Low-complexity distance-dependent time domain resampling technology for high-precision sensing of large-scale motion of FMCW (Frequency Modulated Continuous Wave) radar

Abstract

The invention relates to a range-dependent RDRT (Range-Dependent Resampling Technology) technology, which comprises the following steps of: firstly, carrying out range-fast Fourier transform on an intermediate frequency signal to obtain a range unit and a minimum range unit of a target at different slow time moments, and then, carrying out range information-based RDRT on an original intermediate frequency (IF) signal at different slow time moments. Through the technology, a plurality of distance units of motion migration are aligned into one unit, and phase errors caused by distance unit conversion are effectively eliminated. The effectiveness of the technology is evaluated through simulation and experiments, and the result shows that the root mean square error (RMSE) of motion recovery is improved by 2-3 orders of magnitude. The achievement highlights the importance of the method in large-scale motion sensing, and has great value in the field of FMCW millimeter wave radar displacement sensing.

Description

Technical Field

[0001] This invention relates to a technology in the field of millimeter-wave radar, specifically a large-scale motion sensing technology for frequency modulated continuous wave (FMCW) radar. By using range-dependent resampling technique (RDRT), phase errors caused by range cell migration due to large-scale motion are eliminated, thereby enabling FMCW radar to achieve high-precision motion sensing of large-scale displacement. Background Technology

[0002] Frequency-modulated continuous wave (FMCW) radar has been widely used in autonomous driving, vital sign monitoring, and motion tracking due to its high precision, low power consumption, and robustness. However, traditional techniques are prone to introducing phase error problems when dealing with large-scale target motion. Specifically, when the target motion crosses multiple range cells, due to spectral leakage, traditional techniques introduce phase errors caused by range cell migration when extracting phase from fixed cells. These errors accumulate as the range of range cell migration increases, leading to a decrease in the accuracy of motion information recovery. Existing solutions have inherent flaws. For example, to reduce the impact of range cell migration, existing techniques typically expand the range cell coverage by reducing the effective bandwidth, but this significantly reduces resolution. Furthermore, methods that track and extract target range cells separately cannot eliminate phase errors, hindering subsequent signal processing. Phase compensation techniques, on the other hand, require additional memory space and computational complexity. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes a range-dependent resampling technique. By performing resampling steps on the original intermediate frequency signal of the target at different slow-time points, specifically related to the target's distance at each slow-time node, the motion spanning multiple range units is aligned to a single fixed unit, thereby achieving high-precision large-scale motion sensing. This process does not sacrifice the system's effective bandwidth, i.e., it does not reduce the system's range resolution. Furthermore, it avoids phase errors and subsequent signal processing obstacles caused by individually tracking and extracting target range units. Moreover, this process only involves the Fast Fourier Transform complexity required by traditional range sensing, without introducing additional computational complexity and memory space. This provides a high-precision displacement sensing scheme suitable for FMCW radar systems that is efficient, computationally inefficient, and easy to process in subsequent signal processing.

[0004] This invention is achieved through the following technical solution: Step 1) Target range cell extraction at different slow time points: Perform range-fast Fourier transform on the fast time dimension of the intermediate frequency (IF) signal of the FMCW radar system to obtain the range information corresponding to the target at each slow time point, denoted as R[bin]. n ], and the minimum distance unit number R[bin min ], where n is the slow-time index, laying the foundation for distance-related resampling techniques; Step 2) Distance-related resampling technique: Based on the R[bin] corresponding to the target at each slow time step n ] and R[bin min Resampling is performed on different slow-time intermediate frequency signals, specifically, 4πBR[bin n ]m / Mc ⇒ 4πBR[bin n ]m / Mc×R[bin min ] / R[bin n ] = 4πBR[bin min ]m / Mc, 4πBRm / Mc is the frequency term of the FMCW radar intermediate frequency signal, where M is the number of sampling points in the fast time dimension of the original signal, B is the modulation bandwidth, m is the fast time dimension index, i.e., the sampling point index, and c is the speed of light. It should be noted that the resampling process is equivalent to changing the original sampling interval, i.e., the coefficient 4πBR / Mc multiplied by m in the frequency term, specifically 4πBR[bin n ] / Mc becomes 4πBR[bin min Therefore, the distance information of the intermediate frequency signal at time n (slow time) can be adjusted so that the distances corresponding to different slow times are aligned to R[bin]. min From this point on, the migration of distance cells can be removed, and the target will be aligned to a fixed distance cell; Step 3) Phase extraction and motion demodulation: Phase information is extracted and motion demodulated in the slow time dimension corresponding to the fixed distance cell, thereby eliminating the phase error caused by the cross distance cell. In addition, this technique operates on the entire original signal data matrix, which facilitates subsequent signal processing of the data matrix. It has the characteristics of high efficiency, low computational complexity and ease of subsequent processing.

