A non-contact identity recognition method, system, medium, device and terminal

Abstract

This invention belongs to the field of identity recognition technology and discloses a non-contact identity recognition method, system, medium, device, and terminal. It extracts, demodulates, and differentiates the phase of vital movement signals of a target individual collected by an AWR1642 sensor. Sine fitting is used to denoise the differentiated phase signal to obtain cardiac motion information. Adaptive weight adjustment and local perturbation are used to improve the Wild Dog algorithm's support vector machine optimization, achieving identity recognition. A Butterworth low-pass filter separates respiratory and heartbeat signals, and the obtained heartbeat frequency is compared with that of a smart bracelet to verify the feasibility of using FMCW millimeter-wave radar for vital signal detection. Experimental results show that the heartbeat frequency obtained by this invention has a Pearson correlation coefficient of 0.96 with that obtained by contact methods, indicating the feasibility of FMCW radar collecting vital signals. Furthermore, the accuracy of using the improved Wild Dog optimization algorithm to optimize the SVM for real individual identity recognition can reach 88%, demonstrating that the identity recognition method of this invention can effectively improve the accuracy of recognition.

Description

Technical Field

[0001] This invention belongs to the field of identity recognition technology, and in particular relates to a contactless identity recognition method, system, medium, device and terminal. Background Technology

[0002] Currently, traditional methods for collecting individual cardiac signals are contact-based, such as using patch electrodes or contact sensors to collect the heart's electrical impulse signals. These methods not only limit the application scenarios—for example, individuals with burns or infectious diseases are unsuitable for wearable devices—but also require prolonged wearing of the device during the identification process, leading to a poor user experience. To address these issues, researchers have focused on contactless identification technologies. Since 1975, when Lin et al. proposed a system using radar to detect vital signs, demonstrating the feasibility of using radar for detecting vital signs, individual vital signals collected by radar contain environmental noise and random body movements. These interferences can overwhelm the individual's cardiac motion signals. Therefore, how to suppress environmental noise and eliminate random body movements has become a focus for researchers.

[0003] In 2012, Xu et al. proposed a vital sign signal processing method based on zero higher-order cumulant Gaussian noise. However, some clutter in the same frequency range as breathing and heartbeat signals still existed in the experiment, requiring further suppression. In 2013, Gu et al. proposed a camera-based technique to eliminate random body motion, using camera tracking of body motion to compensate for the acquired radar vital signs and reduce phase shift caused by body motion. In 2015, Lv et al. used gradient descent and extended differential cross-multiplication algorithms to solve the DC offset problem, ensuring a wide range of vital signs, and used curve fitting techniques to compensate for body motion. In 2016, Ren et al. used complex signal demodulation and arctangent demodulation techniques to demodulate the radar echo signal. However, the method was limited by the selection of the window function. In the same year, Hu Xikun et al. applied wavelet transform to separate signal components containing heartbeat and breathing signals from non-contact vital signs. Experimental results showed that this method could effectively solve the problem of separating breathing and heartbeat signals in different scenarios. In 2018, Liu Zhenyu et al. proposed a vital signal detection method based on intrinsic mode component filters. After adaptive ensemble empirical mode decomposition of the vital signals acquired by radar, the method selects the respiratory and heartbeat components separately using component filters, and reconstructs the human respiratory and heartbeat signals using the selected components. Experimental results show that the respiratory and heartbeat signals obtained by this method have good signal-to-noise ratios. In the same year, Lv et al. introduced matched filters to obtain weak vital signals hidden under large-scale body movements. Experimental results show that the proposed method is robust. In 2021, Liu Luyao proposed a signal processing method based on wavelet analysis and autocorrelation calculation. Experiments tested the processing effect of this method at distances of 0.5m, 1m, 1.5m, 2m, 2.5m, and 3m between 10 requesters and the radar. Experimental results show that the error value of the respiratory rate obtained by this method is less than 1.65%, and the error value of the heart rate is less than 1.83%. Moreover, as the distance between the requester and the radar increases, the signal-to-noise ratio of the signal decreases, proving the effectiveness of this method in reducing environmental noise in vital signals.

[0004] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0005] (1) The individual vital signals collected by traditional frequency-modulated continuous wave millimeter-wave radar are subject to interference from environmental noise and random body movements, which can overwhelm the individual's heart movement signals.

