CN110879389A - Multi-body target recognition and localization method based on multistatic IR-UWB bio-radar signal - Google Patents

Abstract

The invention belongs to the technical field of biological radars, and discloses a multi-human body target identification and positioning method based on multi-base IR-UWB biological radar signals. According to the invention, a set of target discrimination program combining two characteristic parameters of energy-to-noise ratio and correlation coefficient mean is adopted according to signal characteristics of human respiration and the like, so that the recognition and the distance direction positioning of a multi-human target are realized, and the target recognition accuracy is improved.

Description

Multi-body target recognition and localization method based on multistatic IR-UWB bio-radar signal

technical field

The invention belongs to the technical field of biological radar, and in particular relates to a multi-body target identification and positioning method based on a multi-base IR-UWB biological radar signal.

Background technique

Bioradar is a non-contact, long-distance, and penetrating medium to realize the functions of life detection, vital sign monitoring, life imaging and positioning by extracting the signals related to vital signs in radar echoes. The principle is that the radar emits electromagnetic waves to the human body, and the electromagnetic waves are reflected back to the radar receiving antenna after modulation of human physiological activities such as breathing, heartbeat, and body movement. Physiological and biological information including physiological parameters, waveforms, images, target locations, etc. Because of the above advantages, bio-radar technology has shown great advantages and broad application prospects in the fields of post-disaster rescue, medical monitoring, anti-terrorism and stability maintenance, and battlefield search and rescue.

Multi-target recognition and positioning in bio-radar detection is a difficult point, and it is also a bottleneck technology that restricts the further development of bio-radar technology, which affects the practical value of existing bio-radar prototypes. The prior art lacks the achievement of multi-human target recognition and localization based on bio-radar. Therefore, the solution to the multi-human target recognition and localization problem can greatly improve the detection efficiency in non-contact life detection and meet the needs of multi-target detection in actual detection. The problem of positioning can expand the application scope of bio-radar, thereby promoting the further development of the bio-radar industry.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a multi-body target identification and positioning method based on multi-base IR-UWB bio-radar signals, so as to solve the problem of lack of identification and positioning of multi-body targets realized by bio-radar in the prior art.

In order to realize the above-mentioned tasks, the present invention adopts the following technical solutions, including the following steps:

Step 1: The transmitting antenna of the three-channel IR-UWB bio-radar transmits radar pulses to the target position. The three channels include a middle channel, a left channel and a right channel. The radar pulses are reflected at the target position and pass through the three-channel IR-UWB. The three receiving antennas of the bioradar obtain the radar echo signal E(m,n) in the middle channel, the radar echo signal E _left (m,n) in the left channel and the radar echo signal E _right (m,n) in the right channel, where , m is the sampling sequence number in the fast time direction, n is the sampling sequence number in the slow time direction, and m and n are positive integers;

Step 2: Use E(m,n), E _left (m,n) and E _right (m,n) obtained in step 1 to obtain the middle channel energy signal E ₆ (l) and the left channel energy signal through signal preprocessing respectively E _6left (l) and right channel energy signal E _6right (l);

Step 3: Inflection point extraction is performed on E ₆ (l), E _6left (l) and E _6right (l) obtained in step 2, and the secondary inflection point signal E ₉ (l) of the middle channel and the secondary inflection point signal E of the left channel are obtained. _9left (l) and right channel secondary inflection point signal E _9right (l);

Step 4: Obtain the position of the top three values and the top three values of each secondary inflection point signal obtained in step 3, and the top three values include the maximum value, the second maximum value and the third maximum value;

Step 5.1: Calculate the peak-to-background ratio and the mean value of the correlation coefficient of the first three values of E ₉ (1), obtain the number of targets by the peak-to-background ratio and mean value of the correlation coefficient of the first three values of E ₉ (l), include:

A) If V _EtoB1 <σ _1N , or σ _{1N ≤V EtoB1} <σ _1Y and r _m1 <σ _rm1 , the judgment result is no target;

B) If σ _{1N ≤V EtoB2} <σ _2N , or σ _{2N ≤V EtoB2} <σ _2Y and r _m2 <σ _rm1 , the discrimination result is a single target;

C) If r _m3 >σ _rm3 , or σ _rm2 <r _m3 ≤σ _rm3 and V _EtoB3 ≥σ _1Y , the discrimination result is three-objective;

D) In addition to A) B) C) other cases, the discrimination result is a dual target;

Wherein, V _EtoB1 represents the peak-to-background ratio of the maximum value of E ₉ (l), V _EtoB2 represents the peak-to-background ratio of the second largest value of E ₉ (l), and V _EtoB3 represents the peak of the third largest value of E ₉ (l) - Background ratio, r _m1 means the mean value of the correlation coefficient of the largest value, r _m2 means the mean value of the correlation coefficient of the second largest value, r _m3 means the mean value of the correlation coefficient of the third largest value, σ _1N means no target threshold, σ _1Y means there is a target Threshold, σ _2N represents dual-target threshold, σ _2Y represents multi-target threshold and σ _1N <σ _2N <σ _1Y <σ _2Y , σ _rm1 , σ _rm2 and σ _rm3 represent correlation threshold and σ _rm2 <σ _rm1 <σ _rm3 ;

Step 5.2: According to the number of targets identified in step 5.1, determine the radial distance of the target and the orientation of the target in combination with the positions of the first three values of E ₉ (l), E _9left (l) and E _9right (l), including:

a) If the recognition result is a single target, the radial distance of the target is determined by the position of the maximum value of E ₉ (1), and the orientation of the target is determined by the position of the maximum value of E _9left (1) and E _9right (1);

b) If the recognition result is a double target, the radial distance of the first target is determined by the position of the maximum value of E ₉ (l), and the first target is determined by the position of the maximum value of E _9left (l) and E _9right (l). The bearing of one target, then the target radial distance of the second by the position of the second largest value of _{E9(l), determined by the position of the second largest value of E9left (} l) and _E9right (l) the orientation of the second target;

c) If the recognition result is three targets, the radial distance of the first target is determined by the position of the maximum value of E ₉ (l), and the first target is determined by the position of the maximum value of E _9left (l) and E _9right (l). The bearing of one target, then the target radial distance of the second by the position of the second largest value of _{E9(l), determined by the position of the second largest value of E9left (} l) and _E9right (l) The bearing of the second target, and finally the radial distance of the third target is determined by the position of the third largest value of _{E9(l), by the position of the third largest value of E9left (} l) and _E9right (l) determine the orientation of the third target;

Identify and locate the target according to the number of targets, the radial distance of the target and the orientation of the target.

Further, in step 5.1, σ _1N =2, σ _1Y =3, σ _2N =2.4, σ _2Y =3.5, σ _rm1 =0.8, σ _rm2 =0.75 and σ _rm3Y =0.88.