[0005] Technical effect This invention comprehensively addresses the shortcomings of existing technologies in processing large-scale, cross-range cell displacement sensing by FMCW radar, including phase errors, resolution loss, high complexity, large memory burden, and inconvenience for subsequent processing. Specifically, it performs a Fast Fourier Transform on the fast time dimension corresponding to each slow time node of the FMCW radar's intermediate frequency (IF) signal to obtain the range spectrum, and extracts the range information R[bin] corresponding to the target at each slow time moment. n] and the minimum value R[bin] among all distance cells of the target min Based on the two methods, resampling techniques are applied to the intermediate frequency signals at different slow time points n. Specifically, 4πBR[bin] n ]m / Mc ⇒ 4πBR[bin n ]m / Mc×R[bin min ] / R[bin n ] = 4πBR[bin min ]m / Mc, correcting the sampling interval to align distance information corresponding to different slow times to the same distance R[bin min This method completely eliminates the phase error caused by range cell migration. While maintaining high resolution, it effectively extends motion sensing accuracy, avoids the impact of separately tracking and extracting target range cells on subsequent signal processing problems, significantly improves processing efficiency, and does not introduce an additional phase compensation matrix that would increase memory burden. Simulations and experiments show that the root mean square error (RMSE) of motion sensing using this method is reduced by 2-3 orders of magnitude compared to traditional methods, and can be widely applied to high-precision, large-scale displacement sensing in FMCW radar. Attached Figure Description

[0006] Figure 1 This is a flowchart of the present invention; Figure 2 This is a schematic diagram illustrating the scenario, system, and distance migration problem of the present invention; Figure 3 This is a flowchart of the technology of the present invention; Figure 4 The figure shows the simulation results of this invention; In the figure: (a) is the distance spectrum after processing with traditional technology, (b) is the distance spectrum after processing with distance-related resampling technology, (c) is a comparison of demodulation displacement after processing with different technologies, and (d) is a comparison of demodulation displacement error after processing with different technologies. Figure 5 This is a diagram of the experimental scenario for this invention; Figure 6 Figure showing the experimental results of this invention. In the figure: (a) is the distance spectrum after processing with traditional technology, (b) is the distance spectrum after processing with distance-related resampling technology, (c) is a comparison of demodulation displacement after processing with different technologies, and (d) is a comparison of demodulation displacement error after processing with different technologies. Detailed Implementation

[0007] like Figure 2The diagram illustrates the scenario, system, and range migration problem. The system is an Infineon BGT60TR13C FMCW radar module. The raw data of the acquired intermediate frequency (IF) signal is transmitted to a PC. For FMCW radar, traditional techniques typically demodulate motion by selecting the range cell corresponding to the target location and extracting the phase along the slow time of that cell. This method performs well in small-scale motion scenarios, where the target's displacement is confined to a single range cell, enabling accurate motion sensing. However, challenges arise when the target traverses multiple range cells. In this case, phase errors are introduced due to spectral leakage from the target's range cell at certain slow-time instants. To address the range cell migration problem, digital downsampling is commonly used to reduce the effective bandwidth, thereby increasing the coverage of a single range cell. By confining the target motion to a single range cell, phase error interference can be effectively reduced. However, this method inevitably leads to a decrease in resolution. Furthermore, although the correct motion can be partially recovered by tracking and extracting target cells at different slow-time instants, slight errors still occur due to the presence of phase errors, and independently extracting cells is not convenient for subsequent processing of the data matrix containing the target. In addition, while phase compensation techniques can mitigate the impact of this error, they require additional memory space for the phase compensation matrix and incur additional computational complexity.

