阅读说明：本技术 一种改进调频连续波雷达目标检测的降噪方法 (Noise reduction method for improving frequency modulation continuous wave radar target detection ) 是由 西格弗雷德·博龙 阮洪宁 黄震 于 2020-09-29 设计创作，主要内容包括：本申请涉及一种改进调频连续波雷达目标检测的降噪方法,应用于汽车电子产品中,所述方法包括：获取并处理雷达的回波混频信号,生成原始数据矩阵；将所述原始数据矩阵通过二维自适应滤波器进行滤波；对滤波后的所述原始数据矩阵进行处理。有益效果是：本申请通过二维自适应滤波器来抑制汽车雷达信号中有效目标周围的噪声；因而在距离多普勒图中大大提高目标幅度与周围噪声幅度的对比度。通过增加这种对比度,目标信号功率与周围噪声功率的比率增加,从而让以幅度或者功率为主要特征的目标检测方法,如恒虚警检测法可以更有效地检测到雷达回波弱目标,从而提高检测率。(The application relates to a noise reduction method for improving frequency modulation continuous wave radar target detection, which is applied to automobile electronic products and comprises the following steps: acquiring and processing an echo mixing signal of a radar to generate an original data matrix; filtering the original data matrix through a two-dimensional adaptive filter; and processing the filtered original data matrix. The beneficial effects are that: the method and the device suppress the noise around the effective target in the automobile radar signal through the two-dimensional adaptive filter; thus greatly improving the contrast of the target amplitude to the surrounding noise amplitude in the range-doppler plot. By increasing the contrast, the ratio of the target signal power to the ambient noise power is increased, so that a target detection method using amplitude or power as a main characteristic, such as a constant false alarm detection method, can more effectively detect a radar echo weak target, thereby improving the detection rate.)

1. A noise reduction method for improving frequency modulation continuous wave radar target detection is characterized by being applied to an automotive electronic product provided with a radar, and the method comprises the following steps:

acquiring and processing an echo mixing signal of a radar to generate an original data matrix;

filtering the original data matrix through a two-dimensional adaptive filter;

processing the filtered original data matrix;

the two-dimensional adaptive filter comprises a radial distance filter for filtering a radial distance dimension and a Doppler filter for filtering a Doppler dimension.

2. A method of noise reduction for improved frequency modulated continuous wave radar target detection as claimed in claim 1, wherein obtaining the echo mix signal comprises:

the radar transmits continuous linear frequency modulation signals, echoes of the linear frequency modulation signals are received by the radar and then are mixed with the signals of the transmitted linear frequency modulation signals, and the echo mixed signals are generated after ADC sampling.

3. A method of noise reduction for improved frequency modulated continuous wave radar target detection as claimed in claim 1, wherein processing the echo mixed signal to generate a raw data matrix comprises:

establishing an original data matrix S from the echo mixing signal_n(k, l), wherein n is the nth receiving antenna of the continuous wave radar, l is the l chirp signal, and k is the th chirp signalk sample points.

4. A method of noise reduction for improving frequency modulated continuous wave radar target detection as claimed in claim 1, wherein said filtering said raw data matrix through a two-dimensional adaptive filter comprises:

filtering the signal of the radial distance of the distance dimension by the radial distance filter;

and filtering the Doppler signals of the Doppler dimension through the Doppler filter.

5. A method of noise reduction for improving frequency modulated continuous wave radar target detection according to claim 4, wherein said filtering a signal for a radial distance in a distance dimension by said radial distance filter comprises:

initializing a first input parameter a (k) of the radial distance filter, a (k) being a filter coefficient of length M +1, i.e. a (k) [ [ a ], (k) ]₀(k)，a₁(k)，…，a_M(k)]；

The kth sampling point and the first M +1 sampling points of the k sampling point of the original data matrix of the echo mixing signal of the ith linear frequency modulation signal form an intermediate input sampleThey and the first input parameter a (k) are input into the radial distance filter, and the estimated value y (k) of the k-th sampling time of the real signal is obtained, and the difference between y (k) and x (k, l) generates the error signal e_n；

According to the error signal e_nAnd intermediate input samplesUpdating the first input parameter a (k +1) at the k +1 th moment by an adaptive algorithm;

and performing iterative computation on the echo mixing signal of the linear frequency modulation signal through a first input parameter a (k +1), and outputting a linear frequency modulation signal matrix of primary filtering.

