Method for reducing ranging data jitter of millimeter wave radar

Abstract

The invention discloses a method for reducing range finding data jitter of a millimeter wave radar, belongs to the technical field of radar signal processing, and aims to reduce target distance data jitter of an FMCW millimeter wave radar in a range finding process. In a target-free state, the definition of a signal is improved by collecting background noise for multiple times, setting a noise threshold value and carrying out de-noising processing on original data. The method comprises the following steps of: weighting an antenna array by utilizing multi-channel antenna data through a designed beam direction and an angle stepping value, generating a distance angle map, and determining a target coarse index position; different from a traditional frequency spectrum refinement algorithm, the method not only depends on frequency spectrum amplitude information, but also combines an improved ZoomFFT algorithm, and uses phase information in a refined frequency spectrum to carry out refinement distance measurement. The use of the phase information can more sensitively capture the distance change of the target, thereby significantly improving the distance measurement precision.

Description

Technical Field

[0001] The present invention belongs to the field of radar signal processing technology, and in particular to a method for reducing jitter of millimeter wave radar ranging data. The method aims to reduce the jitter of target distance data during the ranging process of FMCW millimeter wave radar. Background Art

[0002] High-precision distance measurement is crucial in areas such as autonomous driving and industrial automation. Millimeter-wave radar excels in these applications with its high resolution and strong penetration capabilities. It can provide accurate distance measurement in various environments, especially in bad weather or low light conditions. This makes millimeter-wave radar an ideal choice for meeting the needs of high-precision distance measurement. However, in practical applications, the measurement data of millimeter-wave radar is easily affected by various factors, resulting in jitter.

[0003] The ranging principle of FMCW radar is to convert the propagation time of millimeter waves into the frequency difference between the transmitted signal (Tx) and the received signal (Rx), and calculate the distance of the target by measuring the frequency difference. Spectrum refinement is currently the mainstream way to improve ranging accuracy, but this method will increase the sensitivity to noise while improving the spectrum resolution, resulting in increased data jitter. Therefore, better design of optimization algorithms and reduction of data jitter can improve the effectiveness and reliability of millimeter wave radar ranging in practical applications. Summary of the invention

[0004] The present invention relates to a method for reducing millimeter wave radar ranging data jitter, aiming to improve the accuracy and reliability of target measurement and reduce the jitter of target distance caused by environmental interference, hardware noise and signal processing errors. The method removes background noise, uses multi-channel antennas for beamforming, and combines an improved Zoom-FFT algorithm and phase information for refined ranging, thereby effectively reducing data jitter while ensuring high-precision ranging.

[0005] The present invention proposes a method for reducing millimeter wave radar ranging data jitter, which specifically includes the following aspects:

[0006] 1. Preprocessing of received signals

[0007] In one frame time, the discrete time domain signal X from eight channels is received n [k] (where n = 1, 2, ..., 8 represents the channel index and k represents the sampling point index in the time domain). The time domain signal of each channel is converted into frequency domain data X by fast Fourier transform (FFT). n[m] (where n = 1, 2, ..., 8 represents the channel index, and m represents the frequency point index in the frequency domain). Each channel contains N chirp signals. In order to improve the signal-to-noise ratio, the frequency domain data X of the N chirps in each channel is n,i [m] (i = 1, 2, ..., N represents the i-th chirp index in each channel, n = 1, 2, ..., 8 represents the channel index, m represents the frequency point index in the frequency domain, X n,i [m] represents the data of the ith chirp signal in the nth channel in the frequency domain) and performs arithmetic averaging to obtain the average frequency domain data of the nth channel. For subsequent processing.

[0008]

[0009] 2. Combine beamforming and noise removal to generate distance angle maps

[0010] (1) The present invention mainly measures the distance in fixed scenes. The background noise of the scene will affect the detection accuracy of the target and is a factor of data jitter. In order to improve the quality of the signal, the present invention removes the background noise in the test scene before the beamforming process. In the state without a target, for n channels, the frequency domain data N of the background noise of each channel is collected multiple times (assuming that the collection is Q times) n,q [m] (where n = 1, 2, ..., 8 represents the channel index, q = 1, 2, ..., Q represents the noise data collected for the qth time, and m is the frequency point index in the frequency domain). The noise threshold is set by the following formula to denoise the original data to improve the clarity of the signal. The following are the steps to calculate the threshold based on background noise:

[0011] ① Calculate the arithmetic mean μ of the background noise collected N times at the mth frequency point n [m] (where m is the frequency point index of the frequency domain data, and n=1, 2, ..., 8 represents the channel index).