[0006] (2) Existing vital sign signal processing methods need to suppress noise in the same frequency range as breathing and heartbeat signals, and echo signal demodulation methods are relatively limited by the selection of window functions. Summary of the Invention

[0007] To address the problems existing in the prior art, the present invention provides a contactless identity recognition method, system, medium, device and terminal, and particularly relates to a contactless identity recognition method, system, medium, device and terminal based on optimized support vector machine (SVM).

[0008] This invention is implemented as follows: the non-contact identification method includes: acquiring the vital movement signals of a target individual using an AWR1642; performing phase extraction, demodulation, and differential analysis on the signals; and using sine fitting to denoise the differential phase signals to obtain the target individual's cardiac motion information; improving the Wild Dog algorithm's support vector machine by adaptively adjusting weights and local perturbations to achieve identification; and separating respiratory and heartbeat signals using a Butterworth low-pass filter, comparing the obtained heartbeat frequency with a smart bracelet to verify the feasibility of using FMCW millimeter-wave radar for vital signal detection.

[0009] Furthermore, the contactless identification method includes the following steps:

[0010] Step 1, Acquiring vital signs: Extracting the phase changes caused by the human body's position and heart movement from the signals collected by the FMCW millimeter-wave radar, and obtaining heart movement signals based on the phase information;

[0011] Step 2, optimize the wild dog algorithm: introduce adaptive weight adjustment, add a random perturbation mechanism after the wild dog individual updates its position during the prey encirclement stage, and realize the optimization of the wild dog algorithm;

[0012] Step 3, Feature Extraction: Construct a cardiac motion model, locate the reference points of cardiac motion signals using the extreme value method, and obtain feature vectors by calculating the distance and amplitude between each reference point.

[0013] Furthermore, obtaining the human body's position and the phase changes caused by heart movement in step one includes:

[0014] If the thoracic cavity moves a distance ΔR, the phase change between consecutive measurements is:

[0015]

[0016] λ is the signal wavelength, derived from... f = S*t The relationship between the distance R between the human body and the frequency f of the intermediate frequency signal is derived as follows:

[0017]

[0018] Where B is the signal bandwidth, c is the speed of light, and T is the speed of light. cLet S be the frequency modulation duration and S be the frequency change rate. Further, the acquisition of the cardiac motion signal in step one includes: acquiring human cardiac motion signals using FMCW millimeter-wave radar and estimating the target distance using Fast Fourier Transform; extracting the phase signal caused by cardiac motion at the location of the human body; and finally removing noise from the phase signal.

[0019] Furthermore, the optimization of the wild dog algorithm in step two includes:

[0020] (1) Introduce adaptive weight adjustment, as shown in the following expression:

[0021]

[0022] Where Ite is the iteration number, t is the current iteration number, f(x) is the fitness function, best(f(x)) is the best fitness value, and worst(f(x)) is the worst fitness value. After adding adaptive weight adjustment, the formula for updating the position of the wild dog siege behavior is as follows:

[0023]

[0024] x best For the optimal starting position of the wild dogs, x k x i Let w be the initial position of the k-th and i-th wild dogs, w be the weight, n be the number of wild dogs surrounding the prey, and x be the initial position of the k-th and i-th wild dogs, respectively. newi Let be the updated position of the i-th wild dog, and β be a scaling factor, a random number between -2 and 2, which can change the size of the wild dog's trajectory. The formula for updating the position due to pursuit behavior is as follows:

[0025]

[0026] Where β1 is a uniformly random number in the range [-2, 2], and β2 is a uniformly random number generated in the range [-1, 1]. r This represents the random location of the wild dogs. The formula for updating the location due to scavenging behavior is as follows:

[0027]

[0028] The formula for updating position based on survival behavior is as follows:

[0029]

[0030] Where, x r1 x r2 The positions of two distinct random wild dogs in the group.

[0031] (2) After the wild dog individual updates its position during the prey encirclement phase, a random perturbation mechanism is added.

[0032] The disturbance method is as follows:

[0033]

[0034] In the formula, t is the current iteration number, r is the position of the wild dog population, x'new is the updated position after adding the disturbance mechanism, and xnew is the original position without adding the disturbance.

[0035] Furthermore, the steps of using the improved wild dog optimization algorithm are as follows:

[0036] (1) Initialize the population position;

[0037] x = lb+(ub - lb)rand(p,2);

[0038] In the formula, lb is the lower limit of the optimization variable, ub is the upper limit of the optimization variable, the position of each individual is initialized, p is the size of the wild dog population, and rand returns a random number uniformly distributed within the interval p*2.

