Doppler radar signal processing method, system and Doppler radar

Abstract

The invention discloses a Doppler radar signal processing method, system and Doppler radar, including: collecting original signals; calculating DC bias, subtracting DC bias from the input original signal to obtain a signal without DC bias; sampling the signal without DC bias, and performing N-point summation operations on the points of the positive semi-axis and the negative semi-axis respectively, calculating the absolute value of the difference between the cumulative results of the positive semi-axis and the negative semi-axis, judging whether the current DC bias is accurate according to the absolute value of the difference between the cumulative results of multiple consecutive positive semi-axis and negative semi-axis and the set threshold, and if it is inaccurate, updating the DC bias; taking the absolute value of the signal, and performing noise estimation, comparing the signal with a preset threshold in real time, and outputting a high-level signal when the signal is greater than the preset threshold. The DC bias can be well tracked, the influence of instantaneous interference on DC estimation can be avoided, the adjustment of the chip working environment changes can be taken into account, noise interference can be avoided, and the normal sensing function of the radar is not affected.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing, and in particular relates to a Doppler radar signal processing method, system and Doppler radar. Background Art

[0002] Radar sensors use continuous waves (CW) to detect micro-motion in a set area and frequency modulated continuous waves (FMCW) to detect the distance of objects. The operating frequency bands include but are not limited to 5.8GHz, 9~10GHz, 24GHz, 60GHz, 77GHz and UWB bands. Doppler radar is a type of continuous wave radar. It sends fixed-frequency continuous radio waves, receives radio waves reflected by moving objects, and extracts the frequency components introduced by the Doppler effect of moving objects in the reflected waves to determine whether there are moving objects within a certain range.

[0003] In Doppler radar systems with limited cost, power consumption, and computing power, traditional radar algorithm signal processing algorithms usually use Figure 1 The signal processing flow shown is:

[0004] (1) The original signal collected by the ADC usually takes a certain number of sampling points and calculates the average value of the signal as the DC bias value of the signal. This value will be updated according to certain rules to track the changes in the DC bias of the signal;

[0005] (2) Subtract the DC value from the input original signal to obtain a signal without DC bias. If necessary, you can add an LPF low-pass filter module to filter out some out-of-band interference and noise;

[0006] (3) Take the absolute value of the signal to facilitate subsequent comparison and threshold judgment;

[0007] (4) The signal is compared with the preset threshold in real time. When the signal is greater than the preset threshold, a high-level signal is immediately output, indicating that a moving object has been sensed.

[0008] In actual application scenarios, the two key issues of DC bias offset and noise variation will have a great impact on the performance and working stability of radar sensors.

[0009] The DC bias of the signal is mainly affected by the following factors:

[0010] 1. Due to insufficient isolation between the board level or antenna, or reflection through the casing, the signal from the transmitting end directly returns to the receiving antenna end from the transmitting end. Depending on the strength and phase, the DC bias will be too high or too low.

[0011] 2. Changes in the chip's operating state, such as temperature and power supply voltage fluctuations, will also affect internal components, causing fluctuations in the DC bias of the output intermediate frequency signal. This type of impact is constantly changing during the operation of the radar chip, which will affect the radar's detection accuracy or generate false trigger signals.

[0012] In addition, in radar sensors, noise is one of the main factors affecting the maximum sensing distance and stability of the radar. Noise mainly comes from several sources: power supply noise, device noise, thermal noise, etc. As the operating temperature of the device changes, the noise energy is also constantly changing. In some application scenarios with high requirements for detection distance, accurate estimation of noise energy is crucial to expanding the sensor's extreme performance; in other scenarios with more complex environments, accurate noise estimation can also help reduce false triggering and improve radar stability.

[0013] The DC bias calculation process in common radar signal processing is as follows:

[0014] Usually, the DC bias link will continuously accumulate data from N sampling points, divide it by N after the accumulation is completed, and get the average value of the N sampling points. This value is used as the DC bias estimate for the subsequent period of time and participates in the subsequent calculation process of subtracting DC. The accumulation and averaging process is carried out continuously. Here we record is the value of the intermediate frequency signal sampled by the kth ADC, The calculation result of the current N ADC sampling points is expressed as follows:

[0015] Calculate the average value of the current N sampling points:

[0016]

[0017] Update the newly calculated N sample points to the DC value used in the DC subtraction step:

[0018]

[0019] This process is repeated continuously during radar detection. The DC elimination link continuously updates the DC size to be eliminated based on the new ADC average value, thereby realizing DC tracking under various working conditions of the chip.