Further, the signal preprocessing in step 2 includes the following sub-steps:

Step 2.1: Accumulate distances for E(m,n), E _left (m,n) and E _right (m,n) respectively;

Step 2.2: Multiply the accumulated signal in step 2.1 by the exponential gain curve G(l) of formula I to perform attenuation compensation;

Among them, V _h represents the ratio of the maximum value of the radar echo data to the amplitude of the reflected echo of the human target, P _human is the position of the human target, l represents the fast time sequence number after distance accumulation, l=1,2,...,L and L is a positive integer;

Step 2.3: remove static clutter from the signal after attenuation compensation in step 2.2;

Step 2.4: remove the linear trend of the signal after removing the static clutter in step 2.3;

Step 2.5: perform low-pass filtering on the signal after the linear trend removal in step 2.4 in the slow time dimension;

Step 2.6: Accumulate the low-pass filtered signals in step 2.5 along the slow time axis to obtain the middle channel energy signal E ₆ (1), the left channel energy signal E _6left (1) and the right channel energy signal E _6right (1).

Further, the inflection point extraction in step 3 includes the following sub-steps:

Step 3.1: Remove the direct wave from E ₆ (l), E _6left (l) and E _6right (l) obtained in step 2 to obtain E ₇ (l), E _7left (l) and E _7right (l);

Step 3.2: Extract the inflection point from the signal obtained in step 3.1, and obtain primary inflection point signals E ₈ (l), E _8left (l) and E _8right (l);

Step 3.2: Extract the inflection point from the signal obtained in step 3.2, and obtain the secondary inflection point signal E ₉ (l) of the middle channel, the secondary inflection point signal E _9left (l) of the left channel and the secondary inflection point signal E _9right (l) of the right channel.

Further, step 4 includes the following sub-steps:

Step 4.1: Obtain the maximum value of each secondary inflection point signal obtained in step 3, and remove the tail of the maximum value;

Step 4.2: Obtain the second largest value of each quadratic inflection point signal obtained in step 3, and remove the tail of the second largest value;

Step 4.3: Obtain the third largest value of each quadratic inflection point signal obtained in step 3.

When the first three values are obtained, the adjacent 16 consecutive amplitude signals after the position of the maximum value and the position of the second maximum value are set to zero to remove the tail of the maximum value.

Further, the mean value r _m1 of the correlation coefficient at the maximum value, the mean value r _m2 of the correlation coefficient at the second largest value and the mean value r _m3 of the correlation coefficient at the third largest value are calculated by formula II:

Among them, i represents the serial number of the correlation coefficient and i=1,2,3,4,5,6, k represents the serial number of the first three values and k=1 represents the serial number of the maximum value, k=2 represents the second largest value Serial number, k=3 represents the serial number of the third largest value, r _ik represents the correlation coefficient of the first three values, r _mk represents the mean value of the correlation coefficient of the first three values, E ₅ (l, q) represents the slow time obtained in step 2.5 The signal of the intermediate channel after low-pass filtering in the dimension, E _maxk (q) represents the signal at the position of the first three values of E ₅ (l, q), and E _(maxk+(i-4)) (q) represents the same as The signal of the first three positions adjacent to the first three values of E ₅ (l,q), (E _(maxk+(i-3)) (q) represents the first three values of E ₅ (l, q) Signals at the next three positions adjacent to each other, Q represents the total number of signal sampling points in the slow time direction of E ₅ (l, q) and Q is a positive integer, q represents the qth signal sampling point in the slow time direction and q is positive integer.

Further, in step 5, the position of the target is determined by the position l _left-maxk of the first three values of E _9left (l) and the position l _right-maxk of the first three values of E _9right (l), including:

a) If |l _left-maxk -l _right-maxk |≤2, then the target is on the central axis, and the central axis is the connection between the receiving antenna and the transmitting antenna of the middle channel;

b) If (l _left-maxk -l _right-maxk )≤-2, the target is in the left area of the central axis, and the left area is the area where the left channel transmit antenna on one side of the central axis is located;

c) If (l _left-maxk -l _right-maxk )>2, the target is in the right area of the central axis, and the right area is the area where the right channel transmit antenna on one side of the central axis is located.

Compared with the prior art, the present invention has the following technical characteristics:

1. The present invention adopts a set of target discrimination procedures combining two characteristic parameters of energy-to-noise ratio and mean correlation coefficient according to the signal characteristics of human body breathing, etc., to realize the identification and distance orientation of multiple human targets, and to improve the accuracy of target recognition. .

2. The present invention uses pre-determined human target position, and then according to the calculated exponential gain curve G(l), the signal is attenuated and compensated in the distance direction, which solves the energy attenuation of the radar signal as the distance increases during the transmission process. The problem is that the missed judgment rate of the remote target is effectively reduced. Compared with the segmented linear gain compensation method, the gain compensation adopted in the present invention is more accurate.

3. The present invention adopts the method of three-channel radar detection, the middle channel identifies and locates up to three human targets in the distance direction, and the two channels locate the target in the azimuth direction, which realizes the detection, identification and positioning (positioning) of the three targets. The result contains orientation information), which improves the accuracy of multi-target 2D positioning.

4. The present invention finds the optimal target existence threshold and correlation threshold by studying the relevant factors affecting V _EtoBk and the relevant factors affecting r _mk , and the recognition effect is better improved by setting the optimal threshold.

Description of drawings

Figure 1 is a schematic block diagram of a three-channel IR-UWB bio-radar system;

Figure 2 is a schematic diagram of a three-channel IR-UWB bioradar system for through-wall detection;

Figure 3 is a schematic diagram of a two-dimensional matrix of radar echo signals;

Figure 4(a) is a waveform diagram of a fast time signal;

Figure 4(b) is a waveform diagram of a slow time signal;

Figure 5 is a schematic diagram of IR-UWB radar detecting human breathing;

Fig. 6 is the pulse echo schematic diagram of human respiration;

7 is a flowchart of a signal preprocessing algorithm;

8 is a schematic diagram of a simulated radar echo signal;

9 is a schematic diagram of an analog echo signal after static clutter is removed;

FIG. 10 is a schematic diagram of three maxima of the secondary inflection point signal E ₉ (1) and the phenomenon of “smearing”;

Figure 11 is a schematic diagram of the secondary inflection point signal E(l) after the middle channel has removed "smearing";

Figure 12 is a three-target identification and positioning result diagram.

13 is a flowchart of a method for judging the number of multiple human targets;

Detailed ways

First, the technical terms that appear in the present invention are explained:

Bioradar technology: including continuous wave (CW) radar and ultra-wideband (UWB) radar, among which, ultra-wideband bioradar has become the current The mainstream of bio-radar technology research. Impulse-radio Ultra-wideband (IR-UWB) has become a research hotspot in the field of post-disaster search and rescue due to its excellent performance and simple structure. channel IR-UWB bio-radar system to achieve.

Slow time: The detection time of the radar to the target, the unit is second (s).

Fast time: The time it takes for a pulse to propagate, in nanoseconds (ns).

As shown in Figure 3, in the actual detection, the echo signal sampled by the IR-UWB radar is stored in a two-dimensional matrix R(m,n) after sampling, integration and amplification, where m is the row vector and n is the Column vector, the horizontal axis in the figure represents slow time, and the vertical axis represents fast time. Fast time can be converted into detection distance according to the propagation speed of electromagnetic waves in the medium, and the unit is meters (m).

The calculation relationship between fast time and distance is: distance (m) = fast time (ns) × propagation speed of electromagnetic waves in the medium (m/ns)/2.

Fast time signal: A signal along the fast time dimension at a certain moment, that is, a column vector of a two-dimensional matrix.

Slow time signal: A signal along the slow time dimension at a distance point, i.e. a row vector of a two-dimensional matrix.

