A method for motion target perception based on a UWB CIR signal

Abstract

The application discloses a kind of motion target perception methods of CIR signal based on UWB, belong to UWB ultra-wideband wireless communication perception technical field.The method is according to the information of transmitting UWB and receiving UWB equipment, and the time delay information is solved by filtering, background subtraction and other algorithms, and the position of motion target is perceived and estimated.The application perceives the time delay information of motion target in motion environment, and estimates the position information of motion target.The method is simple and easy to implement, can better resist interference, and can detect and identify the position of living body target in real time in practical application, and the perception effect is more superior.

Description

Technical Field

[0001] This invention belongs to the field of UWB ultra-wideband wireless communication sensing technology, and specifically refers to a moving target sensing method based on UWB CIR (channel impulse response) signals. Background Technology

[0002] In practical production and daily life applications, UWB positioning and sensing are playing an increasingly important role. UWB, or Ultra Wideband, is a carrier-free communication technology. In wireless positioning, the main advantage of using UWB signals over narrowband signals is that UWB signals can accurately separate the first-pass signal and multipath reflection signals in wireless transmission, a capability that narrowband signals lack.

[0003] UWB boasts numerous advantages, including strong penetration, low power consumption, excellent anti-interference capabilities, high security, large spatial capacity, and precise positioning. It is widely used in industries such as intelligent manufacturing, smart construction, elderly care and healthcare, public safety, and logistics. UWB sensing complements UWB systems primarily based on UWB positioning. By learning time delay estimation algorithms, it improves the accuracy of target location estimation, thereby achieving a more accurate representation of UWB sensing capabilities.

[0004] Currently, most of the research areas in UWB are in positioning systems. There are no effective technical solutions in the field of UWB sensing. UWB positioning methods mostly rely on observations to determine the location of a target, such as angle and time of arrival, and there is still room for improvement in positioning accuracy. Summary of the Invention

[0005] In view of this, the present invention proposes a moving target perception method based on UWB CIR signals. This method has high accuracy and can perceive target time delay information. Combined with UWB device location information, the position of the target can be estimated.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A moving target sensing method based on UWB CIR signals includes the following steps:

[0008] Step 1: Measure the CIR over a certain period of time and perform cumulative averaging to obtain the averaged CIR measurement value z. Outliers are directly rejected.

[0009] Step 2: Apply mean filtering to the CIR measurement value z to obtain the filtered signal y;

[0010] Step 3: Perform variance filtering on the CIR measurement value z of the background when no target appears and the y value after the target appears to obtain the CIR variance value;

[0011] Step 4: Subtract the background CIR measurement value z from the current CIR variance value to obtain the variance difference s. Then, perform front-end detection on s to obtain the target time delay τ. TP The estimated value

[0012] Step 5, based on the relation τ TP =(||p tx -p T ||+||p rx -p T ||) / c, using the estimated value from step 4 The target position p is obtained by finding the intersection of ellipses. T In the formula, p tx It is the location of the base station, p rx is the label position, c is the speed of light, and |||| represents the distance to be calculated.

[0013] Furthermore, in step 1, the criteria for determining outliers include:

[0014] (1) Not enough cumulative preamble was collected before CIR measurement. The number of symbols in the cumulative preamble is at least half the number of transmitted preambles.

[0015] (2) At the estimated first path position τ FP The previous CIR sample points were not less than 5 times the first node coefficient h0 representing the inherent noise level;

[0016] (3) The maximum number of sampling points for the measured CIR shall not be less than 5 times the maximum node coefficient and not more than 2 times the maximum node coefficient.

[0017] Furthermore, the specific method for step 2 is as follows:

[0018] With coefficient The form tracks the current CIR measurement z, forming a piecewise linear function with uniform intervals:

[0019]

[0020] Where, τ i ≤τ<τ i +1, where τ represents the specific measured value;

[0021] knots of τ i The piecewise linear parameters are:

[0022] τ i =τ start +Δτ knots *i, i = {0,…,N} knots -1}

[0023]

[0024] Δτ knots It is the CIR measurement sampling period Δτ s divisible by, i.e.

[0025] The CIR measurement value z is then expressed as

[0026]

[0027] Signal The calculation method relative to the current segment parameter is as follows:

[0028] y j =z j -h(τ meas，j ), j = {0, ..., N} S -1}

[0029] This signal y is used to recursively update the coefficient h:

[0030] h←h+Ky

[0031] Wherein, the gain matrix

[0032] The coefficient h is initialized using the first received measurement value z.

[0033]

[0034] For all i∈{0,1,…,N knots -1}, by continuously measuring and updating the coefficient h, and outputting the corresponding signal y.