6. A method of noise reduction for improving frequency modulated continuous wave radar target detection as claimed in claim 5, further comprising, after filtering the same chirp of a matrix of chirps by said radial distance filter:

and obtaining radial distance information by performing fast Fourier transform on the filtered linear frequency modulation signal matrix.

7. A method of noise reduction for improving FM continuous wave radar target detection as claimed in claim 4 wherein said filtering adjacent said chirp signals of a matrix of chirp signals by said Doppler filter comprises:

initializing the second input parameter b (l) of the doppler filter b (l) is the filter coefficient b (l) of length P +1 [ b ]₀(l)，b₁(l)，…，b_P(l)]；

The kth sampling point of the original data matrix of the echo mixing signal of the ith chirp signal and the echo mixing signal of the first P +1 chirp signals form an intermediate input sampleThey and a second input parameter b (l) are input into the Doppler filter, and an error signal e is generated by the difference between the real echo mixing signal y (l) of the first chirp signal at the k-th sampling time and y (l) and X (k, l)_n(l)；

According to the error signal e_n(l) And intermediate input samplesUpdating the second input parameters b (l) by an adaptive algorithm;

and (3) performing iterative calculation on the original data matrix by inputting a second parameter b (l), and outputting a linear frequency modulation signal matrix which is filtered again.

8. The method of claim 7, wherein the filtering the doppler signal in the doppler dimension by the doppler filter further comprises:

and calculating the linear frequency modulation signal matrix through fast Fourier transform to obtain Doppler information.

9. A method of noise reduction for improved frequency modulated continuous wave radar target detection as claimed in any one of claims 5 or 7, wherein the adaptive algorithm is a normalized least mean square algorithm or a time varying least mean square algorithm or a least squares method.

10. A method of noise reduction for improved frequency modulated continuous wave radar target detection as claimed in claim 9, wherein said noise reduction is based on said error signal e_nAnd intermediate input samplesUpdating the parameter a (k +1) of the radial distance filter at the time k +1 by an adaptive algorithm, comprising:

by passingCalculated, where Δ is a step constant.

11. A method of noise reduction for improving frequency modulated continuous wave radar target detection as claimed in claim 1, wherein said processing of said filtered raw data matrix comprises:

carrying out incoherent superposition processing on the linear frequency modulation signal matrix of each channel;

carrying out constant-virtual early warning detection on the result after the incoherent superposition processing;

and carrying out target detection on the result after the constant-deficiency early warning detection.

Technical Field

The application relates to the technical field of automotive electronics, in particular to a noise reduction method for improving frequency modulation continuous wave radar target detection.

Background

Automotive mounted radar is one of the important sensors in advanced driving assistance systems. The radar is used for detecting targets near or far away from the vehicle, and the targets can be other vehicles, pedestrians, surrounding stationary targets and the like. The ability of the radar to detect these targets is directly related to the signal-to-noise ratio, SNR, of these targets to their surrounding background noise. The high SNR can effectively reduce the probability of false detection, thereby improving target detection capability.