[0020] In actual application scenarios, the disadvantages of the simple DC update algorithm above gradually become apparent. When an object moves irregularly within the radar sensing range, the intermediate frequency signal of the radar reflection wave analysis is not symmetrical; or there is instantaneous interference from the air or power supply, which causes glitches in the intermediate frequency signal. At this time, the calculated average value will deviate from the actual DC bias of the signal, causing errors in the subsequent judgment process; when there is no moving object, the uncertainty caused by the presence of noise will cause the DC bias to fluctuate slightly within a certain range, which will cause the radar sensing limit distance to be unstable. Summary of the invention

[0021] In order to solve the above-mentioned technical problems, the present invention provides a Doppler radar signal processing method, system and Doppler radar, which can track direct current bias (DC) well, while avoiding the influence of some instantaneous interference on DC estimation, taking into account the adjustment of changes in the chip working environment, avoiding noise interference, and not affecting the normal sensing function of the radar.

[0022] The technical solution of the present invention is:

[0023] A Doppler radar signal processing method comprises the following steps:

[0024] S01: collect original signal;

[0025] S02: Calculate the DC bias, subtract the DC bias from the input original signal to obtain a signal without DC bias; sample the signal without DC bias, and perform N-point summation operations on the points of the positive semi-axis and the negative semi-axis respectively, calculate the absolute value of the difference between the accumulated results of the positive semi-axis and the negative semi-axis, and judge the current DC bias according to the absolute value of the difference between the accumulated results of multiple consecutive positive semi-axis and negative semi-axis and the set threshold Is it accurate? If not, update it. ;

[0026] S03: Take the absolute value of the signal and perform noise estimation, compare the signal with a preset threshold in real time, and output a high-level signal when the signal is greater than the preset threshold.

[0027] In the preferred technical solution, in step S02, the current DC bias is determined Methods for accuracy include:

[0028] Perform N-point summation operations on the points of the positive and negative semi-axis respectively:

[0029]

[0030]

[0031] in, is the kth intermediate frequency signal after removing the DC bias, and the cumulative result of the positive half axis signal is , the absolute value of the negative half axis signal is accumulated and the result is ;

[0032] Calculate the absolute value of the difference between the accumulated results of the positive and negative semi-axis:

[0033]

[0034] Store the first M-1 consecutive Value, denoted as , ,……, ;

[0035] Calculate the time between two adjacent The absolute value of the difference is:

[0036] )

[0037] when When, update ;

[0038] in, yes Minimum threshold, yes The maximum threshold.

[0039] In the preferred technical solution, the method for estimating noise in step S03 includes:

[0040] The noise of the intermediate frequency signal without DC bias is Gaussian white noise. Since the DC component is subtracted, the expected value of the noise is equivalent to 0, that is:

[0041] E(x) = μ = 0

[0042] is the sample standard deviation, is the sample mean;

[0043] The density function of the normal distribution of Gaussian white noise is:

[0044]

[0045] Average the absolute value of the signal:

[0046]

[0047] Substitution After a simple calculation, we get:

[0048]

[0049]

[0050]

[0051] After conversion:

[0052]

[0053] Calculate the probability that the sample falls into the interval [-R, +R]:

[0054]

[0055] Select the noise floor amplitude , the noise at the predicted sampling point exceeds Number of ;

[0056] Where N is the number of sampling points in the current period, and P is the probability that the noise sampling point falls into the interval.

[0057] In the preferred technical solution, the noise estimation in step S03 also includes calculating the difference between the newly calculated noise value in each cycle and the currently used noise value. ;according to The size of the noise is updated by selecting different noise update coefficients.

[0058] In a preferred technical solution, the method for updating noise includes:

[0059] A lookup table is preset in the algorithm, according to Falling into different ranges ( , ], find the corresponding noise update coefficient :

[0060]

[0061] Among them, the noise update coefficient value The value range of is (0,1), The range interval and the corresponding noise update coefficient have the following relationship: The smaller the value, the The larger the value, that is:

[0062]

[0063] according to choose Then, update the noise amplitude value used by the algorithm according to the following calculation method :

[0064] .

[0065] In the preferred technical solution, the method of comparing the signal with the preset threshold in real time in step S03 includes:

[0066] S31: Set the preset threshold and noise amplitude Add them together to get the trigger threshold determined by the algorithm:

[0067] =

[0068] S32: After taking the absolute value, the signal When the timer is on, open the observation window and start the timer;

[0069] S33: Open the device in the observation window. , trigger the counter Add 1;

[0070] S34: Determine the observation window Is it established? The trigger count threshold is:

[0071] When the observation window has not ended and the counter is triggered , determine that there is a moving object trigger time, pull up the trigger mark, and close the observation window;

[0072] When the observation window timer reaches the observation window length , triggering the counter at the same time , it is considered that there is no valid trigger signal, the trigger mark is not pulled high, and the observation window is closed.

