sight (LOS) environment, but in multipath or non-line-of-sight (NLOS) scenarios, they are easily disturbed by reflected paths, causing phase trajectory deviation and affecting ranging stability. In addition, traditional methods are usually limited by frequency resolution, making it difficult to achieve sub-step-level fine ranging.

[0004] Orthogonal matching pursuit (Orthogonal Matching Pursuit, OMP) is a classic sparse reconstruction algorithm, widely used in compressed sensing, signal recovery, image reconstruction and other fields. Its basic idea is: given a redundant dictionary, select the atoms (dictionary columns) most related to the observed signal step by step through iteration, and update the residual at each step by least squares to approximate the sparse representation of the original signal.

[0005] The application of orthogonal matching pursuit (OMP) algorithm to the sparse reconstruction method in the MCPD ranging scene can estimate the main path delay by constructing a delay-frequency dictionary and selecting the most matching atom through OMP iteration. This kind of method has strong multipath resolution ability in theory, but has the following shortcomings in practical application:

[0006] (1) Large dictionary storage overhead: the traditional OMP method needs to explicitly construct and store the complete delay-frequency dictionary matrix, with a space complexity of O(KN), which is difficult to deploy on embedded platforms.

[0007] (2) Ranging accuracy is limited by dictionary step size: the time delay estimated by OMP is a discrete grid point, and the accuracy is limited by the dictionary resolution. If higher accuracy is required, a denser dictionary is needed, further increasing the computational and storage burden.

[0008] (3) Lack of refinement mechanism: existing methods mostly take OMP output as the final result, without sub-meter level interpolation refinement on the estimated result, making it difficult to achieve sub-meter level ranging accuracy.

[0009] (4) Poor embedded adaptability: traditional sparse reconstruction methods have high computational complexity and many iterations, making it difficult to achieve real-time operation on resource-constrained MCU platforms. SUMMARY

[0010] The application proposes a ranging algorithm combining orthogonal matching pursuit (OMP) sparse solution and three-point interpolation refinement, aiming to improve the multipath resistance and ranging resolution of phase ranging methods on BLE, SLE and other narrowband platforms, especially suitable for wireless positioning scenarios that require low embedded resources but require sub-meter level ranging accuracy.

[0011] TECHNICAL SCHEME

[0012] A star flash phase ranging method based on orthogonal matching pursuit and interpolation refinement, comprising the following steps:

[0013] S1. Constructing a delay-frequency implicit dictionary;

[0014] The constructed delay-frequency implicit dictionary matrix has Fourier basis atoms as atoms;

[0015] S2. Multi-channel phase measurement and preprocessing;

[0016] Obtain and process phase information at multiple frequency points, eliminate clock error through round trip communication between the transmitting end and the receiving end, and obtain raw phase data; expand the raw phase into continuous phase; remove the global offset to construct a complex phase vector with unit amplitude; use filtering method to reduce the error caused by phase mutation; the complex phase vector is provided to step S3;

[0017] S3. Coarsely estimating the time delay using the OMP method;

[0018] Using the OMP method, the sparse representation of the observed signal is gradually approximated through iteration. In each iteration, the algorithm selects the atom that best matches the current residual from the dictionary, adds it to the support set, and performs least squares fitting on the support set to update the residual. Iteration continues until the residual is small enough or the number of iterations reaches a certain number, and the time delay index corresponding to the optimal atom is finally output as the main path estimate.

[0019] S4. Three-point parabolic interpolation refinement;

[0020] Based on the time delay estimation result of step S3, the three-point parabolic interpolation is used to refine the time delay estimation result.

[0021] S5. Calculate the distance estimate value according to the refined time delay estimation result.

[0022] Advantages

[0023] Compared with the traditional method, the method has the following advantages:

[0024] 1. High resolution: sub-grid delay estimation is realized between the dictionary grid by three-point parabolic interpolation, and the ranging accuracy can reach decimeter level (error ≤0.2 m under LOS condition)

[0025] 2. Anti-multipath interference: OMP can effectively identify the main line-of-sight atom and suppress the interference of surrounding multipath components on the phase, and the error remains in the range of 0.8 m~1.2 m in indoor multipath scenarios.

[0026] 3. Low computational complexity: a small number of OMP iterations are performed, and the computational complexity is O(KN), which is greatly reduced compared with traditional MUSIC / ESPRIT algorithms.

[0027] 4. Low storage overhead: no need to explicitly store the complete K x N dictionary, only the corresponding carrier phase vector is dynamically generated, and the memory occupation is less.