In general, various target detection algorithms degrade in performance at low signal-to-noise ratios. These conditions, but not limited to, lead to a decrease in signal-to-noise ratio. For example, when the target is at a long distance, the transmission power is low in the energy of the return signal due to the loss of the distance; or the target is close to the edge of the radar field, the included angle between the target and the radar is a large angle, and the gain of the antenna is lower than the zero-degree azimuth (right ahead) of the antenna when the antenna gain is at the large angle, so that the echo energy is reduced; in addition, in the case that the noise of the channel or the noise of the receiver is relatively large, the noise floor is raised, thereby lowering the signal-to-noise ratio. Therefore, increasing the signal-to-noise ratio is a commonly used main means for improving the target detection performance. Improving receiver noise floor may increase signal-to-noise ratio, but the approach by increasing transmit power may not be suitable for power-limited systems, such as vehicle-mounted radar.

Disclosure of Invention

In order to solve the problem of how to improve the signal-to-noise ratio in the prior art, the application provides a noise reduction method for improving the frequency modulation continuous wave radar target detection.

A noise reduction method for improving frequency modulation continuous wave radar target detection is applied to an automotive electronic product provided with a radar, and comprises the following steps:

acquiring and processing an echo mixing signal of a radar to generate an original data matrix;

filtering the original data matrix through a two-dimensional adaptive filter;

processing the filtered original data matrix;

the two-dimensional adaptive filter comprises a radial distance filter for filtering a radial distance dimension and a Doppler filter for filtering a Doppler dimension.

Optionally, acquiring the echo mixed signal includes:

the radar transmits continuous linear frequency modulation signals, echoes of the linear frequency modulation signals are received by the radar and then are mixed with the signals of the transmitted linear frequency modulation signals, and the echo mixed signals are generated after ADC sampling.

Optionally, processing the echo mixed signal to generate a raw data matrix, includes:

establishing an original data matrix S from the echo mixing signal_n(k, l), wherein n is the nth receiving antenna of the continuous wave radar, l is the ith chirp signal, and k is the kth sampling point on the echo signal of the ith chirp signal.

Optionally, the filtering the raw data matrix through a two-dimensional adaptive filter includes:

filtering the signal of the radial distance of the distance dimension by the radial distance filter;

and filtering the Doppler signals of the Doppler dimension through the Doppler filter.

Optionally, the filtering, by the radial distance filter, a signal of a radial distance of a distance dimension includes:

initializing a first input parameter a (k) of the radial distance filter, a (k) being a filter coefficient of length M +1, i.e. a (k) [ [ a ], (k) ]₀(k)，a₁(k)，…，a_M(k)]；

The kth sampling point of the original data matrix of the echo mixing signal of the ith chirp signalAnd the first M +1 sampling points form an intermediate input sampleThey and the first input parameter a (k) are input into the radial distance filter, and the estimated value y (k) of the true signal at the k-th sampling time is obtained, and the difference between y (k) and x (k, l) generates the error signal e_n；

According to the error signal e_nAnd intermediate input samplesUpdating the first input parameter a (k +1) at the k +1 th moment by an adaptive algorithm;

and performing iterative calculation on the echo mixing signals of the linear frequency modulation signals through a first input parameter a (k +1), and outputting a linear frequency modulation signal matrix of primary filtering.

Optionally, after the filtering, by the radial distance filter, the same chirp signal of the chirp signal matrix, the method further includes:

and obtaining radial distance information by performing fast Fourier transform on the filtered linear frequency modulation signal matrix.

Optionally, the filtering, by the doppler filter, the chirp signals adjacent to the chirp signal matrix includes:

initializing the second input parameter b (l) of the doppler filter b (l) is the filter coefficient b (l) of length P +1 [ b ]₀(l)，b₁(l)，…，b_P(l)]；

The kth sampling point of the original data matrix of the echo mixing signal of the ith linear frequency modulation signal and the echo mixing signals of the first P +1 linear frequency modulation signals form an intermediate input sampleThey and a second input parameter b (l) are input into the Doppler filter to obtain an estimate of the true echo mixing signal of the l-th chirp signal at the k-th sampling instantThe difference between the values y (l), y (l) and X (k, l) generates an error signal e_n(l)；

According to the error signal e_n(l) And intermediate input samplesUpdating the second input parameters b (l) by an adaptive algorithm;

and (3) performing iterative calculation on the original data matrix by inputting a second parameter b (l), and outputting a linear frequency modulation signal matrix which is filtered again.