[0073] The preferred technical solution also includes, when the longest distance is required, Set to 0 to use the noise amplitude value As the trigger threshold, according to the noise level range R selected in the noise estimation algorithm and the probability P of the noise falling into the corresponding range, calculate The number of sampling points that may exceed the noise amplitude value within the observation window:

[0074] There are M sampling points in the observation window, and the number of sampling points exceeding the noise amplitude value is M*(1-P);

[0075] The trigger count threshold Set to be greater than M*(1-P).

[0076] The present invention also discloses a Doppler radar signal processing system, comprising:

[0077] Signal acquisition module, collecting original signals;

[0078] The DC bias estimation and removal module calculates the DC bias and subtracts the DC bias from the input original signal to obtain a signal without DC bias; samples the signal without DC bias and performs N-point summation operations on the points of the positive semi-axis and the negative semi-axis respectively, calculates the absolute value of the difference between the accumulated results of the positive semi-axis and the negative semi-axis, and determines the current DC bias based on the absolute value of the difference between the accumulated results of multiple consecutive positive semi-axis and negative semi-axis and the set threshold Is it accurate? If not, update it. ;

[0079] The noise estimation and discrimination module takes the absolute value of the signal and performs noise estimation, compares the signal with a preset threshold in real time, and outputs a high-level signal when the signal is greater than the preset threshold.

[0080] In the preferred technical solution, the DC bias estimation and removal module determines the current DC bias Methods for accuracy include:

[0081] Perform N-point summation operations on the points of the positive and negative semi-axis respectively:

[0082]

[0083]

[0084] in, is the kth intermediate frequency signal after removing the DC bias, and the cumulative result of the positive half axis signal is , the absolute value of the negative half axis signal is accumulated and the result is ;

[0085] Calculate the absolute value of the difference between the accumulated results of the positive and negative semi-axis:

[0086]

[0087] Store the first M-1 consecutive Value, denoted as , ,……, ;

[0088] Calculate the time between two adjacent The absolute value of the difference is:

[0089] )

[0090] when When, update ;

[0091] in, yes Minimum threshold, yes The maximum threshold.

[0092] The present invention further discloses a Doppler radar, comprising a signal processing chip, wherein any of the above-mentioned Doppler radar signal processing methods is embedded in the signal processing chip.

[0093] Compared with the prior art, the advantages of the present invention are:

[0094] By adding a DC bias (DC) evaluation, Update conditions to restrict and set appropriate Values ​​can be limited The update frequency can reduce the randomness of noise. Frequent changes provide a more stable data basis for the subsequent judgment process. It can track the DC bias (DC) well, while avoiding the impact of some instantaneous interference on DC estimation, taking into account the adjustment of changes in the chip working environment, avoiding noise interference, and not affecting the normal sensing function of the radar.

[0095] A gradual partial update algorithm is proposed, which can effectively filter out some random noise or interference, as well as the drastic changes in waveform caused by moving objects entering the sensing range, so that the noise used by the algorithm to judge remains at a relatively stable value. According to the actual situation of the application scenario, the appropriate update coefficient can be selected to enable the radar to adapt to the changed working conditions in a relatively short time, thus improving the user experience.

[0096] The processing flow of the present invention only uses simple operations such as addition, subtraction, and shift, and rarely uses multiplication and division, which greatly optimizes the radar algorithm operation flow. Even an 8-bit MCU can process radar signals of a certain bandwidth in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0098] Figure 1 It is the signal processing flow chart of the traditional radar algorithm;

[0099] Figure 2 is a flow chart of a Doppler radar signal processing method of the present invention;

[0100] Figure 3 Flowchart for DC offset estimation and elimination;

[0101] Figure 4 It is the normal distribution cumulative probability and normal distribution function curve;

[0102] Figure 5 Update the judgment algorithm flow chart for the gradual partial noise;

[0103] Figure 6 This is the flow chart of the Doppler radar sensing trigger judgment algorithm. DETAILED DESCRIPTION

[0104] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.