[0012]

[0013] Where: q = 1, 2, ..., Q represents the noise data collected for the qth time.

[0014] n=1, 2, ..., 8 represents the channel index.

[0015] ② Calculate the standard deviation σ of the background noise collected N times at the mth frequency point n [m] (where m is the frequency point index of the frequency domain data, and n=1, 2, ..., 8 represents the channel index).

[0016]

[0017] Where: q = 1, 2, ..., Q represents the noise data collected for the qth time.

[0018] n=1, 2, ..., 8 represents the channel index.

[0019] ③The arithmetic mean value μ of the passing noise n [m] and standard deviation σ n [m] Calculate the threshold τ at the mth frequency point n [m],

[0020] τ n [m] = μ n [m]+σ n [m]

[0021] ④Use threshold τ n [m] Average frequency domain data of the nth channel signal Noise removal is performed. The specific method is to remove the data of each frequency domain point of each channel. The threshold τ of the corresponding channel and frequency domain point n [m], if the signal strength is lower than the threshold, the part is considered as noise and filtered out, thereby improving the quality of the signal. The final denoised frequency domain signal X n ’ [m].

[0022]

[0023] (2) Using the array signal X after the above denoising n ’ [m], through the designed beam direction and angle step value, the antenna array is weighted, the distance angle spectrum is generated, and the target coarse index position is determined. The specific steps are as follows:

[0024] ① Design the expected beam pointing and calculate the weight value: Here we plan to scan the angle in the interval θ∈[-30°, 30°] with an angle step of 1°. Under each specified beam direction θ, calculate the corresponding weight value W[n,θ] for each channel n.

[0025]

[0026] forθ=-30°,-29°,...,29°,30°

[0027] for n=1,2,...,8

[0028] Wherein: n = 1, 2..., 8 represents the channel number.

[0029] θ represents the pointing angle of the current beam.

[0030] d represents the antenna spacing.

[0031] e is the base of the natural logarithm (Euler's number, approximately equal to 2.718) and is used to represent the change in the phase of a wave.

[0032] j represents the imaginary unit.

[0033] λ represents the wavelength of the signal, which is related to the frequency f of the signal, according to The formula is obtained (where c is the speed of light, and its value is 3x10 8 m / s).

[0034] W[n,θ] represents the beamforming weight.

[0035] ② For the frequency domain signal X after denoising of the nth channel n ’ [m] (where n = 1, 2, ..., 8 represents the channel index, and m represents the frequency point index in the frequency domain), multiplied by the weight W[n,θ] of the corresponding angle, and then the signals of all array elements are added to obtain the received signal in the θ direction. By traversing θ∈[-30°, 30°], the two-dimensional signal Y[m,θ] can be obtained, and the distance-angle spectrum can be obtained. The calculation formula is as follows. The signal Y[m,θ] reflects the target echo signal strength at different distances and angles. By searching the peak of the signal Y[m,θ] after beamforming, the distance and angle index of the target can be found.

[0036] Y[m,θ]=∑ θ ∑ n X n '[m]W[n,θ]

[0037] for θ=-30°,-29°,...,29°,30°

[0038] Where: θ represents the pointing angle of the beam, θ∈[-30°, 30°].

[0039] n=1,2...,8 represents the channel number.

[0040] 3. Using phase information in the improved ZoomFFT spectrum for refined ranging

[0041] The traditional spectrum refinement algorithm mainly relies on the amplitude information of the spectrum after refinement to improve the ranging accuracy, ignoring the distance information contained in the phase. When the distance of the target changes slightly, the change of phase is more sensitive than that of frequency, so by analyzing the phase information of the peak point, a more accurate ranging result can be obtained. The present invention overcomes the shortcomings of the traditional spectrum refinement algorithm by combining the improved ZoomFFT phase for ranging, and provides a more accurate ranging method.

[0042] (1) Determine the target direction in the distance-angle spectrum: In the generated distance-angle spectrum Y[m,θ], each point represents the signal strength at different distances and angles. By finding the peak value, the intensity peak of the target signal can be found, thereby determining the rough distance index r of the target. max and direction index θ max .

[0043] (r max ,θ max ) = argMax m,θ (Y[m,θ])

[0044] Where: argMax m,θ It means returning the value of index m and index θ where the maximum value is located.

[0045] r max Indicates the distance index corresponding to when Y[m,θ] reaches its maximum value.

[0046] θ max Indicates the angle index corresponding to when Y[m,θ] reaches its maximum value.