[0035] Furthermore, step 3 is performed as follows:

[0036] The CIR variance is parameterized as a piecewise constant function with coefficients. The current CIR variance is updated according to the following formula:

[0037]

[0038] Where, l∈{0,1,…,N} knots -2},τ l ≤τ j <τ l+1 Gain

[0039] Set the background change coefficient to Tracking and updating the change coefficient is set to That is, the background CIR variance value is updated to

[0040]

[0041] in,

[0042] Furthermore, step 4 is specifically implemented as follows:

[0043] Using the current CIR measurement value z and the background CIR variance value, regions with temporarily high CIR variance are detected by subtracting the background variance value from the current variance value:

[0044]

[0045] Among them, β>1 is a constant scalar sensitivity factor;

[0046] Find τ in the region where CIR is higher than the typical variance. l ≤τ<τ l+1 The section, of which

[0047] The position τ of the target path TP The measurement is taken by finding the first cluster value in this segment, i.e.

[0048] and

[0049] Where, l∈{0,1,…,N} knots -2};

[0050] Searching in N win It has a higher variance N than usual under the window length. seg The specific time delay value of the number.

[0051] The beneficial effects of this invention are as follows:

[0052] (1) Based on the actual environment and UWB equipment, this invention senses the time delay information of the CIR signal of the moving target, realizes the function of estimating the position of the moving target, and ensures the convergence of the positioning method through information communication between nodes. This not only improves the positioning accuracy of the target node, but also expands the coverage area, reduces the positioning error of the system, and improves the system performance.

[0053] (2) Interference from simultaneously operating radar systems in some radio frequency bands is an increasingly serious problem. This invention uses UWB sensing, which, thanks to coordinated transmitting and receiving elements, can better resist interference and obtain more accurate target information. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of time delay estimation based on CIR signals in an embodiment of the present invention;

[0055] Figure 2 This is a flowchart of the UWB sensing process for the position of a moving target based on CIR signals in an embodiment of the present invention. Detailed Implementation

[0056] A moving target sensing method based on UWB CIR signals includes the following steps:

[0057] (1) Measure the CIR over a certain period of time and perform cumulative averaging. Some outliers are directly rejected. The mean value is filtered on the measured value z of the average cumulative CIR to obtain the result y.

[0058] (2) Using coefficients The form that tracks the average value of the currently measured CIR is a piecewise linear function with uniform intervals, i.e.

[0059]

[0060] Where τ i ≤τ<τ i +1, and knots of τ i The piecewise linear parameters are

[0061] τ i =τ start +Δτ knots *i for i={0,...,N knots -1}

[0062]

[0063] Δτ knots It is the CIR measurement sampling period Δτ s divisible by, i.e.

[0064]

[0065] The minimum spacing m = 64 is given by the resolution of the leading edge detection algorithm of DWM1000, which is used to estimate the first path estimation position τ within a measured CIR with a resolution of 1 / 64Δτs. FP .

[0066] Received CIR measurements

[0067]

[0068] N with corresponding sampling time S Composed of CIR samples

[0069]

[0070] Signal The parameters are calculated relative to the current segmentation:

[0071] y j =z j -h(τ meas，j for j = {0, ..., N} S -1}

[0072] The signal y is used for recursive updating of coefficient h.

[0073]

[0074]

[0075] Where the gain matrix

[0076] The coefficient h is initialized using the first received measurement value z.

[0077]

[0078] For all i∈{0,1,…,N knots -1}, and then update the mean-filtered observation y according to the formula through continuous measurement.

[0079] (3) Perform variance filtering on the y-values ​​after mean filtering;

[0080] The variance is parameterized as a piecewise constant function with coefficients. Current CIR variance update based on

[0081]

[0082] Where, l∈{0,1,…,N} knots -2},τ l ≤τ j <τ l+1 and gain

[0083] To distinguish the variation in the tracking filter CIR from variations caused by parameter errors, system defects, and environmental background noise, the background variation coefficient is set to... Tracking and updating the change coefficient is set to That is, the background CIR variance is updated to

[0084]

[0085] in,

[0086] (4) Subtract the variance between the background average cumulative CIR and the current average cumulative CIR, and then perform front-end detection on the variance difference s to obtain the target time delay τ. TP The estimated value

[0087] Using the variances of the currently measured CIR and the background CIR, areas with temporarily high variance in the CIR can be detected by subtracting the background variance from the current variance.

[0088]

[0089] β>1 is a constant scalar sensitivity factor. Find τ in the region where the CIR is higher than the typical variance. l ≤τ<τ l+1 The section, of which The position τ of the target path TP The measurement is taken by finding the first cluster value in this segment, i.e.

[0090]

[0091] Where, l∈{0,1,…,N} knots -2}, searching in N win The variance N must be higher than the usual variance for the given window length. seg The specific time delay value of the number.