[0105] Example:

[0106] The preferred embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0107] like Figure 2 As shown, a Doppler radar signal processing method comprises the following steps:

[0108] S01: collect original signal;

[0109] S02: Calculate the DC bias, subtract the DC bias from the input original signal to obtain a signal without DC bias; sample the signal without DC bias, and perform N-point summation operations on the points of the positive semi-axis and the negative semi-axis respectively, calculate the absolute value of the difference between the accumulated results of the positive semi-axis and the negative semi-axis, and judge the current DC bias according to the absolute value of the difference between the accumulated results of multiple consecutive positive semi-axis and negative semi-axis and the set threshold Is it accurate? If not, update it. ;

[0110] S03: Take the absolute value of the signal and perform noise estimation, compare the signal with a preset threshold in real time, and output a high-level signal when the signal is greater than the preset threshold.

[0111] In a preferred embodiment, in step S02, the current DC bias is determined. Methods for accuracy include:

[0112] Perform N-point summation operations on the points of the positive and negative semi-axis respectively:

[0113]

[0114]

[0115] in, is the kth intermediate frequency signal after removing the DC bias, and the cumulative result of the positive half axis signal is , the absolute value of the negative half axis signal is accumulated and the result is ;

[0116] Calculate the absolute value of the difference between the accumulated results of the positive and negative semi-axis:

[0117]

[0118] Store the first M-1 consecutive Value, denoted as , ,……, ;

[0119] Calculate the time between two adjacent The absolute value of the difference is:

[0120] )

[0121] when When, update ;

[0122] in yes Minimum threshold, yes The maximum threshold.

[0123] In a preferred embodiment, the noise estimation method in step S03 includes:

[0124] The noise of the intermediate frequency signal without DC bias is Gaussian white noise. Since the DC component is subtracted, the expected value of the noise is equivalent to 0, that is:

[0125] E(x) = μ = 0

[0126] is the sample standard deviation, is the sample mean;

[0127] The density function of the normal distribution of Gaussian white noise is:

[0128]

[0129] Average the absolute value of the signal:

[0130]

[0131] Substitution After a simple calculation, we get:

[0132]

[0133]

[0134]

[0135] After conversion:

[0136]

[0137] Calculate the probability that a sample falls into a given interval:

[0138]

[0139] Select the noise floor amplitude , the noise at the predicted sampling point exceeds Number of ;

[0140] Where N is the number of sampling points in the current period, and P is the probability that the noise sampling point falls into the value range.

[0141] In a preferred embodiment, the noise estimation in step S03 further includes calculating the difference between the noise value newly calculated in each cycle and the noise value currently used. ;according to The size of the noise is updated by selecting different noise update coefficients.

[0142] In a preferred embodiment, the method for updating noise includes:

[0143] A lookup table is preset in the algorithm, according to Falling into different ranges ( , ], find the corresponding noise update coefficient :

[0144]

[0145] Among them, the noise update coefficient value The value range of is (0,1), The range interval and the corresponding noise update coefficient have the following relationship: The smaller the value, the The larger the value, that is:

[0146]

[0147] according to choose Then, update the noise amplitude value used by the algorithm according to the following calculation method :

[0148] .

[0149] In a preferred embodiment, the method of comparing the signal with a preset threshold in real time in step S03 includes:

[0150] S31: Set the preset threshold and noise amplitude Add them together to get the trigger threshold determined by the algorithm:

[0151] =

[0152] S32: After taking the absolute value, the signal When the timer is on, open the observation window and start the timer;

[0153] S33: Open the device in the observation window. , trigger the counter Add 1;

[0154] S34: Determine the observation window Is it established? The trigger count threshold is:

[0155] When the observation window has not ended and the counter is triggered , determine that there is a moving object trigger time, pull up the trigger mark, and close the observation window;

[0156] When the observation window timer reaches the observation window length , triggering the counter at the same time , it is considered that there is no valid trigger signal, the trigger mark is not pulled high, and the observation window is closed.

[0157] A preferred embodiment also includes, when the longest distance is required, Set to 0 to use the noise amplitude value As the trigger threshold, according to the noise level range R selected in the noise estimation algorithm and the probability P of the noise falling into the corresponding range, calculate The number of sampling points that may exceed the noise amplitude value within the observation window:

[0158] There are M sampling points in the observation window, and the number of sampling points exceeding the noise amplitude value is M*(1-P);

[0159] The trigger count threshold Set to be greater than M*(1-P).