Inflection point: Mathematically refers to the point that changes the upward or downward direction of the curve. Intuitively, the inflection point is the point where the tangent line crosses the curve (ie, the concave-convex boundary of the curve).

The principle realized by the invention is: when the emission signal of the biological radar is irradiated to a stationary object, the radar echo signal is a stable fixed value, but when it is irradiated to a living body such as a human or an animal, the body surface caused by the breathing of the living body will be Micro fluctuations cause fluctuations in the original radar echo signal, so we can detect living bodies through these small fluctuations.

The invention discloses a three-channel bio-radar system with one transmitter and three receivers. The block diagram of the system is shown in Figure 1. Under the control of a PRF oscillator, the pulse generator operates at a certain Pulse Repeat Frequency (PRF). Generate a pulse signal. The generated pulse signal is divided into two channels: one channel is adjusted and adjusted by the transmitting circuit to form a bipolar pulse signal, which is radiated through the transmitting antenna; the other channel is sent to the delay unit to generate a series of delay time under the control of the microprocessor. The distance gate is actually a sampling pulse signal with a very short duration. Under the trigger of this signal, the receiving circuit can selectively receive and sample the radar echo. The signal radiated by the transmitting antenna is reflected when it encounters an object, and the reflected radar echo is received by the receiving antenna and sent to the receiving circuit for selective sampling, integration and amplification under the trigger of the distance gate, and then through analog-to-digital conversion. The ADC (Analog to Digital Converter, ADC) forms the radar echo signal. The radar echo signal is sent to the processing and display terminal through the WiFi module under the control of the microprocessor for signal processing and result display. The IR-UWB bio-radar system consists of three channels: each transmitting antenna and one receiving antenna form a transceiver channel. The three receiving antennas shown in Figure 1 can be combined with the transmitting antennas to form three channels.

Figure 2 is a schematic plan view of the three-channel IR-UWB bioradar system for multi-target penetration detection and identification, each radar antenna is placed close to a 24cm thick brick wall (the penetration range of the multi-channel radar is: within 200cm thickness Brick wall), 1.15 meters from the ground (corresponding to the average height of an adult chest, the height can be adjusted freely, the radar detection area is a cone space with an opening angle of 120 degrees, as long as the target can be detected within the detection area), where The transmitting antenna Tx and the receiving antenna Rx1 are placed next to each other in the middle, forming the middle channel of the radar system. The data in the middle channel is mainly used to determine the number of targets and the radial distance of each target; the receiving antenna Rx2 and the receiving antenna Rx3 are respectively Placed at 1.2 meters on the left and 1.2 meters on the right of the transmitting antenna Tx (the distance between Rx2, Rx3 and the transmitting antenna is adjustable from 0.3m-1.2m, different distances correspond to different lateral resolutions, the larger the distance, the more The larger the lateral resolution of the target, the more accurate the positioning), and the left channel and the right channel are formed with the transmitting antenna. The data of the left channel and the right channel are mainly used to locate the direction of the target.

Figure 4(a) and (b) show the waveforms of the radar echo fast time signal and slow time signal, respectively. The window width of the time window of the IR-UWB radar determines the length of the fast time signal. In the experimental setup of the present invention, the time window width of a fast time signal is set to 80ns, and the corresponding detection distance is 12m. Each fast time signal consists of 8192 sampling points, and the time interval between each two fast time signals is Ts=0.0625s, that is to say, the sampling frequency of the slow time signal is fs=1/Ts=16Hz, which satisfies Nyquis The requirements of the special sampling law for the sampling of human respiratory signals.

Example 1

This embodiment discloses a method for identifying and locating multiple human targets based on a multistatic IR-UWB biological radar signal, including the following steps:

Step 1: The transmitting antenna of the three-channel IR-UWB bio-radar transmits radar pulses to the target position. The three channels include a middle channel, a left channel and a right channel. The radar pulses are reflected at the target position and pass through the three-channel IR-UWB. The three receiving antennas of the bioradar obtain the radar echo signal E(m,n) in the middle channel, the radar echo signal E _left (m,n) in the left channel and the radar echo signal E _right (m,n) in the right channel, where , m is the sampling sequence number in the fast time direction, n is the sampling sequence number in the slow time direction, and m and n are positive integers;

Step 2: Use E(m,n), E _left (m,n) and E _right (m,n) obtained in step 1 to obtain the middle channel energy signal E ₆ (l) and the left channel energy signal through signal preprocessing respectively E _6left (l) and right channel energy signal E _6right (l);

Step 3: Inflection point extraction is performed on E ₆ (l), E _6left (l) and E _6right (l) obtained in step 2, and the secondary inflection point signal E ₉ (l) of the middle channel and the secondary inflection point signal E of the left channel are obtained. _9left (l) and right channel secondary inflection point signal E _9right (l);

Step 4: Obtain the position of the top three values and the top three values of each secondary inflection point signal obtained in step 3, and the top three values include the maximum value, the second maximum value and the third maximum value;

Step 5.1: Calculate the peak-to-background ratio and the mean value of the correlation coefficient of the first three values of E ₉ (1), obtain the number of targets by the peak-to-background ratio and mean value of the correlation coefficient of the first three values of E ₉ (l), include:

A) If V _EtoB1 <σ _1N , or σ _{1N ≤V EtoB1} <σ _1Y and r _m1 <σ _rm1 , the judgment result is no target;

B) If σ _{1N ≤V EtoB2} <σ _2N , or σ _{2N ≤V EtoB2} <σ _2Y and r _m2 <σ _rm1 , the discrimination result is a single target;

C) If r _m3 >σ _rm3 , or σ _rm2 <r _m3 ≤σ _rm3 and V _EtoB3 ≥σ _1Y , the discrimination result is three-objective;

D) In addition to A) B) C) other cases, the discrimination result is a dual target;

Wherein, V _EtoB1 represents the peak-to-background ratio of the maximum value of E ₉ (l), V _EtoB2 represents the peak-to-background ratio of the second largest value of E ₉ (l), and V _EtoB3 represents the peak of the third largest value of E ₉ (l) - Background ratio, r _m1 means the mean value of the correlation coefficient of the largest value, r _m2 means the mean value of the correlation coefficient of the second largest value, r _m3 means the mean value of the correlation coefficient of the third largest value, σ _1N means no target threshold, σ _1Y means there is a target Threshold, σ _2N represents dual-target threshold, σ _2Y represents multi-target threshold and σ _1N <σ _2N <σ _1Y <σ _2Y , σ _rm1 , σ _rm2 and σ _rm3 represent correlation threshold and σ _rm2 <σ _rm1 <σ _rm3 ;

Step 5.2: According to the number of targets identified in step 5.1, determine the radial distance of the target and the orientation of the target in combination with the positions of the first three values of E ₉ (l), E _9left (l) and E _9right (l), including:

a) If the recognition result is a single target, the radial distance of the target is determined by the position of the maximum value of E ₉ (1), and the orientation of the target is determined by the position of the maximum value of E _9left (1) and E _9right (1);

b) If the recognition result is a double target, the radial distance of the first target is determined by the position of the maximum value of E ₉ (l), and the first target is determined by the position of the maximum value of E _9left (l) and E _9right (l). The bearing of one target, then the target radial distance of the second by the position of the second largest value of _{E9(l), determined by the position of the second largest value of E9left (} l) and _E9right (l) the orientation of the second target;

c) If the recognition result is three targets, the radial distance of the first target is determined by the position of the maximum value of E ₉ (l), and the first target is determined by the position of the maximum value of E _9left (l) and E _9right (l). The bearing of one target, then the target radial distance of the second by the position of the second largest value of _{E9(l), determined by the position of the second largest value of E9left (} l) and _E9right (l) The bearing of the second target, and finally the radial distance of the third target is determined by the position of the third largest value of _{E9(l), by the position of the third largest value of E9left (} l) and _E9right (l) determine the orientation of the third target;

Identify and locate the target according to the number of targets, the radial distance of the target and the orientation of the target.