Optionally, after filtering the doppler signal in the doppler dimension by the doppler filter, the method further includes:

and calculating the linear frequency modulation signal matrix through fast Fourier transform to obtain Doppler information.

Alternatively, the adaptive algorithm may be a normalized least mean square algorithm or a time-varying least mean square algorithm or a least squares method.

Optionally, said determining according to said error signal e_nAnd intermediate input samplesUpdating the parameter a (k +1) of the radial distance filter at the time k +1 by an adaptive algorithm, comprising:

by passingCalculated, where Δ is a step constant.

Optionally, the processing the filtered raw data matrix includes:

carrying out incoherent superposition processing on the linear frequency modulation signal matrix of each channel;

carrying out constant-virtual early warning detection on the result after the incoherent superposition processing;

and carrying out target detection on the result after the constant-deficiency early warning detection.

Compared with the prior art, the beneficial effects of this application are: the noise around the effective target in the automobile radar signal is reduced through the two-dimensional adaptive filter; especially in range Doppler maps (range Doppler maps) greatly improve the contrast of the amplitude of the target with the amplitude of the surrounding noise. By increasing this contrast, the ratio of target signal power to ambient noise is increased, allowing amplitude-based target detection algorithms, such as constant false alarm detection (CFAR), to more effectively detect targets with weak radar returns, thereby improving the target detection capabilities of the radar. More scattering points are detected, and the radar can further sense the size of the target by a clustering method, so that more information is provided to help the target classification. At the same time, the present application is done in software, thus eliminating the need for expensive hardware modifications, making it easier to implement on top of existing algorithms.

Drawings

Fig. 1 is a schematic diagram of a method according to an embodiment of the present application.

Fig. 2 is a schematic diagram of a raw data matrix of an echo mixed signal of a chirp signal according to an embodiment of the present application.

Fig. 3 is a block diagram of general radar signal processing with a two-dimensional adaptive filter according to an embodiment of the present application.

Fig. 4 is a block diagram of a two-dimensional adaptive filter for filtering radar data having multiple channels according to an embodiment of the present application.

Fig. 5 is a schematic diagram of a range adaptive filter according to an embodiment of the present application.

Fig. 6 is a schematic structural diagram of a complex doppler filter according to an embodiment of the present application.

FIG. 7 is a model of noise suppression according to an embodiment of the present application

Fig. 8 is a diagram illustrating a filtering process according to an embodiment of the present application. a) RD with strong noise. b) Radial distance filter amplitude response. c) RD map after radial distance filtering. d) The doppler filter magnitude response. e) RD pattern after radial distance filtering and doppler filtering.

Detailed Description

The present application will be further described with reference to the following detailed description.

The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it is to be understood that the terms "upper", "lower", "left", "right", "top", "bottom", "inner", "outer", and the like, if any, are used in the orientations and positional relationships indicated on the basis of the drawings, which are merely for convenience of description and simplicity of description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed in a specific orientation, and be operated, and therefore, the terms describing the positional relationships in the drawings are used for illustrative purposes only and are not to be construed as limiting the present patent.

Furthermore, if the terms "first," "second," and the like are used for descriptive purposes only, they are used for distinguishing different devices, elements, or components (the specific types and configurations may be the same or different), and they are not used for indicating or implying relative importance or quantity among the devices, elements, or components, but are not to be construed as indicating or implying relative importance.