[0028] 5. Good scalability: suitable for BLE / SLE multi-carrier positioning system, and other multi-carrier phase ranging scenarios, K, N, parameters can be adjusted according to hardware performance. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 is a flow chart of the method of the application;

[0030] Figure 2 is a hardware experimental platform diagram of the star flash HI2821E of the embodiment of the application;

[0031] Figure 3 is a ranging data comparison diagram of the indoor parking lot of the embodiment of the application;

[0032] Figure 4 is an outdoor ranging data comparison diagram of the embodiment of the application;

[0033] Figure 5 is a comparison diagram of the memory occupation of the algorithm of the embodiment of the application. DETAILED DESCRIPTION

[0034] The technical solutions provided by the application will be further described below in combination with specific embodiments and their drawings. The advantages and features of the application will be clearer in combination with the following description.

[0035] A star flash phase ranging method based on orthogonal matching pursuit and interpolation refinement, comprising the following steps: Figure 1 )

[0036] S1. Delay-frequency implicit dictionary construction

[0037] Firstly, determine the maximum ranging distance of system requirement , thus calculate the maximum round-trip delay :

[0038]

[0039] Wherein, (unit: meter) is used to determine the maximum delay range covered by the dictionary, is the light speed constant.

[0040] Then, according to the dictionary size N (or the required ranging resolution), calculate the dictionary delay step :

[0041]

[0042] Set a set of candidate delays }, construct a dictionary matrix containing N atoms , wherein the th atom (Fourier basis atom) is defined as:

[0043]

[0044] These atoms constitute the phase response templates of the signal under different delays. The dictionary should cover the entire delay range corresponding to the maximum ranging range.

[0045]

[0046] Wherein, represents the th candidate delay value, K represents the number of subcarriers used in the multicarrier system, represents the frequency of the th subcarrier.

[0047] The final delay-frequency implicit dictionary matrix is constructed , wherein the th atom is:

[0048]

[0049] The traditional OMP algorithm needs to store the dictionary matrix explicitly, and the space complexity is . The implicit dictionary matrix introduced in the present application utilizes the structural characteristics of the dictionary Fourier basis, dynamically generates atoms through functions, and does not need to store the complete matrix, thereby reducing the space memory occupation.

[0050] S2. Multichannel phase measurement and preprocessing

[0051] This step aims to acquire and process phase information at multiple frequency points. By communicating back and forth between the transmitter and receiver, clock errors are eliminated and raw phase data is acquired. The raw phase is expanded into a continuous phase. Global offset is removed to construct a complex phase vector with unit amplitude. Filtering methods are used to reduce errors caused by phase abrupt changes. Since the main path is significantly stronger than other paths, the phase vector exhibits a sparse structure in the time delay domain, providing a foundation for subsequent OMP processing.

[0052] The specific process is as follows:

[0053] (1) Two-way half-duplex measurement: The transmitter sends a carrier signal to the receiver. The receiver samples the carrier phase at K frequency points and sends it back. The transmitter measures the return signal phase again to eliminate the transmit and receive clock error and obtain the round-trip phase. .

[0054] (2) Phase unwrap: unwrap the original phase according to Modulus expansion yields continuous phase. .

[0055] (3) Offset correction and normalization: Remove global offsets (such as phase baseline caused by carrier phase difference or temperature drift) and construct a unit amplitude complex phase vector. .

[0056] There are K carrier frequencies Each frequency corresponds to a measured phase difference. , forming a complex vector:

[0057]

[0058] Under line-of-sight conditions, the phase difference and the propagation distance r satisfy:

[0059]

[0060] in Phase deviation introduced by noise or multipath.

[0061] Since the number of propagation paths is relatively small (main path + several strong reflection paths), a sparse structure is assumed to exist in the time delay domain.

[0062] (4) Noise suppression: for Low-pass or median filtering can be used to suppress noise spikes and amplify errors.

[0063] S3. Use the OMP method to roughly estimate the time delay.

[0064] OMP aims to linearly combine the atoms in the dictionary with minimum number of atoms to represent the observed signal The sparse support set is constructed iteratively. The detailed procedure is as follows:

[0065] ① Initialization of residual vector , support set Λ and iteration number , as follows:

[0066] Residual vector

[0067] Support set Λ = ∅

[0068] Iteration number = 0

[0069] ② Iteration step (select one optimal atom in each round):

[0070] OMP algorithm iteratively approximates the sparse representation of the observed signal. In each iteration, the algorithm selects the atom from the dictionary that best matches the current residual, adds it to the support set, and performs least squares fitting on the support set to update the residual. The iteration continues until the residual is small enough or the number of iterations reaches the set number, and finally outputs the time delay index corresponding to the optimal atom as the main path estimate.

[0071] Calculate the inner product of the current residual and all atoms in the dictionary:

[0072]

[0073] Select the index with the maximum inner product , update the support set:

[0074]

[0075] Solve the least squares problem in the current support set:

[0076]

[0077] where represents the selected atom set extracted from the dictionary.

[0078] Update the residual:

[0079]

[0080] If the number of iterations reaches the upper limit, stop the iteration (the residual can also be judged to be lower than the threshold to stop the iteration).