[0160] Another embodiment, a Doppler radar signal processing system, comprising:

[0161] Signal acquisition module, collecting original signals;

[0162] The DC bias estimation and removal module calculates the DC bias and subtracts the DC bias from the input original signal to obtain a signal without DC bias; samples the signal without DC bias and performs N-point summation operations on the points of the positive semi-axis and the negative semi-axis respectively, calculates the absolute value of the difference between the accumulated results of the positive semi-axis and the negative semi-axis, and determines the current DC bias based on the absolute value of the difference between the accumulated results of multiple consecutive positive semi-axis and negative semi-axis and the set threshold Is it accurate? If not, update it. ;

[0163] The noise estimation and discrimination module takes the absolute value of the signal and performs noise estimation, compares the signal with a preset threshold in real time, and outputs a high-level signal when the signal is greater than the preset threshold.

[0164] The following is a detailed description of the working process of the Doppler radar signal processing system using a best example:

[0165] The present invention optimizes the operation processing of the "calculation of DC bias (DC)" process, and adds a "noise estimation" algorithm in the absolute value processing stage, and passes the noise estimation result to the "discrimination and output" link. With the newly designed motion / existence discrimination algorithm, a better balance can be achieved between misjudgment and sensitivity. The complete algorithm process implemented by the present invention is as follows: Figure 2 shown.

[0166] The above processing flow only uses simple operations such as addition, subtraction, and shift, and rarely uses multiplication and division, which greatly optimizes the radar algorithm operation process. Even an 8-bit MCU can process radar signals of a certain bandwidth in real time.

[0167] 1. DC and noise estimation algorithm for Doppler radar signals

[0168] In actual application scenarios, two key issues, DC bias offset and noise variation, will have a great impact on the performance and working stability of radar sensors:

[0169] 1.1 DC bias estimation and elimination:

[0170] The DC bias of the signal is mainly affected by the following factors:

[0171] The signal from the transmitting end directly returns to the receiving antenna end due to insufficient isolation between the board level or the antenna, or reflection through the casing. Depending on the strength and phase, the DC bias will be too high or too low.

[0172] Changes in the chip's operating state, such as temperature and power supply voltage fluctuations, can also affect internal components, causing fluctuations in the DC bias of the output intermediate frequency signal. This type of influence changes continuously during the radar chip's operation, affecting radar detection accuracy or generating false trigger signals.

[0173] Here we propose a new DC estimation and elimination algorithm. The algorithm structure diagram is shown in Figure 3 As shown:

[0174] The new algorithm samples the output of the DC elimination module and performs N-point summation operations on the points of the positive and negative semi-axis respectively:

[0175]

[0176]

[0177] in is the intermediate frequency signal after removing DC, The cumulative result of the positive half axis signal is , the absolute value of the negative half axis signal is accumulated and the result is .

[0178] The DC evaluation module calculates the absolute value of the difference between the accumulated results of the positive and negative half axes:

[0179]

[0180] At the same time, the module stores the first M-1 consecutive Value, denoted as , ,……, .

[0181] The DC evaluation module determines the current Is it accurate? If not, update it once ,Right now .

[0182] Remember the interval between two adjacent The absolute value of the difference is:

[0183] )

[0184] The judgment conditions are as follows:

[0185]

[0186] in is a preset fixed value in the algorithm, as Ideally, if the DC bias If the calculation is completely accurate, then we have By limiting The meaning is that only when The estimated And the real V DC The error between them is large enough to allow update. Values ​​can be limited The update frequency can reduce the randomness of noise. Frequent changes provide a more stable data basis for subsequent judgment processes.

[0187] in It is a preset fixed value in the algorithm, as two consecutive The difference, that is Sometimes some instantaneous interference signals or small disturbances from moving objects may cause a short-term offset in the DC bias of the intermediate frequency signal, but we do not want the algorithm to track this short-term offset and update . Continuously judge several The change between values ​​can effectively eliminate these instantaneous interferences. The algorithm needs to see M cycles continuously. The waveform is in a very small range. At this time, it can be judged that the current state is stable and can be updated. At the same time, according to the motion characteristics of the target sensing object, setting a suitable M value can also prevent the large signal intermediate frequency waveform generated by the moving object entering the radar sensing range from affecting the Have a greater impact.

[0188] In the radar signal processing process, the DC bias link will continuously accumulate N sampling point data, divide it by N after the accumulation is completed, and get the average value of N sampling points. This value is used as the DC bias estimate for the subsequent period of time and participates in the subsequent calculation process of subtracting DC. The accumulation and averaging process is carried out continuously. Here is recorded is the value of the intermediate frequency signal sampled by the kth ADC, is the average value of the current N ADC sampling points, expressed as follows:

[0189] Calculate the average value of the current N sampling points:

[0190]

[0191] Update the newly calculated N sample points to the DC value used in the DC subtraction step:

[0192]

[0193] By adding the DC evaluation module and By limiting the update conditions, the DC bias (DC) can be tracked well, while avoiding the impact of some instantaneous interference on DC estimation, taking into account the adjustment of changes in the chip's working environment, avoiding noise interference, and not affecting the normal sensing function of the radar.