[0008] like Figure 3 The diagram shown is a flowchart of the range-related resampling technique. Specifically, the modulated linear frequency modulated signal transmitted by the FMCW radar can be represented as: Where t represents fast time, A represents the signal amplitude, and f c The center frequency is represented by γ, and the modulation slope is represented by B / Tc, where B represents the bandwidth and Tc represents the center frequency. c φ0 represents the pulse repetition time (PRT), and φ0 represents the initial phase. When an electromagnetic wave interacts with a target, it causes scattering, and the received signal can be expressed as: S Rx (t) = σS Rx (t - Δt), where σ represents the attenuation factor, Δt is the time delay, defined as Δt = 2R(τ) / c, where R(τ) represents the radial distance between the target and the radar, τ is often referred to as "slow time," and c represents the speed of light. Furthermore, an intermediate frequency (IF) signal is generated by mixing the received signal with the local oscillator (LO) signal.

[0009] The intermediate frequency (IF) signal mentioned above is specifically, , where 4πγR 2 (τ) / c 2This is residual phase (RVP), and its effect can be considered negligible. The intermediate frequency (IF) signal is sampled and organized into... Figure 3 The two-dimensional matrix shown in (a) has rows corresponding to slow time and columns corresponding to fast time. To obtain the distance spectrum, a Fast Fourier Transform (FFT) is performed on the fast time, as shown below. Figure 3 As shown in (b).

[0010] The fast time-dimensional Fast Fourier Transform, specifically, is... Where, sinc(x) = sin(x) / x, f r Indicates frequency. For S 1DFFT Each row yields the spectrum of the intermediate frequency (IF) signal, where f r There is a peak at = 2γR(τ) / c. It can be observed that this frequency is related to the target's distance, which can be expressed as R(τ) = cf. r / 2γ. When the target's range of motion is limited to a single range unit, i.e. When, the range cell corresponding to the target can be selected, where f r = 2γR(τ) / c = f b Then, the phase along the slow time is extracted, and subsequently, the target's motion R(τ) is obtained through phase demodulation.

[0011] The movement of the target is specifically... , where ψ(τ) is S 1DFFT (f b The phase of τ). However, when the target undergoes large-scale motion, extracting the phase along the slow time from a fixed range cell will introduce additional phase errors due to spanning multiple range cells, such as Figure 3 As shown in (b). To analyze this problem, [the following is discussed]... Sampling is performed and a Fast Fourier Transform (FFT) is applied.

[0012] The aforementioned additional phase error specifically refers to: Where k represents the frequency domain index, and n and m represent the indices of the slow and fast time dimensions, respectively. As can be seen from the second term, when the range cell corresponding to the target motion is inconsistent with the selected cell, additional phase errors are introduced. These errors significantly affect the high-precision recovery of large-scale motion, and this effect gradually intensifies as the number of cells traversed by the target increases. To eliminate phase errors and achieve high-precision recovery of large-scale motion, a range-related resampling technique is proposed. After the initial fast Fourier transform of the range, the range information corresponding to the target's motion range at different slow time moments is identified, i.e., R[bin]. nThe minimum distance is denoted as R[bin]. min ], bin n This represents the distance bin where the target is located at time n (slow time). Then, for S... IF_Sampling The intermediate frequency (IF) signals corresponding to different slow times n in [n, m] undergo a resampling step, such as... Figure 3 As shown in (c).

[0013] The resampling step, specifically, is as follows: ,like Figure 3 As shown in (c). After processing, the sampling interval of the intermediate frequency signal corresponding to different slow times changes, from 4πBR[bin] n ] / Mc becomes 4πBR[bin min Therefore, it is possible to effectively transform the large-scale motion of the target across multiple distance units into a single fixed unit, thereby aligning all units corresponding to slow time moments to R[bin]. min ],like Figure 3 As shown in (d). It should be noted that since resampling needs to be performed within the original sampling range, 4πBR[bin] is selected. min The ] / Mc interval is used as a reference because its corresponding sampling range is the smallest, and the sampling range of the intermediate frequency signal at other slower time points can cover this range. Subsequently, by adjusting R[bin min A slow-time phase extraction and demodulation method is used to recover large-scale motion. This method effectively eliminates phase errors and is of great significance for achieving high-precision sensing of large-scale motion.