[0081] ③ Result output:

[0082] The path with the smallest residual is finally selected, and the corresponding to the path delay index is the estimated main path delay.

[0083] Convert it into distance estimation value:

[0084]

[0085] S4. Three-point parabolic interpolation refinement

[0086] The present application further uses three-point parabolic interpolation refinement results on the basis of the OMP method.

[0087] Specifically as follows:

[0088] OMP preliminary estimated time delay index , the two side points are , The corresponding projection amplitude is:

[0089]

[0090] The interpolation formula is used to estimate the real peak value offset position δ, where δ∈[−0.5, 0.5], that is, the relative The refined offset of is:

[0091]

[0092] So the more accurate delay index estimation is:

[0093] .

[0094] This embodiment uses three-point parabolic interpolation, and for higher accuracy, multi-point interpolation refinement can also be used.

[0095] S5. The refined distance estimation value is finally calculated as:

[0096]

[0097] The present method constructs a fitted continuous model around the discrete peak value, thereby effectively breaking through the limitation of the dictionary step size, improving the ranging accuracy, and without the need for additional increase in the dictionary resolution, thereby avoiding the computational complexity introduced by high-dimensional sparse solution.

[0098] To verify the effectiveness of the present method, real measurement comparison is carried out on the star flash SLE platform, and the standard OMP method is selected as the comparison baseline. The test environment includes indoor parking lot (significant multipath) and outdoor open area (LOS), the ranging range is 1-10 meters, the same multi-carrier frequency configuration and hardware platform are used. The star flash hardware experimental platform is as Figure 2 The HI2821E module is used for IQ data acquisition and ranging algorithm running result output.

[0099] Wherein, the standard OMP method (method A) uses an explicit dictionary and OMP iteration to solve, and outputs a coarse time delay estimation; the method of the application (method B) adopts OMP sparse reconstruction combined with an implicit dictionary and three-point interpolation refinement.

[0100] The indoor parking lot ranging data (Fig. 1) and the outdoor scene ranging data (Fig. 2) of the comparison between the standard OMP method and the method of the application are shown in the following table. Figure 3 Figure 4 The memory occupation (Fig. 3) of the comparison between the standard OMP method and the method of the application is shown in the following table. Figure 5

[0101] Wherein, the ranging accuracy is measured by root mean square error (RMSE); the memory occupation is measured by total memory (KB) of dictionary storage and intermediate variables;

[0102] The performance comparison results are shown in Table 1.

[0103] Table 1 Performance comparison

[0104]

[0105] The results show that: it can be seen that the method of the application breaks through the dictionary step limit through interpolation refinement, and the ranging accuracy is improved by 40%-60%; at the same time, the implicit dictionary generation and the light interpolation strategy are adopted, which significantly reduces the memory and calculation burden.

[0106] In addition, the method of the application also has the following advantages:

[0107] Strong anti-multipath capability: OMP sparse modeling can effectively identify the main path and suppress multipath interference;

[0108] Strong embedded deployability: ranging is realized on a 32-bit MCU platform, meeting the real-time positioning demand;

[0109] Wide applicability: can be extended to multi-carrier ranging systems such as BLE, SLE and UWB, and has good universality.

[0110] The above is only a specific embodiment of the application, and it should be pointed out that for ordinary skilled persons in the art, the specific implementation of the application cannot be limited to these descriptions, and any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the application shall be included in the protection scope of the application. Those skilled in the art can make various changes in form and detail, including making a number of simple inferences or replacements, without departing from the spirit and scope of the application.

[0111] The following is a summary of the parameter description of the application:

[0112] I. Dictionary construction and system parameters

[0113] ​​: System maximum ranging distance (unit: meter), used to determine the maximum delay range covered by the dictionary.

[0114] : Light speed constant.

[0115] : Maximum round-trip propagation delay

[0116] : Number of atoms in the dictionary, i.e., the number of candidate delay points.

[0117] Δτ: Delay step size corresponding to a pair of adjacent atoms in the dictionary.

[0118] : The th candidate delay value, constituting the horizontal axis of the dictionary.

[0119] K: Number of subcarriers used in a multicarrier system.

[0120] : Frequency of the th subcarrier.

[0121] : Delay-frequency dictionary matrix, containing atoms, each of which is a -dimensional complex vector.

[0122] : The th dictionary atom,

[0123] II. Phase measurement and preprocessing

[0124] : Phase difference at the kth frequency point.

[0125] : Phase noise or deviation caused by multipath at the kth frequency point.

[0126] : Normalized complex phase vector,

[0127] r: Propagation distance

[0128] III. Parameters related to the OMP algorithm

[0129] : Residual vector of the th iteration.

[0130] : Index set of currently selected atoms (support set). : Inner product value of the jth atom and the current residual, used to select the most matching atom.