[0194] The present invention proposes a noise estimation and updating algorithm which can meet the requirements of different application scenarios.

[0195] 1.2 Noise Estimation Algorithm

[0196] The noise of the intermediate frequency signal output by the RF front-end chip is mainly Gaussian white noise. The amplitude distribution of Gaussian white noise obeys Gaussian distribution (normal distribution). The density function of normal distribution is as follows:

[0197]

[0198] in is the sample standard deviation, is the sample mean.

[0199] For the above radar signal processing flow, since the DC component is subtracted, the expected value equivalent to the noise is 0, that is:

[0200] E(x) = μ = 0

[0201] The density function of the normal distribution becomes:

[0202]

[0203] Since the absolute value of the signal after DC elimination is taken in the radar signal processing process, the average value (expected) of the signal after taking the absolute value is calculated:

[0204]

[0205] Substitution After simple calculation, we can get:

[0206]

[0207]

[0208]

[0209] After conversion:

[0210]

[0211] According to the above formula, the standard deviation of the noise signal can be easily calculated by finding the average value of the absolute value of the noise (E(|x|)). After obtaining the standard deviation of the noise signal, we can clearly understand the distribution of the signal and easily determine the dividing point between the noise and signal amplitudes. According to the cumulative distribution function of the normal distribution:

[0212]

[0213] The probability that a sample falls into the interval [-R, +R] can be calculated:

[0214]

[0215] The relationship between the normal distribution cumulative probability and the normal distribution function curve is as follows Figure 4 As shown, the probability of the noise sampling point falling into the value range is P. When the value range R= When , approximately P = 68.3% of the sample values ​​will fall within the range; when the value range R = When , approximately P = 95.4% of the sample values ​​will fall within the range; when the value range R = 3 When , approximately P = 99.7% of the sample values ​​will fall within the range. According to this rule, a suitable value can be selected as the bottom noise amplitude (hereinafter referred to as ). Select After that, how many sampling points will have noise exceeding Can be Prediction (N is the number of sampling points in the current cycle) makes it easier for subsequent algorithms to filter noise and determine object motion.

[0216] 1.3 Noise Update Algorithm

[0217] In the above radar algorithm system, it is equally important to update the noise at the right time. Usually we hope to calculate the noise level when there are no moving objects in the radar sensing range, and eliminate external interference as much as possible to obtain an accurate noise estimate. However, such conditions are difficult to meet in actual scenarios. If the radar is allowed to continuously calculate the noise and update it to the threshold judgment system, the Doppler signal of the moving object will be mistakenly counted as noise, resulting in a significant increase in the estimated noise value, which in turn affects the radar sensitivity. Another idea is that the radar algorithm updates the noise when it determines that no moving object is sensed, otherwise the previous noise estimation level is maintained. Tests in actual scenarios have found that sometimes due to the sudden increase of various interference noises, the radar will be triggered falsely; after the radar is triggered, the new noise value cannot be updated, and the increased noise cannot be filtered, resulting in repeated false triggering of the radar, and the radar sensor system enters a deadlock state and cannot exit. In response to the above-mentioned problems, the present invention proposes a new noise update judgment algorithm. The algorithm processing flow chart is shown in the figure. Figure 5 shown.

[0218] Compared with binary judgment of whether the noise is updated “yes” or “no”, a gradual partial update algorithm can be considered. The newly calculated noise value in each cycle is The noise value currently used by the algorithm is , the difference between the two can be calculated:

[0219]

[0220] A lookup table can be preset in the algorithm, according to Falling into different ranges, find a corresponding noise update coefficient :

[0221]

[0222] in The value range is (0,1), The range interval and the corresponding noise update coefficient have the following relationship: The smaller the value, the The larger the value, that is:

[0223]

[0224] according to choose After that, the noise amplitude value used by the algorithm can be updated according to the following calculation rules :

[0225]

[0226] Indicates the noise amplitude value obtained in the previous calculation cycle (i-1), which participates in the calculation of the trigger threshold of the previous calculation cycle; It is the noise value calculated based on the new N sampling points and does not participate in the trigger threshold calculation; It is a fusion and The obtained noise threshold will be used in the trigger threshold calculation.