[0014] like Figure 4 The figure shows the simulation results of range-dependent resampling technology. Specifically, for an FMCW radar, the parameters are set as follows: f = 60 GHz, B = 4 GHz, N... chirp = 6000, M sampling = 512, T c = 10ms, signal-to-noise ratio (SNR) of 20 dB. The target was placed 1 meter in front of the radar and moved away from the radar at a speed of 10 cm / s for 60 seconds. Both the conventional method and the proposed method were used to generate [data / data]. Figure 4 (a) and Figure 4 (b) shows the range profile. Analysis reveals that the target's large-scale motion causes it to traverse multiple range cells at different slow time intervals, thus affecting motion recovery. In contrast, the proposed technique effectively aligns the target's range to R[bin]. min This achieves accurate demodulation, such as... Figure 4 (c) and Figure 4As shown in (d), the accumulation of phase error significantly reduces the accuracy of conventional methods as the number of range cells traversed by the target increases. However, the proposed technique effectively mitigates these errors, achieving accurate recovery of large-scale motion. The root mean square error (RMSE) of the recovered motion is 109.18 cm (conventional method), 0.0111 cm (proposed method), 55.01 cm (target tracking technique), and 0.0162 cm (phase compensation technique), showing an improvement of three orders of magnitude (approximately 1000 times) compared to conventional methods. Furthermore, it offers advantages over other existing techniques, including lower complexity, no additional memory burden, and higher accuracy.

[0015] like Figure 5 The image shows an experimental scenario for distance-related resampling technology. Specifically, the slide is located 0.9 meters from the radar and moves within a 30-centimeter range at a speed of 5 cm / s, with an accuracy of 10 micrometers. Data acquisition is performed using an Infineon BGT60TR13C module.

[0016] like Figure 6 The figure shows the experimental results of distance-related resampling techniques. Specifically, the data were analyzed using both traditional methods and the proposed method, generating [data / results / results / etc.]. Figure 6 (a) and Figure 6 The distance spectrum in (b) shows the displacement demodulation performed. Figure 6 (c) and Figure 6 As shown in (d), the root mean square error (RMSE) of the recovered motion is 3.32 cm (traditional method) and 0.0064 cm (proposed method), respectively, showing an improvement of two orders of magnitude (approximately 500 times) compared to the traditional method. Other techniques have been compared and verified in the simulation section, and their advantages have been verified. The experimental section will not provide a detailed comparison.

[0017] Compared with existing technologies, a range-related resampling technique is proposed. This technique uses a range-fast Fourier transform (RF) to estimate the target's range information at different slow-time points, recording the target's range at different times and its minimum range. Then, the RF signal at different slow-time points is resampled using the target's range information, aligning large-scale motion that previously spanned multiple range cells to the minimum range cell. The slow-time dimension phase information of this range cell is then extracted and demodulated to recover the target's large-scale motion. This technique achieves accurate large-scale motion estimation and has advantages such as low computational complexity, no additional memory consumption, and no impact on subsequent signal processing. Experiments verify the effectiveness and efficiency of this technique. Compared with traditional techniques, this technique improves the RMSE by 2-3 orders of magnitude in large-scale displacement recovery, which is of great significance in FMCW millimeter-wave radar displacement sensing applications.

[0018] In summary, this invention adaptively resamples each intermediate frequency signal by using the range information corresponding to the intermediate frequency signal of the FMCW radar. This enables the alignment of the cross-range cell motion caused by large-scale displacement into a fixed range cell, providing a basis for high-precision large-scale displacement demodulation. This technology effectively removes the phase error introduced by the cross-range cell motion caused by large-scale displacement, which is of great significance to the field of displacement sensing in FMCW millimeter-wave radar.

[0019] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A range-related resampling method for FMCW radar, which obtains the range cell and minimum range cell of the target at different times by performing fast Fourier transform on different slow time dimensions of the original FMCW radar signal data matrix, and resamples the intermediate frequency signal at different slow time times, providing a basis for aligning large-scale displacements to a fixed range cell.