In the embodiment shown in fig. 1, the present application provides a noise reduction method for improving the target detection of a frequency modulated continuous wave radar, which is applied to automotive electronics, and the method includes:

100, acquiring and processing an echo mixing signal of a radar to generate an original data matrix; in step 100, acquiring the echo mixed signal includes: the radar transmits continuous linear frequency modulation signals, echoes of the linear frequency modulation signals are received by the radar and then are mixed with the signals of the transmitted linear frequency modulation signals, and the echo mixed signals are generated after ADC sampling. Processing the echo mixed signal to generate a raw data matrix, comprising: establishing an original data matrix S from the echo mixing signal_n(k, l), wherein n is the nth receiving antenna of the continuous wave radar, l is the l chirp signal, and k is the k sampling point on the echo signal of the l chirp signal.

200, filtering the original data matrix through a two-dimensional adaptive filter; in step 200, the two-dimensional adaptive filter includes a radial distance filter for filtering a radial distance dimension, and a doppler filter for filtering a doppler dimension. The filtering the original data matrix through a two-dimensional adaptive filter includes: filtering the signal of the radial distance of the distance dimension by the radial distance filter; and filtering the Doppler signals of the Doppler dimension through the Doppler filter.

300, processing the filtered original data matrix; in step 300, the filtered raw data matrix is processed, which includes: carrying out incoherent superposition processing on the linear frequency modulation signal matrix of each channel; performing constant-virtual early warning detection on the result after the incoherent superposition processing; and carrying out target detection on the result after the constant-deficiency early warning detection.

In the embodiment, the noise around the effective target in the automobile radar signal is reduced through the two-dimensional adaptive filter; in particular, the contrast between the amplitude of the target and the amplitude of the surrounding noise is greatly improved in the range-doppler plot. By increasing the contrast, the ratio of the target signal power to the ambient noise is increased, so that an amplitude-based target detection algorithm, such as a constant false alarm detection method, can more effectively detect a target with weak radar echo, thereby improving the target detection capability of the radar. More scattering points are detected, and the radar can further sense the size of the target by a clustering method, so that more information is provided to help the target classification. At the same time, the present application is done in software, thus eliminating the need for expensive hardware modifications, making it easier to implement on top of existing algorithms.

In some embodiments, acquiring the echo mixed signal comprises: the radar transmits continuous linear frequency modulation signals, echoes of the linear frequency modulation signals are received by the radar and then are mixed with the signals of the transmitted linear frequency modulation signals, and the echo mixed signals are generated after ADC sampling. In this embodiment, the radar is a frequency modulation continuous wave radar, which is a continuous wave radar whose transmitting frequency is modulated by a specific signal. The frequency modulation continuous wave radar obtains the distance information of the target by comparing the difference between the frequency of the echo signal at any moment and the frequency of the transmitting signal at the moment, and the distance is proportional to the frequency difference between the two frequencies. The radial velocity of the target is linearly related to the acquired doppler frequency. Compared with other distance and speed measuring radars, the frequency modulation continuous wave radar has a simpler structure. The echo mixing signal is a signal obtained by mixing an echo of a chirp signal with a transmitting chirp signal after being received by a radar, and is sampled by an analog-to-digital converter (ADC) to generate a discrete echo mixing signal.

In some embodiments, processing the echo mixed signal to generate a raw data matrix comprises: establishing an original data matrix S from the echo mixing signal_n(k, l), wherein n is the nth receiving antenna of the continuous wave radar, l is the l chirp signal, and k is the k sampling point on the echo signal of the l chirp signal chirp. Referring to fig. 2, a raw data matrix is built from the chirp signals. In this embodiment, the echoes returned from each transmitted chirp signal after encountering a target are deposited on each column of the matrix, with the echoes of different chirp being deposited on different columns. There are K samples for each chirp echo signal, so the adaptive filter performs noise suppression on the K samples, which applies to l chirp. Since the K samples contain information about the radial distance, the first step is to optimize the signal-to-noise ratio in the range domain. The adaptive filtering used in the distance dimension is referred to herein as a distance filter.