[0092] (5) Based on the estimated target delay That is, the flight time of the signal reflected from the target, and the UWB position, via τ TP =(||p tx -p T ||+||p rx -p T The location information of the perceived target is obtained by finding the intersection of the ellipses ||) / c.

[0093] The applicable conditions for this method are:

[0094] In a real UWB sensing environment, the signal observation range is not an ideal environment, but the environment should be kept stable for a certain period of time, without any additional environmental noise, and the people or objects being sensed should be in a low-speed motion state.

[0095] UWB sensing devices are required to: locate their own position relative to other devices, synchronize their clocks with the network clock, and locate personnel within the space covered by the network.

[0096] This method makes full use of the information obtained from the transmitting and receiving equipment and the properties of the CIR signal itself. It uses mean filtering algorithm, background subtraction, front estimation and particle filtering to obtain a relatively accurate position of the moving target.

[0097] Here is a more specific example:

[0098] like Figure 1 , 2 As shown, a moving target sensing method based on UWB CIR signals includes the following steps:

[0099] (1) The CIR was measured within a certain time period. The acquisition time period was Δt = 27ms, the sampling frequency was fc = 499.2MHz, and the resolution was Δτs = 1 / (2fc) ≈ 1ns. 50 CIR measurements were obtained using two DWM1000s and the cumulative average was performed. Some outliers were directly rejected.

[0100] Outlier types include:

[0101] 1) CIR measurements must be based on a certain number of cumulative preamble symbols, which must be at least half the number of preamble symbols transmitted;

[0102] 2) At the estimated first path position τ FP The previous CIR sample points must be smaller than a certain multiple of the first node coefficient h0 representing the intrinsic noise level. If the estimated first path position τ FP It's too late; the estimated CIR samples before the first path may be higher than the inherent noise level.

[0103] 3) The maximum sampling point of the measured CIR must not deviate too far from the maximum nodal coefficient;

[0104] 4) Other factors may cause errors in CIR measurements, such as the DWM1000's leading edge detection algorithm failing to correctly detect the first path position τ. FP Or, when two modules are transmitting simultaneously, data packets may collide.

[0105] (2) Apply mean filtering to the measured value z of the average cumulative CIR to obtain the result y;

[0106] With coefficient The form that tracks the average value of the currently measured CIR is a piecewise linear function with uniform intervals, i.e.

[0107]

[0108] Where τ i ≤τ<τ i +1, and knots of τ i The piecewise linear parameters are

[0109] τ i =τ start +Δτ knots*i for i = {0, ..., N} knots -1}

[0110]

[0111] Δτ knots It is the CIR measurement sampling period Δτ s divisible by, i.e.

[0112]

[0113] The minimum spacing m = 64 is given by the resolution of the leading edge detection algorithm of the DWM1000. This algorithm is used to estimate the first path estimation position τ within a measured CIR with a resolution of 1 / 64Δτs. FP To estimate the inherent noise level, τ start Choose a location before the first path position, where h0 represents the inherent noise level. The CIR begins to rise approximately 1 ns before the estimated first path position; therefore, τ... FP -τ start It must be greater than 1 ns. To increase the safety margin, τ can be chosen. FP -τ start = 4ns.

[0114] The received CIR measurement value

[0115]

[0116] This measurement is derived from N with the corresponding sampling time. S Composed of CIR samples, i.e.

[0117]

[0118] Signal The parameters are calculated relative to the current segmentation:

[0119] y j =z j -h(τ meas，j ), j = {0, ..., N} S -1}

[0120] This signal y is used to recursively update the coefficient h:

[0121]

[0122]

[0123] Wherein, the gain matrix

[0124] The coefficient h is initialized using the first received measurement value z.

[0125]

[0126] For all i∈{0,1,…,N knots -1}, and then update the mean-filtered observation y according to the formula through continuous measurement.

[0127] (3) Perform variance filtering on the y-values ​​after mean filtering;

[0128] The variance is parameterized as a piecewise constant function with coefficients. Current CIR variance update based on

[0129]

[0130] Where, l∈{0,1,…,N} knots -2},τ l ≤τ j <τ l+1 and gain

[0131] To distinguish the variation in the tracking filter CIR from variations caused by parameter errors, system defects, and environmental background noise, the background variation coefficient is set to... Tracking and updating the change coefficient is set to That is, the background CIR variance is updated to

[0132]

[0133] in,

[0134] (4) Subtract the variance between the background average cumulative CIR and the current average cumulative CIR, and then perform front-end detection on the variance difference s to obtain the target time delay τ. TP The estimated value

[0135] Using the variances of the currently measured CIR and the background CIR, areas with temporarily high variance in the CIR can be detected by subtracting the background variance from the current variance.

[0136]

[0137] β>1 is a constant scalar sensitivity factor. Find τ in the region where the CIR is higher than the typical variance. l ≤τ<τ l+1 The section, of which The position τ of the target path TP The measurement is taken by finding the first cluster value in this segment, i.e.