[0047] (2) Keep the distance dimension data Z[m] in the target direction.

[0048] Z[m]=Y[m,θ max ](m represents frequency index)

[0049] (3) Use the improved zoomFFT to subdivide the spectrum and improve the resolution. The specific steps are as follows:

[0050] ① Determine the amplification range: According to the maximum distance index r of the signal max , determine the range that needs to be amplified, that is, the range of the coarse frequency point r:

[0051] r=[r max -a,r max +a]

[0052] Wherein: the value of a needs to be selected according to the processor memory and computing power as well as the specific test situation, and generally a=2 is selected.

[0053] ② Determine the refinement multiple M: ​​The refinement multiple M represents the length of the data after each coarse frequency point m is refined. Through refinement, the frequency resolution of the signal will be increased to M times the original. The refinement multiple set in the present invention is the same as the number of sampling points W of the signal. W is determined in the radar parameter configuration and its value is 256.

[0054] ③ Apply the formula to refine the spectrum: At each coarse distance index r and each refined subdivision distance index point p, apply the following formula to calculate the refined spectrum data

[0055]

[0056] Where: Z[m] is the distance dimension data in the direction of the target.

[0057] r represents the coarse distance index of the target.

[0058] p represents the subdivision distance index of the target, and the value of p is determined by the refinement multiple M.

[0059] M is the refinement factor.

[0060] W represents the number of sampling points of the signal.

[0061] e is the base of the natural logarithm (Euler's number, approximately equal to 2.718) and is used to represent the change in the phase of a wave.

[0062] j represents the imaginary unit.

[0063] By thinning, the spectral resolution of the signal changes from Δf to Δf ′ The improved spectral resolution enables more precise extraction of phase information within this frequency range.

[0064]

[0065] Where: f s Indicates the sampling frequency of the signal.

[0066] W represents the number of signal sampling points.

[0067] Compared with the traditional zoomFFT algorithm, the improved zoomFFT algorithm used in the present invention can provide higher resolution for a specific frequency range with lower computational overhead.

[0068] (4) Based on the refined spectrum Extract the target signal.

[0069] ① After ZoomFFT refinement, a higher resolution spectrum data matrix is ​​obtained

[0070] Where r is the coarse distance index and p is the refined distance index. Next, we need to find the peak position p of the target signal in this refined spectrum data. ′ .

[0071]

[0072] Where: argMax p Returns the value of index p where the maximum value is located.

[0073] ② Calculate the frequency position f of the target after refinement z .

[0074]

[0075] Where: p ′ Indicates the peak position of the target after refinement.

[0076] f s Indicates the sampling frequency of the signal.

[0077] Calculate the rough distance R of the target by the frequency position of the target f .

[0078]

[0079] ③ Where: C represents the speed of light, and its value is 3x10 8 Meters / second.

[0080] T C Represents the period of the signal.

[0081] B represents the signal bandwidth of the radar.

[0082] f z Indicates the frequency position of the target after refinement.

[0083] (5) Calculate the phase correction factor: Calculate the phase value according to the following formula The first term in the formula represents the fundamental phase change of the signal, while the second The nonlinear phase change caused by the linear frequency change is compensated. When calculating the correction phase, the signal propagation time, the starting frequency and the frequency modulation slope are considered to eliminate the errors introduced by ADC sampling and frequency modulation, and a more accurate distance estimation can be obtained.

[0084]

[0085] Where: f is the starting frequency of the signal.

[0086] t is the time at which the ADC samples the signal.

[0087] S is the slope of the signal.

[0088] R f Represents the coarse distance value calculated above.

[0089] (6) Demodulate the original mixed signal: by calculating the actual frequency f z and phase value Generate a reference signal Y[m]. Multiply the distance dimension data Z[m] by the reference signal to remove the known beat frequency component in the mixed signal and obtain the signal Thus obtaining Phase information

[0090]

[0091] Where: f z Indicates the frequency position of the target after refinement.

[0092] f s Indicates the sampling frequency of the signal.

[0093] m represents the frequency index, m=1,2,3...,w (w is the number of sampling points of the signal).

[0094] e is the base of the natural logarithm (approximately equal to 2.718) and is used to represent the change in the phase of the wave.

[0095] j represents the imaginary unit.

[0096]

[0097] Where: A[m] is the signal The amplitude.

[0098] It's a signal phase.

[0099] e is the base of the natural logarithm (approximately equal to 2.718) and is used to represent the change in the phase of the wave.

[0100] j represents the imaginary unit.