In some embodiments, said filtering said raw data matrix through a two-dimensional adaptive filter comprises:

filtering the signal of the radial distance of the distance dimension by the radial distance filter;

and filtering the Doppler signals of the Doppler dimension through the Doppler filter.

In this embodiment, a two-dimensional adaptive filter is applied to each channel of the radar, and the chirp signals collected by the radar are arranged into a matrix as shown in fig. 2. The echo of each transmitted chirp signal after encountering the target is stored in each row of the matrix after being subjected to frequency mixing sampling; the echoes of the transmitted different chirp signals are mixed and sampled and then arranged on different columns of the matrix. The fourier transform of the signal in each column yields information about the radial distance, and hence when the adaptive noise filter mentioned in this application is used on the signal in this column, it is called a radial distance filter. The relative motion information of the moving object is obtained by the correlation of the respective received signals of the continuous chirp signals. Therefore, the adaptive filter proposed in this patent is called doppler filter when applied between chirp signals. Referring to fig. 4, fig. 4 is a block diagram of a two-dimensional adaptive filter of the present application; the two-dimensional adaptive filter is used for the channel 1, the channel 2, the … and the channel N; each channel adopts the same two-dimensional adaptive filter, wherein the two-dimensional adaptive filter comprises a Range filter and a Doppler filter, and the output ends of the Range filter and the Doppler filter are respectively subjected to fast Fourier transform to obtain filtered distance dimension information and Doppler dimension information. After radial distance filtering, a fast fourier transform FFT is performed along the distance dimension, i.e. the row in fig. 2. The result of the FFT is a complex matrix signal. Therefore, to apply the second filter to the velocity dimension, i.e., the doppler dimension, the filter must be complex. Unlike the distance filter, the doppler filter is applied to the doppler dimension, i.e., the column in fig. 2, and its operation requires complex operations. Also, the doppler filter will further suppress the noise in the doppler domain. It is with these two filters that this patent constructs a two-dimensional adaptive filter, thereby effectively suppressing noise in two dimensions.

In an implementation of the foregoing embodiment, referring to fig. 5, the filtering, by the radial distance filter, the signal of the radial distance in the distance dimension includes:

initializing a first input parameter a (k) from the radial filter, a (k) being a filter coefficient of length M +1, i.e. a (k) [ [ a ], (k) ]₀(k)，a₁(k)，…，a_M(k)]；

The first oneThe kth sampling point of the original data matrix of the echo mixing signal of the linear frequency modulation signal and the first M +1 sampling points form an intermediate input sampleThey and the first input parameter a (k) are input into the radial distance filter, and the estimated value y (k) of the true signal at the k-th sampling time is obtained, and the difference between y (k) and x (k, l) generates the error signal e_n；

According to the error signal e_nAnd intermediate input samplesUpdating the first input parameter a (k +1) at the k +1 th moment by an adaptive algorithm;

and performing iterative calculation on the echo mixing signals of the linear frequency modulation signals through a first input parameter a (k +1), and outputting a linear frequency modulation signal matrix of primary filtering.

Fig. 5 is a schematic structural diagram of a radial distance filter, where a chirp matrix x (k, l) is s (k, l) + n (k, l), where s (k, l) is a real signal, n (k, l) is noise, and y is_n(k) Is an estimate of the true signal s (k, l). The filter module for the nth channel (shown in dashed outline in fig. 5) outputs the parameter e_nAndas input to the adaptive algorithm. Parameter e_nIs an error signal that approximates the noise in the channel. Parameter(s)Is the intermediate input sample. These two parameters are necessary to calculate the filter coefficients to be used in the next iteration. The first input parameters a (k) are updated in the adaptation algorithm, i.e. the adaptation process. The filter coefficients a are initialized at the beginning of each chirp, even if a₀(k) Has a value of 1. A first input of the radial distance filter as the echo from the chirp is filteredThe parameters a (k) are updated. The shift register comprisesValues, which are intermediate input samples used by the algorithm in the adaptation. And (3) carrying out iterative calculation on the linear frequency modulation signal matrix through the first input parameters a (k) and outputting the linear frequency modulation signal matrix subjected to primary filtering. After the filtering the signal of the radial distance in the distance dimension by the radial distance filter, the method further includes: and obtaining radial distance information by performing fast Fourier transform on the filtered linear frequency modulation signal matrix. In this embodiment, the transmitted signal is a chirp signal, the echo of which when it encounters the target is a delayed chirp signal, the mixing signal with the transmitted signal is a sine wave, and the fourier transform yields the frequency of this sine wave, which has a direct correspondence to the distance to the target. Thus, radial distance information of the target is obtained through fourier transform.