[0227] This update rule can be intuitively understood as follows: is the difference between the newly calculated noise level of the current period and the noise level of the previous period. The larger the value, the greater the difference between the two, indicating that the noise mutation is more frequent, and the greater the noise calculation result is, the more likely it is due to external interference, or the Doppler waveform change caused by a moving object entering the sensing range. At this time, a smaller update coefficient should be selected, and only a small part of the newly calculated noise result should be used as the update value, and the rest should still use the noise result calculated last week. This can well filter out some random noise or interference, as well as the drastic waveform changes caused by moving objects entering the sensing range, so that the noise used in the algorithm judgment remains at a relatively stable value.

[0228] if The smaller it is, the greater the probability that the newly calculated noise value is close to the real noise value. The changed part is the difference caused by the random fluctuation of the noise. At this time, a larger update coefficient should be selected to take a larger part of the newly calculated noise coefficient as the calculation result of the new noise. In this way, the real noise changes can be better tracked without interference or moving objects.

[0229] Another extreme case is that when there is no interference or motion object induction, some reason causes the real noise to suddenly change dramatically. Since this judgment algorithm does not regard this drastic change as an induction signal and prohibits updating the noise, it uses a smaller coefficient to slowly update the noise. Gradually catching up with the changed noise, thus avoiding the self-locking problem of repeated false triggering of the radar sensor. Selecting a suitable update coefficient according to the actual situation of the application scenario can make the radar adapt to the changed working conditions in a relatively short time and improve the user experience.

[0230] 2. Doppler radar sensing trigger judgment algorithm

[0231] Based on the above radar signal processing flow, the present invention proposes a corresponding micro-motion sensing algorithm. The algorithm flow is as follows:

[0232] Several parameters are preset in the algorithm:

[0233] Threshold : The threshold can be adjusted according to the application scenario or different sensing requirements;

[0234] Trigger count threshold ;

[0235] Length of observation window after triggering ;

[0236] Trigger and judgment process Figure 6 As shown:

[0237] 1) Set the preset threshold and noise amplitude Add them together to get the trigger threshold determined by the algorithm:

[0238] =

[0239] 2) When you find When the timer is on, open the observation window and start the timer;

[0240] 3) Open the device in the observation window and whenever a , trigger the counter Add 1;

[0241] 4) Determine the observation window There may be two situations:

[0242] The observation window has not ended and the counter is triggered , it is considered that there is a moving object trigger time, the trigger flag is pulled high, and the observation window is closed;

[0243] The observation window timer counts up to the observation window length , triggering the counter at the same time , it is considered that there is no valid trigger signal, the trigger mark is not pulled high, and the observation window is closed.

[0244] When the maximum distance is required, you can Set to 0 to actually use the noise amplitude value At this time, according to the noise level range R selected in the aforementioned noise amplitude value estimation algorithm and the probability P of the noise falling into the corresponding range, the The number of sampling points that may exceed the noise amplitude value in the observation window: Assume There are M sampling points in the observation window, so the number of sampling points exceeding the noise amplitude value is approximately M*(1-P). By setting it to be greater than M*(1-P), false triggering caused by noise can be effectively reduced, and a lower trigger threshold can be used to obtain a longer sensing distance.

[0245] It should be understood that the above specific embodiments of the present invention are only used to illustrate or explain the principles of the present invention, and do not constitute a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.

Claims

1. A Doppler radar signal processing method, characterized in that: The following steps are involved: S01: collect original signal; S02: Calculate the DC bias, subtract the DC bias from the input original signal, and obtain a signal without DC bias; Sample the signal without DC bias, perform N-point summation operation on the points of the positive and negative axes respectively, calculate the absolute value of the difference between the accumulated results of the positive and negative axes, and judge the current DC bias based on the absolute value of the difference between the accumulated results of multiple consecutive positive and negative axes and the set threshold. Is it accurate? If not, update it. ; The determination of the current DC bias Methods for accuracy include: Perform N-point summation operations on the points of the positive and negative semi-axis respectively: ； ； in, is the kth intermediate frequency signal after removing the DC bias, and the cumulative result of the positive half axis signal is , the absolute value of the negative half axis signal is accumulated and the result is ; Calculate the absolute value of the difference between the accumulated results of the positive and negative semi-axis: ； Store the first M-1 consecutive Value, denoted as , ,……, ; Calculate the time between two adjacent The absolute value of the difference is: )； when When, update ; in, yes Minimum threshold, yes The maximum threshold of S03: Take the absolute value of the signal and perform noise estimation, compare the signal with a preset threshold in real time, and output a high-level signal when the signal is greater than the preset threshold.