Preferably, in step 5.1, σ _1N =2, σ _1Y =3, σ _2N =2.4, σ _2Y =3.5, σ _rm1 =0.8, σ _rm2 =0.75 and σ _rm3Y =0.88.

Specifically, the method for obtaining the radar echo signal in step 1 is as follows:

The schematic diagram of IR-UWB radar detecting human breathing is shown in Figure 5. Assuming that the initial distance between the chest wall surface of the human target and the radar is d ₀ , the breathing of the human body will cause the thoracic cavity to expand and contract periodically. When the displacement of the chest wall is a sinusoidal function _x (t) about the slow time, then the actual distance d( _t ) between the chest wall surface of the human target and the radar will be periodic around d0 according to the breathing frequency fr of the human body ground changes:

d(t)=d ₀ +x(t)=d ₀ +Arsin(2πf _r t)

In the formula, t represents the slow time, x(t) represents the change of the displacement of the chest wall when the human body is breathing, and Ar represents the maximum amplitude of the human body breathing.

Because the environment within the detection range is static, and the human target remains static, only the chest wall motion caused by breathing, then the impulse response h(t,τ) of the radar system will change with time like the breathing motion:

表示静态背景目标的脉冲回波成分，其中α_i和τ_i分别为第i个静态目标脉冲回波的幅度和在快时间维度上的延时，α_vδ(τ-τ_v(t)表示人体目标呼吸运动的脉冲回波成分，其中α_v为脉冲回波的幅度，τ_v(t)为人体目标脉冲回波在快时间维度上的延时变化，可以表示为：where t represents slow time, τ represents fast time,

Represents the pulse echo component of the static background target, where α _i and τ _i are the amplitude of the ith static target pulse echo and the delay in the fast time dimension, respectively, α _v δ(τ-τ _v (t) represents The pulse echo component of the breathing motion of the human target, where α _v is the amplitude of the pulse echo, and τ _v (t) is the delay change of the human target pulse echo in the fast time dimension, which can be expressed as:

where c is the propagation speed of the electromagnetic wave in the vacuum, τ _r is the maximum delay of the breathing motion in the fast time dimension, τ ₀ is the delay time between the radar wave on the surface of the human chest wall and the radar (initial distance), namely

If impulse distortion and other nonlinear effects are ignored, the radar echo signal can be regarded as the convolution of the radar transmit impulse and the system impulse response. Then, without considering the existence of noise, the echo signal of the radar at time t is:

In the formula, p(τ) is the transmitted pulse of the radar, and "*" represents the convolution operation.

In order to explain the signal model more clearly, the schematic diagram of the pulse echo of human respiration is shown in Figure 6. It can be seen from the figure that the delay of the pulse echo of human breathing in the fast time dimension changes with the slow time, while the delay of the pulse echo of the static target is unchanged.

In actual detection, the IR-UWB radar system samples each point on each pulse waveform at each discrete time τ=mT _f (m=1,2,...,M) along the fast time direction, while the slow time The direction samples the pulse waveform once at each discrete time instant t= _nTs (n=1, 2, . . . , N). The sampled echo signal is stored as a (M×N) two-dimensional array E, and the elements in the array E are represented by E(m,n):

The signal E(m,n) is a two-dimensional signal, m is the sampling sequence number in the fast time direction, and n is the sampling sequence number in the slow time direction.

Specifically, the signal preprocessing in step 2 includes the following sub-steps:

Step 2.1: Accumulate distances for E(m,n), E _left (m,n) and E _right (m,n) respectively, taking the middle channel as an example:

The IR-UWB radar system used in this study has 8192 sampling points and a time window of 80ns. If the original radar echo data is directly processed, the calculation amount is large and the operation is slow, which is unfavorable for the real-time detection and identification. In the two-dimensional original echo data E(m,n) received by the IR-UWB radar, since the modulation methods of the radar echoes at the adjacent distance points in the fast time dimension are roughly the same and have a certain correlation, it can be On the premise of not affecting the useful information, the original echo data E(m,n) of the radar is firstly accumulated along the fast time dimension:

其中

表示向下取整。通过大量的实验研究表明，窗宽Q＝40时，算法取得最优效果。那么经过距离累积后，原始回波数据E(m,n)的8192个对应距离点上的慢时间信号，就减少到E₁(l,n)的200个(即L＝200)对应距离点上的快时间信号，大大降低了雷达数据处理过程的运算量，减少了探测所需的运算时间，提高了探测效率。同时，沿着快时间维度的距离累积，也相当于对雷达回波的快时间信号做平滑滤波，一定程度上可以抑制快时间信号上的高频干扰。In the formula, E ₁ (l,n) (l=1,2,...L) is the echo data after distance accumulation, Q is the window width accumulated along the fast time dimension, and L is the accumulated distance in the fast time dimension points, and

in

Indicates rounded down. A large number of experimental studies show that the algorithm achieves the optimal effect when the window width Q=40. Then after distance accumulation, the slow time signals at the 8192 corresponding distance points of the original echo data E(m,n) are reduced to 200 corresponding distance points (ie L=200) of E ₁ (l,n) The fast time signal on the radar greatly reduces the calculation amount of the radar data processing process, reduces the calculation time required for detection, and improves the detection efficiency. At the same time, the distance accumulation along the fast time dimension is also equivalent to smoothing and filtering the fast time signal of the radar echo, which can suppress the high frequency interference on the fast time signal to a certain extent.

Step 2.2: Multiply the accumulated signal in step 2.1 by the exponential gain curve G(l) of formula I to perform attenuation compensation;

Among them, V _h represents the ratio of the maximum value of the radar echo data to the amplitude of the reflected echo of the human target, P _human is the position of the human target, l represents the fast time sequence number after distance accumulation, l=1,2,...,L and L is a positive integer;

Since the radar wave is severely attenuated during the propagation of the medium, the amplitude of the reflected echo from the far-end object interface will be greatly reduced, which will make it difficult for the far-end object to be detected. Therefore, it is necessary to identify the reflected echo from the interface. , and compensate the radar echo E ₁ (l,n) after distance accumulation. The current ultra-wide-spectrum radars (mainly ground penetrating radars) are equipped with automatic gain adjustment functions, which can amplify the far-end echo data through segmented linear or exponential gain adjustment of the radar echoes. However, due to the lack of electromagnetic wave propagation The prior knowledge of the medium interface information, the accuracy of the gain calculation is not high, and the noise is often excessively amplified due to the inaccurate gain, while the real interface reflection echo cannot be properly amplified because the gain is small. Ultimately, the probability of the target being misjudged and missed is greatly increased.