In an implementation of the foregoing embodiment, referring to fig. 6, the filtering adjacent chirp signals of a chirp signal matrix by the doppler filter includes:

initializing a second input parameter b (l) of the doppler filter, b (l) being a filter coefficient b (l) of length P +1 [ b ]₀(l)，b₁(l)，…，b_P(l)]；

The kth sampling point of the original data matrix of the echo mixing signal of the ith linear frequency modulation signal and the echo mixing signals of the first P +1 linear frequency modulation signals form an intermediate input sampleThey and a second input parameter b (l) are input into the Doppler filter, and an error signal e is generated by the difference between the estimated value y (l), y (l) and X (k, l) of the echo mixing signal real signal of the first chirp signal at the k-th sampling moment_n(l)；

According to the error signal e_n(l) And intermediate input samplesUpdating the second input parameters b (l) by an adaptive algorithm;

and (3) performing iterative calculation on the original data matrix by inputting a second parameter b (l), and outputting a linear frequency modulation signal matrix which is filtered again.

In this embodiment, fig. 6 is a schematic structural diagram of a doppler filter, where a matrix of chirps X (k, l) is S (k, l) + n (k, l), where S (k, l) is a fourier transform of a real signal passing through a radial distance, n (k, l) is noise, y is noise, and y is a linear function of the real signal_n(l) Is an estimate of the true signal S (k, l). The filter module for the nth channel (shown in dashed outline in fig. 6) outputs the parameter e_nAndas input to the adaptive algorithm. Parameter e_nIs an error signal that approximates the noise in the channel. Parameter(s)Is the intermediate input sample. These two parameters are necessary to calculate the filter coefficients to be used in the next iteration. The second input parameters b (l) are updated in the adaptation algorithm, i.e. the adaptation process. The filter coefficients b are initialized at the beginning of each chirp, i.e. such that b is₀(l) Has a value of 1. As the echo from the chirp is filtered, the second input parameter b (l) of the doppler filter is updated. The shift register comprisesValues, which are intermediate input samples used by the algorithm in the adaptation. And (4) performing iterative calculation on the linear frequency modulation signal matrix through a second input parameter b (l), and outputting the linear frequency modulation signal matrix subjected to secondary filtering. After filtering the doppler signal in the doppler dimension by the doppler filter, the method further includes: calculating the linear frequency modulation signal matrix through fast Fourier transform to obtain the DopplerAnd (4) information.

In some embodiments, the adaptive algorithm is a normalized least mean square algorithm or a time-varying least mean square algorithm or a least squares method. According to the error signal e_nAnd intermediate input samplesUpdating the parameter a (k +1) of the radial distance filter at the time k +1 by an adaptive algorithm, comprising:

by passingCalculated, where Δ is a step constant.

Figure 7 is a block diagram that discloses an example of a least mean square based adaptive algorithm. First, the signal y is an estimate of the true signal s. The goal is to minimize the error e in the least mean square sense:

e＝y-(s+n)

taking the square of e:

e²＝(y-(s+n))²

＝y²+s²+n²-2(sy+ny)+2sn

find e²Average value of E (E)²):

E(e²)＝E(y²)+E(s²)+E(n²)-2E(sy+ny)+2E(sn).