2. The Doppler radar signal processing method according to claim 1, characterized in that: The noise estimation method in step S03 includes: The noise of the intermediate frequency signal without DC bias is Gaussian white noise. Since the DC component is subtracted, the expected value of the noise is equivalent to 0, that is: E(x) = μ = 0 is the sample standard deviation, is the sample mean; The density function of the normal distribution of Gaussian white noise is: ； Average the absolute value of the signal: ； Substitution After a simple calculation, we get: ； ； ； After conversion: ； Calculate the probability that the noise sample falls into the interval [-R, +R]: ； Selecting the Noise Amplitude , the noise at the predicted sampling point exceeds Number of ; Where N is the number of sampling points in the current period, and P is the probability that the noise sampling point falls into the interval.

3. The Doppler radar signal processing method according to claim 1, characterized in that: The noise estimation in step S03 also includes calculating the difference between the noise value newly calculated in each cycle and the noise value currently used. ;according to The size of the noise is updated by selecting different noise update coefficients.

4. The Doppler radar signal processing method according to claim 3, characterized in that: The method for updating noise includes: A lookup table is preset in the algorithm, according to fall into different intervals ( , ], find the corresponding noise update coefficient : ； Among them, the noise update coefficient value The value range of is (0,1), The interval and the corresponding noise update coefficient have the following relationship: The smaller the value, the larger the corresponding value range. The larger the value, that is: ； according to choose Then, update the noise amplitude value used by the algorithm according to the following calculation method : ； in, represents the noise amplitude value obtained in the previous calculation cycle i-1, is the noise value calculated based on the new N sampling points.

5. The Doppler radar signal processing method according to claim 4, characterized in that: The method of comparing the signal with a preset threshold in real time in step S03 includes: S31: Set the preset threshold and noise amplitude Add them together to get the trigger threshold determined by the algorithm: = ； S32: After taking the absolute value, the signal When the timer is on, open the observation window and start the timer; S33: Open the device in the observation window. , trigger the counter Add 1; S34: Determine the observation window Is it established? The trigger count threshold is: When the observation window has not ended and the counter is triggered , determine that there is a moving object trigger time, pull up the trigger mark, and close the observation window; When the observation window timer reaches the observation window length , triggering the counter at the same time , it is considered that there is no valid trigger signal, the trigger mark is not pulled high, and the observation window is closed.

6. The Doppler radar signal processing method according to claim 5, characterized in that: Also included, when the maximum distance is required, Set to 0 to use the noise amplitude value As the trigger threshold, according to the noise level interval R selected in the noise estimation algorithm and the probability P of the noise falling into the corresponding interval, calculate The number of sampling points that may exceed the noise amplitude value within the observation window: There are M sampling points in the observation window, and the number of sampling points exceeding the noise amplitude value is M*(1-P); The trigger count threshold Set to be greater than M*(1-P).

7. A Doppler radar signal processing system, characterized in that: include: Signal acquisition module, collecting original signals; A DC bias estimation and removal module calculates the DC bias and subtracts the DC bias from the input original signal to obtain a signal without DC bias; Sample the signal without DC bias, perform N-point summation operation on the points of the positive and negative axes respectively, calculate the absolute value of the difference between the accumulated results of the positive and negative axes, and judge the current DC bias based on the absolute value of the difference between the accumulated results of multiple consecutive positive and negative axes and the set threshold. Is it accurate? If not, update it. ; The determination of the current DC bias Methods for accuracy include: Perform N-point summation operations on the points of the positive and negative semi-axis respectively: ； ； in, is the kth intermediate frequency signal after removing the DC bias, and the cumulative result of the positive half axis signal is , the absolute value of the negative half axis signal is accumulated and the result is ; Calculate the absolute value of the difference between the accumulated results of the positive and negative semi-axis: ； Store the first M-1 consecutive Value, denoted as , ,……, ; Calculate the time between two adjacent The absolute value of the difference is: )； when When, update ; in, yes Minimum threshold, yes The maximum threshold of The noise estimation and discrimination module takes the absolute value of the signal and performs noise estimation, compares the signal with a preset threshold in real time, and outputs a high-level signal when the signal is greater than the preset threshold.

8. A Doppler radar, comprising a signal processing chip, characterized in that: The signal processing chip is embedded with the Doppler radar signal processing method according to any one of claims 1 to 6.

Images