[0101] Through phase difference Calculate the distance correction value ΔR corresponding to the phase difference, and add the correction value to the initially measured distance R f Add them together to get a more accurate distance value R.

[0102]

[0103] Where: C represents the speed of light, and its value is 3x10 8 Meters / second.

[0104] f z Indicates the frequency position of the target after refinement.

[0105] R f Indicates the rough distance to the target. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 This is the overall block diagram of the sensor system used in the present invention.

[0107] Figure 2 The array antenna distribution diagram used in the present invention is

[0108] Figure 3 This is a flow chart of combining beamforming and noise removal to generate a distance angle map in the present invention.

[0109] Figure 4 The present invention is a flow chart of using phase information to perform ranging in the improved ZoomFFT spectrum DETAILED DESCRIPTION

[0110] The present invention proposes a method for reducing millimeter wave radar ranging data jitter, which specifically includes the following steps:

[0111] 1. Hardware device configuration

[0112] The overall block diagram of the system used in the present invention is as follows Figure 1 As shown in the figure. The entire distance measurement system is mainly composed of millimeter wave radar sensors, acceleration sensors and NB200 modules. The millimeter wave radar sensor is responsible for transmitting and receiving radar signals, executing algorithm processing and connecting external devices. The acceleration sensor provides calibration information to the distance calibration module in real time by monitoring the changes in the acceleration of the measurement board on three axes, thereby improving the stability of the system test. The NB200 module uploads the device status information and the final test results to the Alibaba Cloud platform.

[0113] The antenna used by the radar sensor module is a high-gain narrow-beam array antenna. The array structure of the transceiver antenna adopts a 4×24 series-fed array antenna structure. The RF front-end circuit uses the AWR1642 main control chip, which is equipped with two sets of transmitting antennas and four sets of receiving antennas. Figure 2 As shown, the horizontal spacing between the antennas is d = 2.09λ. The two antennas transmit in time segments and synthesize eight virtual receiving channels to enhance the system's ability to accurately estimate the target's angular position.

[0114] 2. Radar signal parameter configuration

[0115] The present invention provides two sets of parameter configurations, long distance and long distance, to meet the distance measurement requirements of different distance ranges. The maximum distance measurement range reaches 200 meters, and the distance measurement accuracy reaches 1mm.

[0116] 3. Receive signal acquisition and preprocessing

[0117] In one frame time, the discrete time domain signal X from eight channels is received n [k] (where n = 1, 2, ..., 8 represents the channel index and k represents the sampling point index in the time domain). The time domain signal of each channel is converted into frequency domain data X by fast Fourier transform (FFT). n [m] (where n = 1, 2, ..., 8 represents the channel index, and m represents the frequency point index in the frequency domain). Each channel contains N chirp signals. In order to improve the signal-to-noise ratio, the frequency domain data X of the N chirps in each channel is n,i [m] (i = 1, 2, ..., N represents the i-th chirp index in each channel, n = 1, 2, ..., 8 represents the channel index, m represents the frequency point index in the frequency domain, X n,i [m] represents the data of the ith chirp signal in the nth channel in the frequency domain) and performs arithmetic averaging to obtain the average frequency domain data of the nth channel. For subsequent processing.

[0118]

[0119] 4. Combine beamforming and noise removal to generate distance angle maps

[0120] (1) The present invention mainly measures the distance in fixed scenes. The background noise of the scene will affect the detection accuracy of the target and is a factor of data jitter. In order to improve the quality of the signal, the present invention removes the background noise in the test scene before the beamforming process. In the state without a target, for n channels, the frequency domain data N of the background noise of each channel is collected multiple times (assuming that the collection is Q times) n,q [m] (where n = 1, 2, ..., 8 represents the channel index, q = 1, 2, ..., Q represents the noise data collected for the qth time, and m is the frequency point index in the frequency domain). The noise threshold is set by the following formula to denoise the original data to improve the clarity of the signal. The following are the steps to calculate the threshold based on background noise:

[0121] ① Calculate the arithmetic mean μ of the background noise collected N times at the mth frequency point n [m] (where m is the frequency point index of the frequency domain data, and n=1, 2, ..., 8 represents the channel index).

[0122]

[0123] Where: q = 1, 2, ..., Q represents the noise data collected for the qth time.

[0124] n=1, 2, ..., 8 represents the channel index.

[0125] ② Calculate the standard deviation σ of the background noise collected N times at the mth frequency point n [m] (where m is the frequency point index of the frequency domain data, and n=1, 2, ..., 8 represents the channel index).