The method of segmental compensation for attenuation needs to calculate different gains according to the possible attenuation of each segment of the echo. The calculation process is too complicated, but it is difficult to calculate the gain accurately, and it is easily affected by noise, resulting in erroneous compensation. In actual detection, we can first detect and calculate the position of the human target in an uncompensated form, and take the position information of the human target as prior knowledge. The position and the corresponding reflected echo amplitude are the compensation benchmarks, and the exponential gain compensation method is used to compensate the attenuation of the radar echo data in the fast time dimension. After compensation, the signal is processed again according to the signal processing process, and the target is detected and distinguished. .

The gain curve is calculated as follows:

Assume an ideal exponential gain curve of the form e ^K×τ , where K is an unknown constant. For the preprocessed data, divide the maximum value A _max of E ₁ (l,n) (usually the maximum value of radar echo data) by the amplitude A _human of the reflected echo of the human target (that is, the position of the human target P _human the corresponding amplitude in the radar echo data), the resulting ratio is counted as V _h . Taking this ratio V _h as the ideal gain value where the radar echo position is P _human , the exponential gain curve that changes with the fast time sequence number l can be calculated, and the calculated exponential gain curve is correlated with the radar echo data on the fast time axis. Multiplication, the attenuation compensation for radar echo data is realized, the signal after attenuation compensation is E ₂ (l,n), and E ₂ (l,n)=G(l)E ₁ (l,n).

Step 2.3: remove static clutter from the signal after attenuation compensation in step 2.2;

In the process of radar-type life detection, the direct wave of radar and the reflection of stationary objects within the detection range will form strong background clutter in the radar echo signal. Overwhelmed by clutter, as shown in Figure 3, in the original echo of the radar, almost no life signal of the human target can be seen, only background clutter can be seen. But under ideal conditions, these background clutters are static, called static clutter, and only the vital signal of the human target changes with time. The waves are completely filtered out, leaving only human vital signs:

where E ₃ (l,n) is the radar echo signal after background removal.

Figure 8 shows the two-dimensional radar signal simulated by matlab software. It can be seen from the figure that the static clutter near 15ns and 65ns does not change with the slow time, while the breathing signal of the human target near 40ns is along the slow time dimension. regular changes. After the static clutter is eliminated by the de-averaging method on the simulated two-dimensional radar signal, the static clutter components in the echo are completely removed, leaving only the breathing signal of the human target, as shown in Figure 9.

Step 2.4: remove the linear trend of the signal after removing the static clutter in step 2.3;

The hardware of the IR-UWB radar system is often accompanied by the drift of the echo baseline in the process of collecting data. Linear baseline drift will lead to energy leakage of echo data in low frequency bands, which will affect the detection and recognition of human target breathing signals. Therefore, the present invention adopts Linear Trend Subtraction (LTS) to remove the linear baseline drift in the radar echo signal. LTS estimates the DC component and low-frequency linear drift trend of the echo signal E ₃ (l,n) in the slow time dimension by linear least squares fitting, and then subtracts it from the echo data:

n＝[0,1,2...,N-1]^T，这里y为一个N行2列的行列式，1_N为一个长度是N且元素都是1的列向量，N为E₃中快时间信号的个数。线性趋势消除以后再将E₄ ^T转置得到E₄(l，n)。In the formula, E ₄ represents the radar data processed by LTS, and E ₃ represents the radar data E ₃ (l,n) after de-averaging; E ₄ ^T and E ₃ ^T are their transposed determinants respectively.

n=[0,1,2...,N-1] ^T , where y is a determinant with N rows and 2 columns, 1 _N is a column vector of length N and all elements are 1, and N is E ₃ The number of medium and fast time signals. E ₄ (l, n) is obtained by transposing E ₄ ^T after the linear trend is eliminated.

Step 2.5: perform low-pass filtering on the signal after the linear trend removal in step 2.4 in the slow time dimension;

Since the hardware of the IR-UWB radar system will inevitably generate noise during the working process, these noises are high-frequency noises compared to the human breathing signal, and the breathing signal of the human target is a narrow-band low-frequency quasi-periodic signal, so , in order to effectively filter out high-frequency interference and further improve the signal-to-noise ratio of the radar echo, the present invention performs low-pass filtering on the radar echo signal in the slow time dimension:

E ₅ (l,q)=E ₄ (l,n)*h(t)

In the formula, E ₅ (l,q) is the filtered radar data, "*" represents the convolution operation, and h(t) is the impulse function of the Finite Impulse Response (FIR) filter. According to the breathing frequency of the human body, the cutoff frequency of the low-pass filter is set to 0.5Hz, and the order of the filter is 120. The low-pass filtered radar echo signal is E ₅ (l,q).

Step 2.6: Accumulate the low-pass filtered signals in step 2.5 along the slow time axis to obtain the middle channel energy signal E ₆ (1), the left channel energy signal E _6left (1) and the right channel energy signal E _6right (1).

In the experiment, the experimental data t _s = 80 seconds are collected each time. According to the sampling frequency of the slow time signal f _s = 16 Hz, it can be known that the data collected each time contains 16 × 80 = 1280 fast time signals, that is, the low-pass filtered The radar echo signal is Q = t _s f _s = 1280 in E ₅ (l, q), L is the value after distance accumulation 200 (200 is obtained from 8192 sampling points of the fast time signal through distance accumulation, mainly In order to reduce the amount of calculation, the value can be determined freely, and it can be accumulated to 200-1000 points with less calculation without affecting the signal quality).

Because the steps of de-averaging and low-pass filtering in preprocessing require a convergence process, the first 200 (200 here is related to the sum of the orders of de-averaging and low-pass filtering in preprocessing. The lower the order, the value can be taken The smaller the value is, the higher the order is, and the larger the value needs to be.) The fast time signal is not used as the basis for target detection and identification, and is eliminated. We set 1000 in E ₅ (l, q) (the 1000 here is determined by the length of the sampling time, the sampling frequency of 16Hz corresponds to about 62.5 seconds of data, the longer the signal is collected, the greater the value) fast time After the signal (200-1200) takes the absolute value, it is accumulated along the slow time axis to form the energy signal E ₆ (l).

The energy signal E ₆ (l), (l=1,2,...,200) is a one-dimensional signal, the abscissa is the fast time, corresponding to the distance (m), and the ordinate is formed by the accumulation along the slow time energy magnitude. After the aforementioned series of signal processing, the amplitude of the energy signal is closely related to the life signal of the living body.