[0039] (2) Obtain the initial fitness value. Taking the accuracy rate of the support vector machine prediction value as the fitness function, substitute the initialized position information into the fitness function to obtain the initial fitness value.

[0040] (3) Update the population position: In the African wild dog algorithm, the probabilities of each wild dog choosing the three behaviors of siege, pursuit, and scavenging are determined by p and q. Select p = 0.5 and q = 0.7. rand1 and rand2 are uniformly random numbers within [0,1]. If rand1 < p && rand2 < q, the wild dog chooses to siege, and then update the position according to the position update formula for the wild dog siege behavior; if rand1 < p && rand2 >= q, the wild dog chooses to pursue, and then update the position according to the position update formula for the pursuit behavior; if rand1 >= p, the wild dog chooses to scavenge, and then update the position according to the position update formula for the scavenging behavior. If the survival rate < 0.3, then update the position according to the position update formula for the survival behavior.

[0041] (4) Add local disturbance and update the population position.

[0042] (5) Judge the maximum number of iterations. If the maximum number of iterations is not satisfied, repeat steps (3) and (4). If satisfied, exit.

[0043] Furthermore, the feature extraction in step three includes:

[0044] (1) Construct a heart motion model

[0045] Select the simulated heart motion signal. Equivalent the life signal model to the superposition of sine signals, then:

[0046] y(t) = A hsin(f h t)+A b sin(f b t);

[0047] In the formula, f h f represents the respiratory signal frequency. b Represents the heartbeat signal frequency, A h Represents the amplitude of the heartbeat, A b This represents the amplitude of breathing.

[0048] (2) Feature extraction

[0049] The baseline points of cardiac motion signals are located using the extreme value method; the distances and amplitudes between each baseline point are calculated to obtain time and amplitude characteristics, which are then combined to form a feature vector.

[0050] Another object of the present invention is to provide a contactless identity recognition system applying the aforementioned contactless identity recognition method, the contactless identity recognition system comprising:

[0051] The vital signal acquisition module is used to extract the phase changes caused by the human body's position and heart movement from the signals collected by the FMCW millimeter-wave radar, and to acquire the heart movement signal based on the phase information.

[0052] The Wild Dog Algorithm Optimization Module is used to optimize the Wild Dog Algorithm by introducing adaptive weight adjustment and adding a random perturbation mechanism after the position update of individual wild dogs during the prey encirclement stage.

[0053] The feature extraction module is used to construct a cardiac motion model. It locates the reference points of cardiac motion signals using the extreme value method and obtains feature vectors by calculating the distance and amplitude between each reference point.

[0054] Another object of the present invention is to provide a computer device including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the contactless identification method.

[0055] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the contactless identification method.

[0056] Another objective of this invention is to provide an information data processing terminal for implementing the aforementioned contactless identity recognition system.

[0057] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0058] First, addressing the technical problems existing in the prior art and the difficulty of solving them, this paper closely analyzes, in conjunction with the technical solution to be protected by this invention and the results and data obtained during the research and development process, how the technical solution of this invention solves the technical problems, and the inventive technical effects brought about by solving these problems. The specific description is as follows:

[0059] To address the problem that environmental noise and random bodily movements can obscure individual cardiac motion signals in individual vital signals acquired by Frequency Modulated Continuous Wave (FMCW) millimeter-wave radar, this invention proposes using sinusoidal fitting to denoise the differentiated phase signal and employing adaptive weight adjustment and local perturbation to improve the Wild Dog algorithm's support vector machine optimization. To verify the effectiveness of the denoising method, this invention uses frequency filtering to separate respiratory and heart rate signals, comparing the obtained heart rate with that acquired by a smart bracelet. To verify the effectiveness of the improved Wild Dog algorithm, this invention establishes a vital signal model using cardiac motion characteristics, comparing the recognition accuracy of the improved Wild Dog algorithm, FOA, WOA, GWO, and DOA on analog signals. Experimental results show that the sinusoidal fitting technique provided by this invention can extract weak heartbeats from vital signals. Furthermore, the improved Wild Dog optimization algorithm achieves an accuracy of 95.6731% in identifying the model signals of 11 test subjects, which is superior to other optimization algorithms. The accuracy in identifying the real signals of test subjects can reach 88%, proving the effectiveness of the non-contact identification method of this invention.