[0126]

[0127] Where: q = 1, 2, ..., Q represents the noise data collected for the qth time.

[0128] n=1, 2, ..., 8 represents the channel index.

[0129] ③The arithmetic mean value μ of the passing noise n [m] and standard deviation σ n [m] Calculate the threshold τ at the mth frequency point n [m],

[0130] τ n [m] = μ n [m]+σ n [m]

[0131] ④Use threshold τ n [m] Average frequency domain data of the nth channel signal Noise removal is performed. The specific method is to remove the data of each frequency domain point of each channel. The threshold τ of the corresponding channel and frequency domain point n [m], if the signal strength is lower than the threshold, the part is considered as noise and filtered out, thereby improving the quality of the signal. The final denoised frequency domain signal X n ’ [m].

[0132]

[0133] (2) Using the array signal X after the above denoising n ’ [m], through the designed beam direction and angle step value, the antenna array is weighted, the distance angle spectrum is generated, and the target coarse index position is determined. The specific steps are as follows:

[0134] ① Design the expected beam pointing and calculate the weight value: Here we plan to scan the angle in the interval θ∈[-30°, 30°] with an angle step of 1°. Under each specified beam direction θ, calculate the corresponding weight value W[n,θ] for each channel n.

[0135]

[0136] forθ=-30°,-29°,...,29°,30°

[0137] for n=1,2,...,8

[0138] Wherein: n = 1, 2..., 8 represents the channel number.

[0139] θ represents the pointing angle of the current beam.

[0140] d represents the antenna spacing.

[0141] e is the base of the natural logarithm (Euler's number, approximately equal to 2.718) and is used to represent the change in the phase of a wave.

[0142] j represents the imaginary unit.

[0143] λ represents the wavelength of the signal, which is related to the frequency f of the signal, according to The formula is obtained (where c is the speed of light, and its value is 3x10 8 m / s).

[0144] W[n,θ] represents the beamforming weight.

[0145] ② For the frequency domain signal X after denoising of the nth channel n ’ [m] (where n = 1, 2, ..., 8 represents the channel index, and m represents the frequency point index in the frequency domain), multiplied by the weight W[n,θ] of the corresponding angle, and then the signals of all array elements are added to obtain the received signal in the θ direction. By traversing θ∈[-30°, 30°], the two-dimensional signal Y[m,θ] can be obtained, and the distance-angle spectrum can be obtained. The calculation formula is as follows. The signal Y[m,θ] reflects the target echo signal strength at different distances and angles. By searching the peak of the signal Y[m,θ] after beamforming, the distance and angle index of the target can be found.

[0146] Y[m,θ]=∑ θ ∑ n X n '[m]W[n,θ]

[0147] forθ=-30°,-29°,...,29°,30°

[0148] Where: θ represents the pointing angle of the beam, θ∈[-30°, 30°].

[0149] n=1,2...,8 represents the channel number.

[0150] 5. Using phase information in the improved ZoomFFT spectrum for refined ranging

[0151] The traditional spectrum refinement algorithm mainly relies on the amplitude information of the spectrum after refinement to improve the ranging accuracy, ignoring the distance information contained in the phase. When the distance of the target changes slightly, the change of phase is more sensitive than that of frequency, so by analyzing the phase information of the peak point, a more accurate ranging result can be obtained. The present invention overcomes the shortcomings of the traditional spectrum refinement algorithm by combining the improved ZoomFFT phase for ranging, and provides a more accurate ranging method.

[0152] (1) Determine the target direction in the distance-angle spectrum: In the generated distance-angle spectrum Y[m,θ], each point represents the signal strength at different distances and angles. By finding the peak value, the intensity peak of the target signal can be found, thereby determining the rough distance index r of the target. max and direction index θ max .

[0153] (r max ,θ max ) = argMax m,θ (Y[m,θ])

[0154] Where: argMax m,θ It means returning the value of index m and index θ where the maximum value is located.

[0155] r max Indicates the distance index corresponding to when Y[m,θ] reaches its maximum value.

[0156] θ max Indicates the angle index corresponding to when Y[m,θ] reaches its maximum value.

[0157] (2) Keep the distance dimension data Z[m] in the target direction.

[0158] Z[m]=Y[m,θ max ](m represents frequency index)

[0159] (3) Use the improved zoomFFT to subdivide the spectrum and improve the resolution. The specific steps are as follows:

[0160] ① Determine the amplification range: According to the maximum distance index r of the signal max, determine the range that needs to be amplified, that is, the range of the coarse frequency point r:

[0161] r=[r max -a,r max +a]

[0162] Wherein: the value of a needs to be selected according to the processor memory and computing power as well as the specific test situation, and generally a=2 is selected.