[0131] : Least squares solution solved under the current support set.

[0132] : Coarse principal path latency estimate output by OMP.

[0133] Four, three-point interpolation refinement parameters

[0134] : OMP output the projection amplitudes (i.e. matching strengths) of the points on both sides of the index and the center point.

[0135] delta: Interpolation offset, representing the sub-pixel offset of the true peak relative to the coarse index.

Claims

1. A star flash phase ranging method based on orthogonal matching pursuit and interpolation refinement, characterized in that, Includes the following steps: S1. Construct a delay-frequency implicit dictionary; The constructed delay-frequency implicit dictionary matrix has Fourier basis atoms. S2. Multi-channel phase measurement and preprocessing; The system acquires and processes phase information at multiple frequency points, eliminates clock errors, and obtains raw phase data through round-trip communication between the transmitter and receiver. The original phase is expanded into a continuous phase; global offset is removed, and a complex phase vector with unit amplitude is constructed; a filtering method is used to reduce the error caused by phase abrupt changes; the complex phase vector is provided to step S3. Specifically as follows: (1) Two-way half-duplex measurement: The transmitter sends a carrier signal to the receiver. The receiver samples the carrier phase at K frequency points and sends it back. The transmitter measures the phase of the returned signal again to eliminate the transmit and receive clock error and obtain the round-trip phase. ; (2) Phase unwrapping: unwrapping the original phase according to Modulus expansion yields continuous phase. ; (3) Offset correction and normalization: Remove global offset and construct a unit magnitude complex phase vector ; There are K carrier frequencies Each frequency corresponds to a measured phase difference. , forming a complex vector: Under line-of-sight conditions, the phase difference and the propagation distance r satisfy: in Phase deviation introduced by noise or multipath; (4) Noise suppression: for Perform low-pass or median filtering to suppress sudden noise changes; S3. Use the OMP method to roughly estimate the time delay; The OMP method is used to iteratively approximate the sparse representation of the observed signal. In each iteration, the algorithm selects the atom that best matches the current residual from the dictionary, adds it to the support set, and performs least squares fitting on the support set to update the residual; the iteration continues until the residual is small enough or the set number of iterations is reached, and finally outputs the time delay index corresponding to the optimal atom as the main path estimate. S4. Refinement of three-point parabolic interpolation; Based on the time delay estimation results in step S3, the time delay estimation results are refined using three-point parabolic interpolation. Specifically as follows: Delay index based on the initial OMP estimate in step S3 Its two sides are , The corresponding projection amplitude is: The true peak offset position δ is estimated using an interpolation formula, where δ∈[ [0.5, 0.5], that is, relative Refinement offset: This leads to a more accurate estimate of the delayed index: ； S5. Calculate the distance estimate based on the refined time delay estimation results.

2. The star flash phase ranging method based on orthogonal matching pursuit and interpolation refinement according to claim 1, characterized in that, Step S1 is as follows: First, determine the maximum ranging distance required by the system. Therefore, the maximum round-trip time delay is calculated. : in, Used to determine the maximum latency range of the dictionary coverage, in meters. The speed of light is constant; Then, the dictionary delay step size is calculated based on the dictionary size N. : Define a set of candidate delays { Construct a dictionary matrix containing N atoms. , of which Atoms Defined as: These atoms constitute the phase response template of the signal under different time delays; the dictionary covers the entire time delay range corresponding to the maximum ranging range; in, Indicates the first There are 1 candidate delay value, where K represents the number of subcarriers used in a multicarrier system. Indicates the first The frequency of each subcarrier.

3. The star flash phase ranging method based on orthogonal matching pursuit and interpolation refinement according to claim 1, characterized in that, Step S3 is as follows: (1) Initialize the residual vector Support set Λ and number of iterations ,as follows: residual vector Support set Λ= Number of iterations =0 (2) Iteration steps: In each iteration, the algorithm selects the atom that best matches the current residual from the dictionary, adds it to the support set, and performs least-squares fitting on this support set to update the residual. The iteration continues until the residual is sufficiently small or a set number of iterations is reached. Finally, the algorithm outputs the delay index corresponding to the optimal atom as the main path estimate. Specifically: Calculate the inner product of the current residual and all atoms in the dictionary: Select the index with the largest inner product Updated support sets: In the current context, the following solutions to least squares problems are supported: in This represents the set of selected atoms extracted from the dictionary; Update residuals: If the maximum number of iterations is reached, the iteration will stop. (3) Output of results: Finally, the path with the smallest residual is selected, corresponding to the path delay index. This is the estimated main path delay; Convert it to a distance estimate: 。 4. The star flash phase ranging method based on orthogonal matching pursuit and interpolation refinement according to claim 1, characterized in that, In step S5, the distance estimate for: in, The refined time delay estimation result obtained in step S4, is the speed of light constant.

Images