[0060] This invention, based on the analysis of vital signal processing acquired by FMCW millimeter-wave radar, proposes a method for extracting cardiac motion signals from noisy radar signals by analyzing the periodicity of cardiac motion. Furthermore, addressing the low accuracy of traditional wilddog optimization algorithms, this invention proposes an improved wilddog optimization algorithm to optimize the support vector machine. This invention first introduces the working principle of FMCW millimeter-wave radar; secondly, it uses sine fitting to acquire cardiac motion information and separates respiratory and heartbeat signals using a Butterworth low-pass filter. The obtained heartbeat frequency is compared with that of a smart bracelet to verify the feasibility of using FMCW millimeter-wave radar for vital signal detection. Finally, this invention uses the improved wilddog optimization algorithm to optimize the support vector machine for identity recognition. Experimental results show that the improved wilddog optimization algorithm provided by this invention can effectively improve the accuracy of identity recognition.

[0061] Second, considering the technical solution as a whole or from a product perspective, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows:

[0062] This invention proposes a method for identifying human vital signs using FMCW millimeter-wave radar. The method utilizes sinusoidal fitting to extract cardiac motion signals and employs a wilddog optimization algorithm to create an identification model. Experimental results show that the heart rate obtained using this non-contact identification method has a Pearson correlation coefficient of 0.96 with that obtained using a contact method, demonstrating the feasibility of using FMCW radar to collect vital signs. Furthermore, the improved wilddog optimization algorithm proposed in this invention achieves an accuracy of 88% in SVM-optimized identification, indicating that the proposed identification method effectively improves the accuracy of identification.

[0063] Third, as supplementary evidence of the inventive step of the claims of this invention, it is also reflected in the following important aspects:

[0064] The technical solution of this invention solves a technical problem that people have long wanted to solve but have never been able to: In this invention, sinusoidal fitting technology is used to denoise the differential vital signals acquired by AWR1642, thereby realizing the extraction of cardiac motion signals of the target individual. Attached Figure Description

[0065] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0066] Figure 1 This is a flowchart of the contactless identity recognition method provided in the embodiments of the present invention;

[0067] Figure 2 This is a flowchart of the process for obtaining vital signs provided in an embodiment of the present invention;

[0068] Figure 3 This is a flowchart of the improved Wild Dog optimization algorithm for optimizing SVM provided in an embodiment of the present invention;

[0069] Figure 4A This is a simulated cardiac motion waveform diagram provided in an embodiment of the present invention;

[0070] Figure 4B This is a feature point map of the signal provided in the embodiments of the present invention;

[0071] Figure 5 This is a schematic diagram of the contactless identity recognition system provided in an embodiment of the present invention;

[0072] Figure 6 This is a schematic diagram showing the location of the target individual provided in an embodiment of the present invention;

[0073] Figure 7 This is a schematic diagram of separate respiration and heartbeat provided in an embodiment of the present invention;

[0074] Figure 8 This is a comparison chart of the average heart rate of 11 individuals tested using millimeter-wave radar and smart bracelets, provided by an embodiment of the present invention.

[0075] Figure 9 This is a comparison chart of the original heart signal acquired by millimeter-wave radar, the heart signal fitted by sine wave, and the heart signal fitted by polynomial wave, provided in the embodiments of the present invention.

[0076] Figure 10 This is a graph showing the accuracy of the improved wild dog optimized SVM for recognizing real individual signals, provided in an embodiment of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0078] To address the problems existing in the prior art, the present invention provides a contactless identity recognition method, system, medium, device, and terminal. The present invention will be described in detail below with reference to the accompanying drawings.

[0079] like Figure 1 As shown, the contactless identity recognition method provided in this embodiment of the invention includes the following steps:

[0080] S101, Acquiring vital signs: Extracting the phase changes caused by the human body's position and heart movement from signals collected by FMCW millimeter-wave radar, and acquiring heart movement signals based on the phase information;

[0081] S102, Optimize the wild dog algorithm: Introduce adaptive weight adjustment, and add a random perturbation mechanism after the wild dog individual updates its position during the prey encirclement stage to optimize the wild dog algorithm;

[0082] S103, Feature Extraction: Construct a cardiac motion model, locate the reference points of cardiac motion signals using the extreme value method, and obtain feature vectors by calculating the distance and amplitude between each reference point.