[0163] ② Determine the refinement multiple M: ​​The refinement multiple M represents the length of the data after each coarse frequency point m is refined. Through refinement, the frequency resolution of the signal will be increased to M times the original. The refinement multiple set in the present invention is the same as the number of sampling points W of the signal. W is determined in the radar parameter configuration and its value is 256.

[0164] ③ Apply the formula to refine the spectrum: At each coarse distance index r and each refined subdivision distance index point p, apply the following formula to calculate the refined spectrum data

[0165]

[0166] Where: Z[m] is the distance dimension data in the direction of the target.

[0167] r represents the coarse distance index of the target.

[0168] p represents the subdivision distance index of the target, and the value of p is determined by the refinement multiple M.

[0169] M is the refinement factor.

[0170] W represents the number of sampling points of the signal.

[0171] e is the base of the natural logarithm (Euler's number, approximately equal to 2.718) and is used to represent the change in the phase of a wave.

[0172] j represents the imaginary unit.

[0173] By thinning, the spectral resolution of the signal changes from Δf to Δf ′ The improved spectral resolution enables more precise extraction of phase information within this frequency range.

[0174]

[0175] Where: f s Indicates the sampling frequency of the signal.

[0176] W represents the number of signal sampling points.

[0177] Compared with the traditional zoomFFT algorithm, the improved zoomFFT algorithm used in the present invention can provide higher resolution for a specific frequency range with lower computational overhead.

[0178] (4) Based on the refined spectrum Extract the target signal.

[0179] ① After ZoomFFT refinement, a higher resolution spectrum data matrix is ​​obtained

[0180] Where r is the coarse distance index and p is the refined distance index. Next, we need to find the peak position p of the target signal in this refined spectrum data. ′ .

[0181]

[0182] Where: argMax p Returns the value of index p where the maximum value is located.

[0183] ② Calculate the frequency position f of the target after refinement z .

[0184]

[0185] Where: p ′ Indicates the peak position of the target after refinement.

[0186] f s Indicates the sampling frequency of the signal.

[0187] Calculate the rough distance R of the target by the frequency position of the target f .

[0188]

[0189] ③ Where: C represents the speed of light, and its value is 3x10 8 Meters / second.

[0190] T C Represents the period of the signal.

[0191] B represents the signal bandwidth of the radar.

[0192] f z Indicates the frequency position of the target after refinement.

[0193] (5) Calculate the phase correction factor: Calculate the phase value according to the following formula The first term in the formula represents the fundamental phase change of the signal, while the second The nonlinear phase change caused by the linear frequency change is compensated. When calculating the correction phase, the signal propagation time, the starting frequency and the frequency modulation slope are considered to eliminate the errors introduced by ADC sampling and frequency modulation, and a more accurate distance estimation can be obtained.

[0194]

[0195] Where: f is the starting frequency of the signal.

[0196] t is the time at which the ADC samples the signal.

[0197] S is the slope of the signal.

[0198] R f Represents the coarse distance value calculated above.

[0199] (6) Demodulate the original mixed signal: by calculating the actual frequency f z and phase value Generate a reference signal Y[m]. Multiply the distance dimension data Z[m] by the reference signal to remove the known beat frequency component in the mixed signal and obtain the signal Thus obtaining Phase information

[0200]

[0201] Where: f z Indicates the frequency position of the target after refinement.

[0202] f s Indicates the sampling frequency of the signal.

[0203] m represents the frequency index, m=1,2,3...,w (w is the number of sampling points of the signal).

[0204] e is the base of the natural logarithm (approximately equal to 2.718) and is used to represent the change in the phase of the wave.

[0205] j represents the imaginary unit.

[0206]

[0207] Where: A[m] is the signal The amplitude.

[0208] It's a signal phase.

[0209] e is the base of the natural logarithm (approximately equal to 2.718) and is used to represent the change in the phase of the wave.

[0210] j represents the imaginary unit.

[0211] Through phase difference Calculate the distance correction value ΔR corresponding to the phase difference, and add the correction value to the initially measured distance R f Add them together to get a more accurate distance value R.

[0212]

[0213] Where: C represents the speed of light, and its value is 3x10 8 Meters / second.

[0214] f z Indicates the frequency position of the target after refinement.

[0215] R f Indicates the rough distance to the target.