Further, the inflection point extraction in step 3 includes the following sub-steps:

Step 3.1: Remove the direct waves from E ₆ (l), E _6left (l) and E _6right (l) obtained in step 2, to obtain E ₇ (l), E _7left (l) and E _7right (l), with the middle An example of a channel:

Remove the direct wave to eliminate the influence of the direct wave between the transmitting and receiving antennas on the target discrimination, that is, set the data of the first 50 points of E ₆ (l) to zero (the range of zero-setting is the first 10-50 points), and form a new after removal of the direct wave. The obtained energy signal E ₇ (l);

Step 3.2: Extract the inflection point from the signal obtained in Step 3.1, and obtain the primary inflection point signals E ₈ (l), E _8left (l) and E _8right (l), taking the middle channel as an example:

Find all the primary inflection points of the energy signal E ₇ (l), all the points that satisfy the following conditions are called primary inflection points, set the values corresponding to the points that do not meet the conditions of the inflection points to zero, and the points that satisfy the conditions are based on the original amplitude value , the original position is stored in a new one-dimensional array to form an inflection point signal E ₈ (l);

Due to the complexity of electromagnetic wave propagation, by removing the energy signal E ₇ (l) of the direct wave, it is difficult to directly find the detected target from the signal. Therefore, it is necessary to remove the interference. One of the main interferences in the radar echo that affects target recognition is the side lobe interference near the peak, and the side lobes of the peak show a characteristic of attenuation to both sides. Therefore, in this step, we identify the peak signal by inflection point extraction, that is, when the amplitude of the energy signal at a certain point is larger than that of the adjacent points on the left and right sides of it, the point is judged as a primary inflection point. The first inflection point discrimination rule is as follows:

E ₇ (l)>E ₇ (l+1)∩E ₇ (l)>E ₇ (l-1)

Step 3.2: Extract the inflection point from the signal obtained in step 3.2, and obtain the secondary inflection point signal E ₉ (l) of the middle channel, the secondary inflection point signal E _9left (l) of the left channel and the secondary inflection point signal E _9right (l) of the right channel, with For example, the middle channel:

Find all the quadratic inflection points of the primary inflection point signal E ₈ (l). All the points that satisfy the following conditions are called quadratic inflection points. Click the original amplitude and store the original position in a new one-dimensional array to form a secondary inflection point signal E ₉ (l).

Judging from the actual processing results, the primary inflection point signal E ₈ (l) still has interference, which makes it impossible to accurately identify targets, especially multiple targets. Therefore, as an optimized implementation, the primary inflection point signal E ₈ (1) is extracted twice to obtain a secondary inflection point signal, that is, all points in the primary inflection point signal E ₈ (1) that satisfy the following conditions are reserved , identify these points as quadratic inflection points:

E ₈ (l)>E ₈ (l+a)∩E ₈ (l)>E ₈ (lb)

where E ₈ (l+a) is the first non-zero value to the right of E ₈ (l), and E ₈ (lb) is the first non-zero value to the left of E ₈ (l).

The energy signal E ₉ (l) after taking the second inflection point extraction removes most of the peak side lobe interference, and preserves the signal energy at each target position to the greatest extent, which can greatly improve the multi-target recognition ability.

Specifically, step 4.1: obtain the maximum value of each quadratic inflection point signal obtained in step 3 through the bubble sort algorithm, and remove the maximum value by zeroing 16 consecutive amplitude signals after the maximum value position. trailing;

Step 4.2: Obtain the second largest value of each quadratic inflection point signal through the bubble sort algorithm, and remove the drag of the second largest value by setting the adjacent 16 consecutive amplitude signals after the second largest value position to zero. tail;

Step 4.3: Obtain the third largest value of each quadratic inflection point signal through the bubble sort algorithm.

Through the bubble sort algorithm, find the maximum value E _9-max1 on the secondary inflection point signal E ₉ (l) of the intermediate channel, the position is recorded as l _max1 , and the adjacent 16 consecutive amplitudes after the maximum position l _max1 The value signal is set to zero to remove the "tail" of the maximum value, and then the second largest value E _9-max2 is found by the bubble sort algorithm, and the position is recorded as l _max2 , and the adjacent 16 after the position of E ₉ (l) The consecutive amplitude signals are set to zero to remove the "tail" of the second largest value, and the third largest value E _9-max3 is found by the bubble sort algorithm, and the position is l _max3 ;

Because the human chest wall has a certain thickness, the echo signal E ₉ (l) not only has a high energy amplitude at the position of the human chest wall surface, but also has a higher energy amplitude at a distance behind the position. We will It is called "smearing", and the phenomenon of "smearing" affects the discrimination of the number of multiple targets. Because the marked E ₉ (l) maximum value E _9-max1 is located at the position l _max1 cannot be other targets within a certain distance range, but can only be the "smear" generated by the target at this position, in order to remove this "smear""Tailing", where the adjacent 16 consecutive amplitude signals after the maximum position l _max1 are set to zero.

The left channel and the right channel are also processed according to the above method, and the first three maximum values E _9left-max1 , E _9left-max2 , and E _9left-max3 on the secondary inflection point signal E _9left (l) of the left channel are successively found. (that is, the maximum value, the second maximum value and the third maximum value in the signal E _9left (l)), where the maximum value and the second maximum value need to be removed after finding the "tail" and then look for the next maximum value. The first three large value positions are respectively recorded as E _9left-max1 , E _9left-max2 , and E _9left-max3 .

Find the first three maxima E _9rihgt-max1 , E _9rihgt-max2 , E _9rihgt-max3 on the secondary inflection point signal E _9right (l) of the right channel (that is, the maximum value in the signal E _9right (l), the second The largest value and the third largest value), where the largest value and the second largest value need to be removed after finding the "tail" and then look for the next large value. The positions of the first three large values found are recorded as E _9right-max1 , E _9right-max2 , E _9right-max3 .

且E_9-max1表示E₉(l)的最大值、

且E_9-max2表示E₉(l)的第二大值、

且E_9-max3表示E₉(l)的第三大值；Specifically, obtain the peak-to-background ratio V _EtoB1 of the maximum value of E ₉ (1), the second-largest peak-to-background ratio V _EtoB2 , and the third-largest peak-to-background ratio V _EtoB3 , where _Bave represents the background mean ,

And E _9-max1 represents the maximum value of E ₉ (l),

And E _9-max2 represents the second largest value of E ₉ (l),

And E _9-max3 represents the third largest value of E ₉ (l);

Preferably, the mean correlation coefficient _rm1 at the maximum value, the mean correlation coefficient _rm2 at the second maximum value, and the mean correlation coefficient _rm3 at the third maximum value are obtained through formula II:

Among them, i represents the serial number of the correlation coefficient and i=1,2,3,4,5,6, k represents the serial number of the first three values and k=1 represents the serial number of the maximum value, k=2 represents the second largest value Serial number, k=3 represents the serial number of the third largest value, r _ik represents the correlation coefficient of the first three values, r _mk represents the mean value of the correlation coefficient of the first three values, E ₅ (l, q) represents the slow time obtained in step 2.5 The signal of the intermediate channel after low-pass filtering in the dimension, E _maxk (q) represents the signal at the position of the first three values of E ₅ (l, q), and E _(maxk+(i-4)) (q) represents the same as The signal of the first three positions adjacent to the first three values of E ₅ (l,q), (E _(maxk+(i-3)) (q) represents the first three values of E ₅ (l, q) Signals at the next three positions adjacent to each other, Q represents the total number of signal sampling points in the slow time direction of E ₅ (l, q) and Q is a positive integer, q represents the qth signal sampling point in the slow time direction and q is positive integer.

If the sampling duration is 60s, and the sampling frequency of the slow-time signal in this example is fs=16Hz, the total number of sampling points Q of the signal is 16×60=960 points.

Through the aforementioned steps, we have calculated the first three maximum values E ₉ -max1 , E _9-max2 , E _9-max3 on the secondary inflection point signal E ₉ (l) of the intermediate channel, and their corresponding positions l _max1 , l _max2 , l _max3 ; the peak-to-background ratio V _EtoBk of the target is calculated, and the mean values r _m1 , r _m2 and r _m3 of the correlation coefficients between the slow signal at this position and its six adjacent slow signals are calculated.