2. The FMCW radar range-related resampling method according to claim 1, characterized in that, The modulated linear frequency modulated signal transmitted by the FMCW radar can be expressed as: Where t represents fast time, A represents the signal amplitude, and f c Let φ0 represent the center frequency, γ be the modulation slope, which can be expressed as B / Tc, where B represents the bandwidth, Tc represents the pulse repetition time (PRT), and φ0 represents the initial phase. When an electromagnetic wave interacts with a target, it causes scattering, and the received signal can be expressed as: S Rx (t) = σS Rx (t - Δt), where σ represents the attenuation factor, Δt is the time delay, defined as Δt = 2R(τ) / c, where R(τ) represents the radial distance between the target and the radar, τ is often referred to as "slow time," and c represents the speed of light. Furthermore, an intermediate frequency (IF) signal is generated by mixing the received signal with the local oscillator (LO) signal.

3. The FMCW radar range-related resampling method according to claim 1, characterized in that, , where 4πγR 2 (τ) / c 2 The residual phase (RVP) can be considered negligible. The intermediate frequency (IF) signal is sampled and organized into a two-dimensional matrix, where rows correspond to slow time and columns correspond to fast time. To obtain the distance spectrum, a Fast Fourier Transform (FFT) is performed on the fast time.

4. The FMCW radar range-related resampling method according to claim 1, characterized in that, Where, sinc(x) = sin(x) / x, f r Indicates frequency. For S 1DFFT Each row yields the spectrum of the intermediate frequency (IF) signal, where f r There is a peak at = 2γR(τ) / c. It can be observed that this frequency is related to the target's distance, which can be expressed as R(τ) = cf. r / 2γ. When the target's range of motion is limited to a single range unit, i.e. When, the range cell corresponding to the target can be selected, where f r = 2γR(τ) / c = f b Then, the phase along the slow time axis is extracted. Subsequently, the target's motion R(τ) is obtained.

5. The FMCW radar range-related resampling method according to claim 1, characterized in that, , where ψ(τ) is S 1DFFT (f b The phase of the target (τ). However, when the target undergoes large-scale motion, extracting the phase along the slow time from a fixed range cell introduces additional phase errors due to spanning multiple range cells. To analyze this problem, the intermediate frequency signal is sampled and a Fast Fourier Transform (FFT) is applied.

6. Additional phase error, specifically: ,in, k represents the frequency domain index, and n and m represent the indices for the slow and fast time dimensions, respectively. As can be seen from the second term, when the range cell corresponding to the target motion is inconsistent with the selected cell, additional phase errors are introduced. These errors significantly affect the high-precision recovery of large-scale motion, and this effect gradually intensifies as the number of cells traversed by the target increases.

7. The method for eliminating additional phase error according to claim 6, characterized in that, In order to eliminate the phase error and achieve high-precision recovery of large-scale motion, a distance-dependent resampling technique is proposed. After the initial range fast Fourier transform, the distance information corresponding to the motion range of the target at different slow time instants is identified, i.e., R[bin n ], where the smallest distance is denoted as R[bin min ], and bin n represents the distance bin where the target is located at the slow time n. Subsequently, the resampling step is performed on the intermediate frequency (IF) signals corresponding to different slow time n in S IF_Sampling [n, m].

8. The method for eliminating additional phase error according to claim 6, characterized in that, 4πBR[bin n ]m / Mc ⇒ 4πBR[bin n ]m / Mc×R[bin min ] / R[bin n ] = 4πBR[bin min ]m / Mc. After processing, the sampling interval of the intermediate frequency signal corresponding to different slow times changed, from 4πBR[bin n ] / Mc becomes 4πBR[bin min Therefore, it is possible to effectively transform the large-scale motion of the target across multiple distance units into a single fixed unit, thereby aligning all units corresponding to slow time moments to R[bin]. min It should be noted that, since resampling needs to be performed within the original sampling range, 4πBR[bin] is selected. min The ] / Mc interval is used as a reference because its corresponding sampling range is the smallest, and the sampling range of the intermediate frequency signal at other slower time points can cover this range. Subsequently, by adjusting R[bin min A slow-time phase extraction and demodulation method is used to recover large-scale motion. This method effectively eliminates phase errors and is of great significance for achieving high-precision sensing of large-scale motion.

Images