Claims

1. A method for reducing jitter of millimeter wave radar ranging data, characterized in that: The following steps are involved: 1) Preprocessing of received signals In one frame time, the discrete time domain signal X from eight channels is received n [k], where n = 1, 2, ..., 8 represents the channel index, and k represents the sampling point index in the time domain; the time domain signal of each channel is converted into frequency domain data X through fast Fourier transform. n [m], where n = 1, 2, ..., 8 represents the channel index, and m represents the frequency point index in the frequency domain; each channel contains N chirp signals. In order to improve the signal-to-noise ratio, the frequency domain data X of the N chirps in each channel is n,i [m] is the arithmetic mean, i = 1, 2, ..., N represents the i-th chirp index in each channel, n = 1, 2, ..., 8 represents the channel index, m represents the frequency point index in the frequency domain, X n,i [m] represents the data of the i-th chirp signal in the n-th channel in the frequency domain; Get the average frequency domain data of the nth channel For subsequent processing; 2) Combining beamforming and noise removal to generate distance angle maps (1) Before the beamforming process, the background noise in the test scene was removed; In the state without a target, for n channels, the frequency domain data of the background noise of each channel is collected Q times N n,q [m], where n = 1, 2, ..., 8 represents the channel index, q = 1, 2, ..., Q represents the noise data collected for the qth time, and m is the frequency point index in the frequency domain; the noise threshold is set by the following formula to denoise the original data to improve the clarity of the signal; Here are the steps to calculate the threshold using background noise: ① Calculate the arithmetic mean μ of the background noise collected N times at the mth frequency point n [m] (where m is the frequency point index of the frequency domain data, and n = 1, 2, ..., 8 represents the channel index); Where: q=1,2,...,Q represents the noise data collected for the qth time; n=1,2,...,8 represents the channel index; ② Calculate the standard deviation σ of the background noise collected N times at the mth frequency point n [m], where m is the frequency point index of the frequency domain data, and n=1, 2, ..., 8 represents the channel index; Where: q = 1, 2, ..., Q represents the noise data collected for the qth time; n=1, 2, ..., 8 represents the channel index; ③The arithmetic mean value μ of the passing noise n [m] and standard deviation σ n [m] Calculate the threshold τ at the mth frequency point n [m]; t n [m]=μ n [m]+s n [m] ④Use threshold Tn[m] to average frequency domain data of nth channel signal To remove noise; the specific method is to remove the data of each frequency domain point of each channel The signal is compared with the threshold τn[m] of the corresponding channel and frequency domain point. If the signal strength is lower than the threshold, the part is considered as noise and filtered out, thereby improving the quality of the signal. The final denoised frequency domain signal X n [m]; (2) Using the array signal X after the above denoising n [m], through the designed beam direction and angle step value, the antenna array is weighted to generate a distance angle map and determine the target coarse index position; the specific steps are as follows: ① Design the expected beam pointing and calculate the weight value: Here, we plan to scan the angle in the interval θ∈[-30°, 30°] with an angle step of 1°. Under each specified beam direction θ, calculate the corresponding weight value W[n, θ] for each channel n; for θ=-30°,-29°,...,29°,30° for n=1,2,...,8 Where: n=1,2...,8 represents the channel number; θ represents the pointing angle of the current beam; d represents the antenna spacing; e is the base of the natural logarithm, approximately equal to 2.718, and is used to represent the change in the phase of a wave; j represents the imaginary unit; λ represents the wavelength of the signal, which is related to the frequency f of the signal, according to The formula is found; where c is the speed of light, Its value is 3x10 8 m / s; W[n,θ] represents the beamforming weight; ② For the frequency domain signal X after denoising of the nth channel n '[m], where n = 1, 2, ..., 8 represents the channel index, m represents the frequency point index in the frequency domain, and multiplies them by the weight W[n, θ] of the corresponding angle respectively, and then adds the signals of all array elements to obtain the received signal in the θ direction; by traversing θ∈[-30°, 30°], the two-dimensional signal Y[m, θ] is obtained, and the distance-angle spectrum is obtained. The calculation formula is as follows; the signal Y[m, θ] reflects the target echo signal strength at different distances and angles. By searching the peak of the signal Y[m, θ] after beamforming, the distance and angle index of the target can be found; Y[m,θ]=∑ θ ∑ n X n '[m]W[n,θ] forθ=-30°,-29°,...,29°,30° Where: θ represents the pointing angle of the beam, θ∈[-30°, 30°]; n=1,2...,8 represents the channel number; 3) Using phase information in the improved ZoomFFT spectrum for refined ranging By combining the phase after improved ZoomFFT for ranging, the shortcomings of the traditional spectrum refinement algorithm are overcome, providing a more accurate ranging method; (1) Determine the target direction