The relationship formula of the relevant factors affecting V _EtoBk is as follows. The number of targets is related to the serial number of the first three values, and each large value corresponds to a target:

a _k is a constant coefficient. Under the same background energy level (including noise), the larger the chest wall area of the human target k, the stronger the reflected energy, the larger the breathing amplitude, the stronger the breathing signal energy, and the closer the human target k is to the radar. The stronger the respiratory signal energy is, the larger the peak-to-background ratio V _EtoBk of the target k is at this time, on the contrary, the smaller V _EtoBk is. In the multi-target scenario, the amplitudes of the first few maximum values E _9- max1 , E _9-max2 , and E _9-max3 of the secondary inflection point signal decrease in turn, while the background mean value _Bave of the secondary inflection point signal is the same value, Therefore, for different targets, we need to set different thresholds for identification.

The relational formula for the relevant factors affecting r _mk is as follows:

b _k is a constant coefficient. Under the condition of the same noise level, the thicker the chest wall thickness (the distance from the front chest to the back) of the human target k, the more regular the breathing signal (slow time signal) corresponding to each distance point is reflected. The breathing signals of r _1k ~ r _6k are all high and consistent, so the average value r _mk is higher; and the higher the degree of breathing regularity of the human target k, the breathing signal at the target distance point is adjacent to its left and right. The higher the correlation of several slow-time signals of , the larger the mean r _mk value. According to these characteristics, we determine the following principles to discriminate the number of human targets.

Specifically, in step 5, the position of the target is determined by the position l _left-maxk of the first three values of E _9left (l) and the position l _right-maxk of the first three values of E _9right (l), including:

a) If |l _left-maxk -l _right-maxk |≤2, then the target is on the central axis, and the central axis is the connection between the receiving antenna and the transmitting antenna of the middle channel;

b) If (l _left-maxk -l _right-maxk )≤-2, the target is in the left area of the central axis, and the left area is the area where the left channel transmit antenna on one side of the central axis is located;

c) If (l _left-maxk -l _right-maxk )>2, the target is in the right area of the central axis, and the right area is the area where the right channel transmit antenna on one side of the central axis is located.

Specifically, the positioning method is described by taking target 2 as an example. Previously, we have calculated the positions l _max2 , l _left-max2 , and l _right-max2 of the second largest value of the secondary inflection point signal of the middle channel, the left channel, and the right channel, respectively. The radial distance of the target 2 is the distance corresponding to l _max2 : 12*l _max2 /200 meters. The following principles are used to determine whether the target 2 is located in the left half area, the right half area or the radial center axis of the detection area.

a) If |l _left-max2 -l _right-max2 |≤2 (that is, the absolute value of the distance difference between the second largest position of the two channels is less than 0.12 meters), then the target 2 is on the central axis;

b) If (l _left-max2 -l _right-max2 )≤-2, then target 2 is in the left half area of the central axis;

c) If (l _left-max2 -l _right-max2 )>2, then target 2 is in the right half area of the central axis;

According to the same principle, use the positions l _max1 , l _left-max1 , l _right-max1 of the maximum value of the secondary inflection point signal of the middle channel, left channel and right channel to locate target 1, and use the middle channel, left channel and right channel to locate target 1. Position l _max3 , l _left-max3 , l _right-max3 of the third largest value of the secondary inflection point signal to locate target 3, thus completing the positioning of all three targets.

Example 2

This embodiment adopts the multi-human target recognition and positioning method and the three-channel IR-UWB biological radar. On the basis of Embodiment 1, the following technical features are also disclosed, and the verification experiment of through-wall detection and positioning is carried out in the laboratory. The experimental subjects were three healthy young men (target 1, target 2, and target 3) to carry out the detection experiment under the condition of penetrating a single brick wall, and gave the target number discrimination and positioning results. Among them, target 1 is standing still at a distance of 2.3 meters behind the wall on the right side; target 2 is standing still at a distance of 5.7 meters behind the wall and towards the left; target 3 is standing still at a distance of 7.8 meters behind the wall in the direction of the central axis.

After using the three-channel IR-UWB bio-radar to detect through the wall, the radar echo signals of the three channels are processed and calculated to obtain V _EtoB1 = 4.01, V _EtoB2 = 2.93, V _EtoB = 2.17, r _m1 = 0.97, r _m2 =0.95, r _m1 =0.89. According to the multi-target number identification method process, it can be determined that the detection result is three human targets. According to the positions l _max1 , l _max2 , and l _max3 of the first three maximum values of the secondary inflection point signal after removing the "smear" in the middle channel, the radial distances of the three targets are 2.34 meters, 5.64 meters, and 7.86 meters, respectively. After determining the radial distance of each target, combine the position of the first three maximum values of the left channel and the first three maximum values of the right channel to determine that target 1 is located in the right direction of the detection area, target 2 is located in the left direction of the detection area, and target 3 is located in the detection area. The direction of the central axis of the area. The detection results are consistent with the actual standing distribution of the three targets, the radar identification and positioning results are correct, and the three-target identification and positioning results are shown in Figure 13.

Claims (7)

1. A multi-human target identification and positioning method based on multi-base IR-UWB biological radar signals is characterized by comprising the following steps:

step 1: the transmitting antenna of the three-channel IR-UWB biological radar transmits radar pulse to a target position, the three channels comprise a middle channel, a left channel and a right channel, the radar pulse is reflected at the target position, and a middle channel radar echo signal E (m, n) and a left channel radar echo signal E are obtained through three receiving antennas of the three-channel IR-UWB biological radar_left(m, n) and a right channel radar echo signal E_right(m, n), wherein m is the sampling ordinal number in the fast time direction, n is the sampling ordinal number in the slow time direction, and m and n are positive integers;

step 2: e (m, n) and E obtained in the step 1_left(m, n) and E_right(m, n) respectively obtaining intermediate channel energy signals E through signal preprocessing₆(l) Left channel energy signal E_6left(l) And right channel energy signal E_6right(l)；

And step 3: for E obtained in step 2₆(l)、E_6left(l) And E_6right(l) Performing inflection point extraction to obtain a secondary inflection point signal E of the middle channel₉(l) Left channel quadratic inflection point signal E_9left(l) And the right channel quadratic inflection point signal E_9right(l)；

And 4, step 4: obtaining the first three values and the positions of the first three values of each secondary inflection point signal obtained in the step 3, wherein the first three values comprise a maximum value, a second large value and a third large value;

step 5.1: calculation of E₉(l) The peak-to-background ratio and the correlation coefficient mean of the first three values of (a) by E₉(l) The number of the targets obtained by the peak-to-background ratio and the mean value of the correlation coefficient of the first three values comprises:

A) if V_EtoB1<σ_1NOr σ_1N≤V_EtoB1<σ_1YAnd r is_m1<σ_rm1If the result is no target, judging that the target is not available;

B) if σ is_1N≤V_EtoB2<σ_2NOr σ_2N≤V_EtoB2<σ_2YAnd r is_m2<σ_rm1If yes, judging that the result is a single target;

C) if r is_m3>σ_rm3Or σ_rm2<r_m3≤σ_rm3And V is_EtoB3≥σ_1YIf yes, judging that the result is three targets;