It is assumed here that s and n are uncorrelated. When s is a sinusoidal signal and n is wideband noise, there is no correlation between them. The above assumption is therefore true. E (E)²) Has a minimum value of

E(e²)_min＝E(y²)_min+E(s²)+E(n²)-2E(sy+ny)_max+0.

Since s and n are uncorrelated, e (sn) ═ 0. When the signal power is greater than the noise power and y is very close to the true signal s, then E (E)²) The minimum value is reached.

It can be seen that the filter is effectively a band pass filter which brings the output y close to the true signal s, so that in another sense the power of the noise is suppressed. If s is a single frequency sinusoidal signal, then this filter is a very narrow band pass filter, allowing only s to pass. And E (E)²) Min equals the noise power of the actual channel. This means that the error signal e is an optimal estimate of the noise and can thus be effectively suppressed.

For the normalized LMS, the coefficients are updated as follows:where Δ is a step constant.

Similarly, the second input parameter b (l) can be passedAnd (6) performing calculation.

In some implementations, the processing the filtered raw data matrix includes:

carrying out incoherent superposition processing on the linear frequency modulation signal matrix of each channel; in this embodiment, the range-doppler maps (range-doppler maps) obtained by fourier transform of the N channels are combined to obtain a combined range-doppler map. An effective and fast method is to perform non-coherent superposition processing, i.e. non-coherent sum, on the data of the N channels, so that the data of the N matrices are combined into one matrix.

Carrying out constant-virtual early warning detection on the result after the incoherent superposition processing; in the embodiment, in the radar signal detection, when the external interference strength changes continuously, the radar can automatically adjust the sensitivity thereof, so that the false alarm probability of the radar remains unchanged, and the characteristic is called as a constant false alarm rate characteristic. That is, the threshold value for detection is not fixed in advance, but is adjusted according to the external strength, so that it can detect the target by finding the adaptive threshold value.

And carrying out target detection on the result after the constant-deficiency early warning detection. In the embodiment, in the radar signal detection, when the external interference strength changes continuously, the radar can automatically adjust the sensitivity thereof, so that the false alarm probability of the radar remains unchanged, and the characteristic is called as a constant false alarm rate characteristic. That is, the threshold value for detection is not fixed in advance, but adjusted accordingly according to the external strength, so that it is used to detect the target by finding the adaptive threshold value.

Referring to fig. 3, the present application discloses a noise reduction method for improving frequency modulated continuous wave radar target detection, which obtains a chirp signal matrix containing signals and noise through frequency modulated continuous waves, and performs primary filtering on the chirp signal matrix through a radial distance filter, and then performs secondary filtering on the chirp signal matrix through a doppler filter. If a plurality of channels exist, incoherent superposition can be carried out on the filtered multi-channel linear frequency modulation signal matrix, and a target list is obtained through constant-virtual early warning detection and target detection. See fig. 8, a) RD plot with higher noise. b) The radial distance filter magnitude response. c) RD pattern after radial distance filter. d) The doppler filter magnitude response. e) Radial distance and RD plot after doppler filter. It is clear from fig. 8 that the SNR of the target is significantly increased, thus improving the detection capability of such weak targets. The noise around the effective target in the automobile radar signal is reduced through the two-dimensional self-adaptive filter; the contrast of the target amplitude to the surrounding noise in the surrounding doppler RD pattern is greatly improved. By increasing this contrast, the ratio of the target signal power to the ambient noise is increased, allowing the CFAR to detect the target more efficiently. The clusters are made sensitively observable by the radar, providing more information on the possible sizes of the targets to aid in target classification. At the same time, the present application is done in software without expensive hardware modifications, making it easier to implement on top of existing algorithms.

It should be understood that the above examples of the present application are only examples for clearly illustrating the present application, and are not intended to limit the embodiments of the present application. It will be apparent to those skilled in the art that other variations and modifications can be made based on the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the claims of the present application.