[0083] As a preferred embodiment, the contactless identity recognition method provided by this invention specifically includes the following steps:

[0084] 1. Obtaining life signals

[0085] 1.1 Working principle of FMCW millimeter-wave radar

[0086] The FMCW millimeter-wave radar system includes components such as a transmitter, receiver, radio frequency (RF) module, and analog-to-digital converter (ADC). Its frequency increases periodically and linearly. The radar operation is as follows: the transmitter RF module transmits linearly frequency-modulated (FM) pulses generated by the synthesizer through an antenna. When the pulse encounters a human body, it is reflected back. The receiver antenna receives the reflected FM pulses. The transmitted and received signals are mixed, and finally, an intermediate frequency (IF) signal is generated in the mixer. The IF signal can be represented as:

[0087]

[0088] In the formula, λ is the signal wavelength; B is the bandwidth; f c T is the starting frequency of the linear frequency modulated signal; c t is the bandwidth of the linear frequency modulated signal; c is the speed of light; j is the complex unit; and R is the distance between the radar and the human body.

[0089]

[0090] The phase is:

[0091]

[0092] Vibrations caused by human vital signals are hidden in phase changes. The movement of the human heart is a vibration with a small amplitude. To obtain information about chest cavity movement, this invention needs to obtain the phase change of the FMCW signal over time. That is, if the chest cavity moves a distance ΔR, then the phase change between consecutive measurements is:

[0093]

[0094] Depend on f = S*t The relationship between the distance R to the human body and the frequency f of the intermediate frequency signal can be derived:

[0095]

[0096] Therefore, the present invention obtains the phase change caused by the movement of the heart and the position of the human body through equations (4) and (5).

[0097] 1.2 Acquisition of cardiac motion signals

[0098] Environmental noise and random body movements contained in individual vital signals acquired by radar can mask cardiac motion signals. The impact of noise on vital signals can be reduced by extracting phase information indicating the body's location from signals acquired by FMCW millimeter-wave radar. The flowchart for acquiring vital signals in this invention is as follows: Figure 2As shown, firstly, the human heart motion signal is acquired using FMCW millimeter-wave radar, and the target distance is estimated using Fast Fourier Transform (FFT); then, the phase signal caused by the heart motion at the location of the human body is extracted, and finally, noise is removed from the phase signal.

[0099] Because the sinoatrial node generates electrophysiological waves, which are transmitted through branches of each physiologically active conduction bundle, myocardial cells undergo electrophysiological transmission, causing the heart to contract and expand rhythmically. During the heartbeat cycle, the rapid compression of the ventricular volume generates short pulse motions, causing displacement on the chest wall. The vital signals collected by radar are chest displacement signals, which are caused by both respiration and heartbeat. The waveforms of respiration and heartbeat are non-stationary signals, approximating sine waves. Utilizing this characteristic, sine fitting technology is used to extract cardiac motion signals.

[0100] 2. Optimize the Wild Dog algorithm

[0101] 2.1 Wild Dog Optimization Algorithm

[0102] During a hunt, African wild dogs in a pack use vocalizations for location. Pack members use vocalizations to help coordinate actions and track prey until a successful hunt is achieved. The Dingo Optimization Algorithm (DOA) simulates this behavior of hunting dogs, mimicking the hunting behavior of African wild dogs: encirclement, pursuit, carrion cannibalism, and survival. It iteratively simulates group hunting behavior, i.e., finding the optimal value.

[0103] 2.1.1 Siege behavior

[0104] A siege involves several wild dogs attacking the target. The African Wild Dogs algorithm randomly selects a certain number of wild dogs to search around the globally optimal location and arrives at a new location.

[0105]

[0106]

[0107] In the formula, n is the number of wild dogs attacking, and β is... random integers in x; best The optimal starting position for the wild dog, x k x i These are the initial positions of the kth and ith wild dogs, respectively.

[0108] 2.1.2 Pursuit and Manhunt

[0109] The pursuit behavior involves searching the vicinity of the globally optimal individual, and its formula is as follows:

[0110]

[0111] In the formula, β1 is a uniformly random number in [-2, 2], and β2 is a uniformly random number in [-1, 1]; x r It refers to a random individual within a pack of wild dogs.

[0112] 2.1.3 Carrion-eating behavior

[0113] The formula for achieving scavenging behavior is as follows:

[0114]

[0115] 2.1.4 Survival

[0116] Each wild dog's survival probability is related to its fitness value. Individuals with a lower survival probability need to return to the vicinity of the current best individual to forage for food in order to improve their survival probability.