in the range-angle spectrum: In the generated range-angle spectrum Y[m,θ], each point represents the signal strength at different distances and angles. By finding the peak value, the intensity peak of the target signal is found, thereby determining the rough range index r of the target. max and direction index θ max ; (r max ,the max )=argMax m，θ (Y[m,θ]) Where: argMax m，θ It means returning the value of index m and index θ where the maximum value is located; r max Indicates the distance index corresponding to when Y[m,θ] reaches its maximum value; θ max Indicates the angle index corresponding to when Y[m,θ] reaches its maximum value; (2) Retain the distance dimension data Z[m] in the target direction; Z[m]=Y[m,θ max ](m represents the frequency index) (3) Apply the improved zoomFFT to subdivide the spectrum and improve the resolution; the specific steps are as follows: ① Determine the amplification range: According to the maximum distance index r of the signal max , determine the range that needs to be amplified, that is, the range of the coarse frequency point r: r=[r max -a,r max +a] Where a = 2; ② Determine the refinement multiple M: ​​The refinement multiple M represents the length of the data after each coarse frequency point m is refined; through refinement, the frequency resolution of the signal will be increased to M times of the original; the set refinement multiple is the same as the number of sampling points W of the signal, which is determined in the radar parameter configuration and its value is 256; ③ Apply the formula to refine the spectrum: At each coarse distance index r and each refined subdivision distance index point p, apply the following formula to calculate the refined spectrum data Where: Z[m] is the distance dimension data in the direction of the target; r represents the coarse distance index of the target; p represents the subdivision distance index of the target, and the value of p is determined by the refinement multiple M; M is the refinement factor; W represents the number of sampling points of the signal; e is the base of the natural logarithm used to represent the change in the phase of a wave; j represents the imaginary unit; ④Through refinement, the spectral resolution of the signal changes from Δf to Δf′; the improvement of spectral resolution enables more accurate extraction of phase information within the frequency range; Where: f s Indicates the sampling frequency of the signal; W represents the number of signal sampling points; (4) Based on the refined spectrum Extracting target signals; ① After ZoomFFT refinement, a higher resolution spectrum data matrix is ​​obtained Where r is the coarse distance index, and p is the refined distance index. Next, we need to find the peak position p′ of the target signal in the refined spectrum data. Where: argMax p It means returning the value of index p where the maximum value is located; ② Calculate the frequency position f of the target after refinement z ; Where: p′ represents the peak position of the target after refinement; f s Indicates the sampling frequency of the signal; ③ Calculate the rough distance R of the target by the frequency position of the target f ; Where: C represents the speed of light, and its value is 3x10 8 m / s; T C Indicates the period of the signal; B represents the signal bandwidth of the radar; f z Indicates the frequency position of the target after refinement; (5) Calculate the phase correction factor: Calculate the phase value according to the following formula The first term in the formula represents the basic phase change of the signal, while the second It compensates for the nonlinear phase change caused by the linear change of frequency. When calculating the correction phase, it eliminates the errors introduced by ADC sampling and frequency modulation by considering factors such as signal propagation time, starting frequency and frequency modulation slope, and can obtain a more accurate distance estimation. Where: f is the starting frequency of the signal; t is the ADC sampling time of the signal; S is the slope of the signal; R f Represents the rough distance value calculated above; (6) Demodulate the original mixed signal: by calculating the actual frequency f z and phase value Generate a reference signal Y[m]; multiply the distance dimension data Z[m] by the reference signal to remove the known beat frequency component in the mixed signal and obtain the signal Thus obtaining Phase information Where: f z Indicates the frequency position of the target after refinement; f s Indicates the sampling frequency of the signal; m represents the frequency index, m = 1, 2, 3..., w, w is the number of sampling points of the signal; e is the base of the natural logarithm (approximately equal to 2.718) and is used to represent the change in the phase of the wave; j represents the imaginary unit; Where: A[m] is the signal the magnitude of It's a signal The phase of e is the base of the natural logarithm and is used to represent the change in the phase of a wave; j represents the imaginary unit; Through phase difference Calculate the distance correction value ΔR corresponding to the phase difference, and add the correction value to the initially measured distance R f Add together to get a more accurate distance value R; Where: C represents the speed of light, and its value is 3x10 8 m / s; f z Indicates the frequency position of the target after refinement; R f Indicates the rough distance to the target.

Images