D) under other conditions except A) B) C), the judgment result is a dual target;

wherein, V_EtoB1Represents E₉(l) Peak-to-background ratio of maximum, V_EtoB2Represents E₉(l) Second largest peak-to-background ratio, V_EtoB3Represents E₉(l) Third highest peak-to-background ratio, r_m1Mean value of the correlation coefficient representing the maximum value, r_m2Mean value of the correlation coefficient, r, representing the second largest value_m3Mean value of the correlation coefficient, σ, representing the third largest value_1NRepresenting a no target threshold, σ_1YIndicating a target threshold, σ_2NRepresenting two target thresholds, σ_2YRepresents a multi-target threshold value and_1N<σ_2N<σ_1Y<σ_2Y，σ_rm1、σ_rm2and σ_rm3Represents a correlation threshold value and σ_rm2<σ_rm1<σ_rm3；

Step 5.2: according to the number of the targets identified in the step 5.1, combining with E₉(l)、E_9left(l) And E_9right(l) The positions of the first three values determine the radial distance of the target and the azimuth of the target, and the method comprises the following steps:

a) if the recognition result is a single target, pass E₉(l) Determines the radial distance of the target by E_9left(l) And E_9right(l) The position of the maximum value of (a) determines the orientation of the target;

b) if the identification result is a double target, passing through E₉(l) Determines the radial distance of the first target by E_9left(l) AndE_9right(l) Determines the position of the first object and then passes E₉(l) Determines the target radial distance of the second, by E_9left(l) And E_9right(l) Determines the bearing of the second object;

c) if the recognition result is three targets, pass E₉(l) Determines the radial distance of the first target by E_9left(l) And E_9right(l) Determines the position of the first object and then passes E₉(l) Determines the target radial distance of the second, by E_9left(l) And E_9right(l) Determines the position of the second object, and finally passes through E₉(l) Determines the radial distance of a third target by E_9left(l) And E_9right(l) Determines the bearing of a third target;

and identifying and positioning the targets according to the number of the targets, the radial distance of the targets and the azimuth of the targets.

2. The multi-human target recognition and positioning method based on multi-base IR-UWB biological radar signals of claim 1, wherein in step 5.1, σ is_1N＝2，σ_1Y＝3，σ_2N＝2.4，σ_2Y＝3.5，σ_rm1＝0.8、σ_rm20.75 and σ_rm3＝0.88。

3. The multi-human target recognition and positioning method based on multi-base IR-UWB biological radar signals as claimed in claim 1, wherein the signal preprocessing in step 2 comprises the following sub-steps:

step 2.1: for E (m, n), E_left(m, n) and E_right(m, n) performing distance accumulation respectively;

step 2.2: multiplying the signals accumulated in the step 2.1 by an exponential gain curve G (l) of a formula I to perform attenuation compensation;

wherein, V_hRepresenting the ratio, P, of the maximum value of the radar echo data to the amplitude of the reflected echo of the human target_humanIs the human target position, L represents the fast time sequence number after the accumulation of the distance, L is 1,2, …, L is a positive integer;

step 2.3: removing static clutter from the signal subjected to attenuation compensation in the step 2.2;

step 2.4: performing linear trend elimination on the signals after the static clutter is removed in the step 2.3;

step 2.5: low-pass filtering the signal with the linear trend eliminated in the step 2.4 in a slow time dimension;

step 2.6: accumulating the signals subjected to low-pass filtering in the step 2.5 along a slow time axis to obtain an intermediate channel energy signal E₆(l) Left channel energy signal E_6left(l) And right channel energy signal E_6right(l)。

4. The multi-human target recognition and positioning method based on multi-base IR-UWB biological radar signals of claim 1, wherein the inflection point extraction of step 3 comprises the following sub-steps:

step 3.1: for E obtained in step 2₆(l)、E_6left(l) And E_6right(l) Removing direct wave to obtain E₇(l)、E_7left(l) And E_7right(l)；

Step 3.2: extracting inflection points from the signals obtained in the step 3.1 to obtain a primary inflection point signal E₈(l)、E_8left(l) And E_8right(l)；

Step 3.2: extracting inflection points from the signals obtained in the step 3.2 to obtain a secondary inflection point signal E of the intermediate channel₉(l) Left channel quadratic inflection point signal E_9left(l) And the right channel quadratic inflection point signal E_9right(l)。

5. The multi-human target recognition and positioning method based on multi-base IR-UWB biological radar signals as claimed in claim 1, wherein the step 4 comprises the following sub-steps:

step 4.1: obtaining the maximum value of each secondary inflection point signal obtained in the step 3 through a bubble sorting algorithm, and removing the tailing of the maximum value by setting adjacent 16 continuous amplitude signals behind the position of the maximum value to zero;

step 4.2: obtaining a second large value of each secondary inflection point signal through a bubble sorting algorithm, and removing the tailing of the second large value by setting adjacent 16 continuous amplitude signals behind the second large value to zero;

step 4.3: and obtaining a third maximum value of each quadratic inflection point signal through a bubble sorting algorithm.

When the first three maximum values are obtained, the adjacent 16 continuous amplitude signals after the maximum value position and the second maximum value position are set to zero, and the most trailing is removed.

6. The multi-human target recognition and positioning method based on multi-base IR-UWB biological radar signals of claim 3, wherein the mean value r of the correlation coefficient at the maximum is obtained by formula II_m1The mean value r of the correlation coefficient at the second largest value_m2And the mean value r of the correlation coefficient at the third largest value_m3：

Where i denotes the index of the correlation coefficient and i is 1,2,3,4,5,6, k denotes the index of the first three largest values and k is 1 denotes the index of the largest value, k is 2 denotes the index of the second largest value, k is 3 denotes the index of the third largest value, r denotes the index of the third largest value_ikThe correlation coefficient, r, representing the first three largest values_mkMean value of the correlation coefficients representing the first three major values, E₅(l, q) represents the signal of the intermediate channel obtained in step 2.5 after low-pass filtering in the slow time dimension, E_maxk(q) represents E₅Signals at the first three maximum positions of (l, q), E_(maxk+(i-4))(q) represents a group represented by₅(l, q) signals of the first three positions adjacent to the first three maximum positions, (E)_(maxk+(i-3))(q) represents a group represented by₅(l, Q) signals at the last three positions adjacent to the first three maximum positions, Q representing E₅(l, Q) the total number of sampling points of the signal in the slow time direction and Q is a positive integer, Q represents the Q-th signal sampling point in the slow time direction and Q is a positive integer.

7. The multi-human target recognition and positioning method based on multi-base IR-UWB biological radar signals of claim 6, wherein in step 5, through E_9left(l) Position l of the first three major values_left-maxkAnd E_9right(l) Position l of the first three major values_right-maxkDetermining the orientation of the target, comprising:

a) if | l_left-maxk-l_right-maxkIf the | is less than or equal to 2, the target is on a central axis, and the central axis is a connecting line of the receiving antenna and the middle channel transmitting antenna;

b) if (l)_left-maxk-l_right-maxk) If the central axis is less than or equal to-2, the target is in the left area of the central axis, and the left area is the area where the left channel transmitting antenna on one side of the central axis is located;

c) if (l)_left-maxk-l_right-maxk) And if the target is in the right area of the central axis, the right area is the area where the right channel transmitting antenna on one side of the central axis is located.

Images