[0117]

[0118]

[0119] Formula (10) calculates the survival probability of the wild dog; where f(x) is the fitness function, best(f(x)) is the best fitness value, and worst(f(x)) is the worst fitness value. If sr < 0.3, then formula (11) is used to update the position; where x r1 x r2 These are two distinct random individuals within a pack of wild dogs.

[0120] 2.2 Optimize the Wild Dog Algorithm

[0121] 2.2.1 Adaptive Weight Adjustment

[0122] To address the imbalance between the early-stage search and later-stage optimization in the algorithm iteration, an adaptive weight adjustment method is introduced. Its specific expression is as follows:

[0123]

[0124] Where Ite is the iteration number, and t is the current iteration number. After incorporating adaptive weight adjustment, the formula for updating the position in the wild dog siege behavior is as follows:

[0125]

[0126] The formula for updating the location during a pursuit is as follows:

[0127]

[0128] The formula for updating the position of scavenging behavior is as follows:

[0129]

[0130] When the wild dog population forages while surrounding the prey, adaptive adjustment of weights can accelerate the convergence speed of the algorithm and effectively avoid falling into local optima.

[0131] The formula for updating the position in the survival behavior is as follows:

[0132]

[0133] 2.2.2 Local perturbation

[0134] To further improve the global convergence accuracy of the algorithm and avoid the algorithm falling into local extrema. After the wild dog individuals update their positions during the stage of surrounding the prey, a perturbation mechanism is added. The specific perturbation method is as follows:

[0135]

[0136] In the formula, t is the current iteration number, and r is the position of the wild dog population. The flow chart of optimizing SVM using the improved wild dog optimization algorithm is as Figure 3 shown, and the specific steps are as follows:

[0137] Step1: Initialize the population position according to the following formula:

[0138] x = lb + (ub - lb)rand(p, 2) (18)

[0139] In the formula, lb is the lower limit of the optimization variable, ub is the upper limit of the optimization variable, and the position of each individual is initialized; where p is the size of the wild dog population.

[0140] Step2: Obtain the initial fitness value. In this invention, the accuracy rate of the support vector machine prediction value is used as the fitness function. The initialized position information is brought into the fitness function to obtain the initialized fitness value.

[0141] Step3: Update the population position. In the African wild dog algorithm, the probabilities for each wild dog to choose the three behaviors of surrounding, pursuing, and scavenging are determined by p and q. Select p = 0.5, q = 0.7, rand1 and rand2 are uniform random numbers within [0, 1]. If rand1 < p && rand2 < q, the wild dog chooses to surround and updates the position according to formula (13). If rand1 < p && rand2 >= q, the wild dog chooses to pursue and updates the position according to formula (14). If rand1 >= p, the wild dog chooses to scavenge and updates the position according to formula (15). If the survival rate < 0.3, the position is updated according to formula (16).

[0142] Step 4: Add local perturbations and update the population position according to formula (17).

[0143] Step 5: Determine the maximum number of iterations. If the maximum number of iterations is not met, repeat steps 3 and 4. If the maximum number of iterations is met, exit.

[0144] 3. Feature Extraction

[0145] 3.1 Cardiac Exercise Model

[0146] To verify the effectiveness of the algorithm, this invention selected simulated cardiac motion signals for experiments. Based on the analysis in section 1.2, the vital signal model is considered to be a simple superposition of sinusoidal signals.

[0147] y(t)=A h sin(f h t)+A b sin(f b t) (19)

[0148] In the formula, f h f represents the respiratory signal frequency. b Represents the heartbeat signal frequency, A h Represents the amplitude of the heartbeat, A b This represents the amplitude of breathing. Figure 4A This is a simulated heartbeat waveform. The heart rate, respiratory rate, and amplitude information for a typical adult are shown in Table 1.

[0149] Table 1. Body surface displacement parameters related to respiration and heartbeat.

[0150]

[0151] 3.2 Feature Extraction

[0152] This invention uses the extreme value method to locate five reference points for cardiac motion signals, namely A, B, C, D, and E, as follows: Figure 4B As shown.

[0153] By calculating the distance and amplitude between each base point, four time features and four amplitude features are obtained, which form a feature vector as shown in Table 2.