[0138] and

[0139] Where, l∈{0,1,…,N} knots -2}, searching in N win The variance N must be higher than the usual variance for the given window length. seg The specific time delay value of the number.

[0140] (5) The target path position τ within the CIR is known. TP That is, the flight time of the signal reflected from the target, and the UWB position, via τ TP =(||p tx -p T ||+||p rx -p T The location information of the sensing target can be obtained by finding the intersection of the ellipses ||) / c. Specifically, the least squares method or the particle filter algorithm can be used. The particle filter algorithm is as follows:

[0141] Assume the motion of the particles involves a random walk, and the position of each particle p is p. p =(x p ,y p Assuming

[0142]

[0143]

[0144] Δt is the time period since the last forecast, where in each forecast step... Taken from zero mean, with a standard deviation of σ η The normal distribution. When a message is received, the particle weights will be adjusted according to the measurement. Update the possible conditions at that time, that is

[0145]

[0146] Where, assuming The measurement error follows a Cauchy distribution centered at zero with a scaling parameter γ. This distribution is more representative of the actual measurement error distribution than the normal distribution. (Measured target path location) Included in the message, and the expectation for each particle p. Calculate according to the following formula:

[0147]

[0148] Receiver position p rx and transmitter p txThis can be extracted from both the current message and previously received messages. Furthermore, for each received message, a CIR measurement is obtained, which can also be filtered and used to update the particle filter. In all N... p After the weights of each particle are calculated, the particles are resampled to obtain N. p In the later stages, all particles have equal weight.

[0149] In summary, this invention uses information from the transmitting and receiving UWB devices to solve for the time delay information τ through algorithms such as filtering and background subtraction. TP This invention perceives and estimates the position of a moving target in a moving environment by sensing the time delay information of the target. The method is simple and easy to implement, effectively resists interference, and provides superior real-time detection and identification of living targets in practical applications.

Claims

1. A moving target sensing method based on UWB CIR signals, characterized in that, Includes the following steps: Step 1: Measure the CIR over a certain period and perform a cumulative average to obtain the averaged CIR measurement value z. Outliers are rejected directly. Outliers are identified as follows: (1) Not enough cumulative preamble was collected before CIR measurement. The number of symbols in the cumulative preamble is at least half the number of transmitted preambles. (2) At the estimated first path location The previous CIR sample points were not less than 5 times the first node coefficient h0 representing the inherent noise level; (3) The maximum number of sampling points for the measured CIR shall not be less than 5 times the maximum node coefficient, or not greater than 2 times the maximum node coefficient; Step 2: Apply mean filtering to the CIR measurement value z to obtain the filtered signal y; the specific method is as follows: With coefficient The form tracks the current CIR measurement z, forming a piecewise linear function with uniform intervals: in, ≤τ< +1, where τ represents the specific measured value. This represents the minimum average interval after each sampling of the measured value; indivual The piecewise linear parameters are: It is the CIR measurement sampling period divisible by, i.e. m∈{1,2,⋯,64}; This represents the initial measurement value. This indicates the final measurement value; The CIR measurement value z is then expressed as Signal The calculation method relative to the current segment parameter is as follows: This signal y is used to recursively update the coefficient h: Wherein, the gain matrix ; The coefficient h is initialized using the first received measurement value z. For all i∈{0,1,⋯, The coefficient h is updated by continuous measurement, and the output signal y is generated. Step 3: Perform variance filtering on the CIR measurement value z of the background when the target is not present and the y value after the target appears to obtain the CIR variance value; the specific method is as follows: The CIR variance is parameterized as a piecewise constant function with coefficients. The current CIR variance is updated according to the following formula: Where, l∈{0,1,⋯ }, Gain ∈(0,1); Set the background change coefficient to The tracking and updating change coefficient is set to That is, the background CIR variance value is updated to in, ≪ ; Step 4: Subtract the background CIR measurement value z from the current CIR variance value to obtain the variance difference value s. Then, perform front-end detection on s to obtain the target time delay. The estimated value The specific method is as follows: Using the current CIR measurement z and the background CIR variance, areas with temporarily high CIR variance are detected by subtracting the background variance from the current variance: in, >1 is a constant scalar sensitivity factor; CIR is higher than the typical variance Find in the area The section, of which >0; Target latency The measurement is taken by finding the first cluster value in this segment, i.e. Where, l∈{0,1,⋯ }; Searching in It has a higher variance than usual at window lengths. The specific time delay value of the number; Step 5, based on the relational formula =( ) / c, using the estimated value from step 4 The target position p is obtained by intersecting the ellipses. T In the formula, p tx It is the location of the base station, p rx is the label position, c is the speed of light, and || represents the distance to be calculated.

Images