[0154] Table 2 Extracted Feature Information

[0155]

[0156] The contactless identity recognition system provided in this embodiment of the invention includes:

[0157] The vital signal acquisition module is used to extract the phase changes caused by the human body's position and heart movement from the signals collected by the FMCW millimeter-wave radar, and to acquire the heart movement signal based on the phase information.

[0158] The Wild Dog Algorithm Optimization Module is used to optimize the Wild Dog Algorithm by introducing adaptive weight adjustment and adding a random perturbation mechanism after the position update of individual wild dogs during the prey encirclement stage.

[0159] The feature extraction module is used to construct a cardiac motion model. It locates the reference points of cardiac motion signals using the extreme value method and obtains feature vectors by calculating the distance and amplitude between each reference point.

[0160] To demonstrate the inventiveness and technical value of the technical solution of this invention, this section provides specific product or related technology application examples of the technical solution claimed.

[0161] This invention first extracts, demodulates, and differentiates the phase of the vital movement signals of the target individual collected by the AWR1642; then, it uses sine fitting technology to denoise the differentiated phase signals; next, it extracts features from the cardiac motion signals of the target individual; finally, it uses an improved wild dog optimization algorithm to optimize the support vector machine, thereby more accurately identifying the individual's identity.

[0162] 1. Effectiveness Analysis of the Improved Wild Dog Algorithm

[0163] This invention simulates 11 people, with 10 data points for each person, forming 110 samples. The feature matrix composed of feature information is then input into an improved Wild Dog Optimization Algorithm for identity recognition.

[0164] Table 4 summarizes the experimental results of this invention, using recognition accuracy as the performance parameter. To demonstrate the fairness of the results, the parameters for the five algorithms were set as follows: population size was set to 30, and the maximum number of iterations was set to 100. The experimental results show that the accuracy of the Wild Dog optimized algorithm can reach 95.6731%, proving the effectiveness of the Wild Dog optimized algorithm.

[0165] 2. Radar Parameter Design

[0166] This invention uses a Texas Instruments (TI) millimeter-wave AWR1642 radar sensor, which operates in the frequency range of 77–81 GHz. When acquiring data, the data is transmitted to a PC via a USB interface, and subsequent signal processing is performed on MATLAB.

[0167] The FMCW radar parameters are configured as follows: each pulse contains 128 chirs, and each chirp acquires 100 data points. The pulse transmission interval is 50ms, i.e., the slow sampling rate fslow is 20Hz, the ADC sampling rate (fast sampling rate) is 2MHz, and the duration of one chirp is Td = 50μs. The sweep bandwidth is 4GHz. Figure 5This is a schematic diagram of a contactless identity recognition system provided in an embodiment of the present invention.

[0168] 3. Feasibility Analysis

[0169] Figure 6 This is a map showing the location of the human body. Phase information at the target individual's location is extracted. To verify the accuracy and effectiveness of the method, radar sensors were used to collect 3 minutes of heart rate and respiratory data from 6 adult men and 5 adult women, with 10 data points collected from each subject, for a total of 110 data points. This constitutes the experimental database of this invention. The operating system used in the experiment was 64-bit Windows 10. The heart rate was calculated using the method of this invention and compared with the heart rate data collected by a smart bracelet. The collected data were processed in 1-minute intervals, and then a Butterworth low-pass filter was used to separate the respiratory and heart rate signals. The frequency with the largest heart rate signal amplitude was found to obtain the heart rate. Figure 7 These are the respiratory and heartbeat signals separated after passing through a low-pass filter. Figure 8 This is a comparison chart of the average heart rate of 11 individuals tested using millimeter-wave radar and smart bracelets.

[0170] This invention uses the Pearson correlation coefficient to calculate the similarity between the average heart rate obtained from radar and the average heart rate measured by a smart bracelet, using the following formula:

[0171]

[0172] In the formula, x i For the non-contact heart rate of the i-th individual, y i Let i be the contact heart rate of the i-th individual. The values ​​are the individual's average non-contact heart rate and contact heart rate, respectively. Using formula (20), r = 0.96 is obtained, and the experimental results demonstrate the feasibility of the FMCW millimeter-wave radar of this invention for detecting vital signs.

[0173] 4. Obtaining life signals

[0174] Figure 9 A comparison of raw heart signals acquired by millimeter-wave radar with heart signals fitted by sine and polynomial fitting.

[0175] To compare the quality of curve fitting techniques, three parameters are used as evaluation criteria: sum of squares (SSE), root mean square error (RMSE), and coefficient of determination (R-squared). SSE is the sum of squares of the errors between corresponding points in the fitted data and the original data. The closer the SSE is to 0, the better the model selection and fit, and the more successful the data prediction. The formula is as follows:

[0176]

[0177]

[0178]

[0179] Where y i The value at point i is the true value. The fitted value at point i. The coefficient of determination is the average of the true values. As shown in the formula for the coefficient of determination, the closer the coefficient of determination is to 1, the stronger the explanatory power of the equation's variables on the original curve, and the better the model fits the data. Table 3 shows the parameter comparison between polynomial fitting and sine fitting obtained from formulas (21), (22), and (23). Table 3 shows that using sine fitting for cardiac motion signals is superior to polynomial fitting.

[0180] Table 3 Comparison of Curve Fitting Results

[0181]

[0182] 5. Identity Recognition

[0183] This invention uses an improved Wild Dog optimization algorithm to optimize a support vector machine based on collected human body data to achieve identity recognition, and the resulting recognition accuracy is as follows: Figure 10 As shown.

[0184] Table 4 Comparison of Accuracy of Different Algorithms

[0185]

[0186] As shown in Table 4, the accuracy of the model signal identification achieved by optimizing SVM using the improved Wild Dog optimization algorithm can reach 95.6731%, which is significantly higher than the accuracy of 95.1923% achieved by the standard Wild Dog optimization algorithm, demonstrating the effectiveness of the improved Wild Dog optimization algorithm.

[0187] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0188] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A non-contact identity recognition method, characterized by, The non-contact identity recognition method includes the following steps: Step 1, obtaining vital signs: extracting the position of the human body and the phase change caused by the heart movement from the signals collected by the FMCW millimeter-wave radar, and obtaining the heart movement signal according to the phase information; Step 2, feature extraction: constructing a heart movement model, locating the reference point of the heart movement signal by the extreme value method, and obtaining the feature vector by calculating the distance and amplitude between each base point; Step 3, optimizing the wild dog algorithm: introducing an adaptive adjustment weight, adding a random perturbation mechanism after the wild dog individuals update their positions in the prey surrounding stage to realize the optimization of the wild dog algorithm, optimizing the support vector machine by using the improved wild dog optimization algorithm, and using the feature vector as the input to more accurately identify the individual identity; The optimization of the wild dog algorithm in Step 3 includes: (1) Introducing an adaptive adjustment weight, and the expression is as follows: ； where Ite is the number of iterations, and t is the current iteration number; After adding the adaptive adjustment weight, the position update formula for the wild dog siege behavior is as follows: ； The position update formula for the pursuit behavior is as follows: ； The position update formula for the scavenging behavior is as follows: ； The position update formula for the survival behavior is as follows: ； (2) Adding a random perturbation mechanism after the wild dog individuals update their positions in the prey surrounding stage; The perturbation method is as follows: ； In the formula, t is the current iteration number, and r is the position of the wild dog population; The steps of using the improved wild dog optimization algorithm are as follows: (1) Initializing the population position; ； In the formula, lb is the lower limit of the optimization variable, ub is the upper limit of the optimization variable, initializing the position of each individual, and p is the size of the wild dog population; (2) Obtaining the initial fitness value, using the accuracy rate of the support vector machine prediction value as the fitness function, and substituting the initialized position information into the fitness function to obtain the initialized fitness value; (3) Updating the population position: In the African wild dog algorithm, the probabilities of each wild dog choosing the three behaviors of siege, pursuit, and scavenging are determined by p and q. Select p = 0.5 and q = 0.

7. Rand1 and rand2 are uniformly distributed random numbers within [0, 1]. If rand1 < p && rand2 < q, the wild dog chooses to siege, and then updates the position according to the position update formula for the wild dog siege behavior; if rand1 < p && rand2 >= q, the wild dog chooses to pursue, and then updates the position according to the position update formula for the pursuit behavior; if rand1 >= p, the wild dog chooses to scavenge, and then updates the position according to the position update formula for the scavenging behavior. If the survival rate < 0.3, then update the position according to the position update formula for the survival behavior; (4) Adding local perturbation and updating the population position; (5) Judging the maximum number of iterations. If the maximum number of iterations is not satisfied, repeat steps (3) and (4). If satisfied, exit.

2. The non-contact identity recognition method of claim 1, wherein, The obtaining of the position of the human body and the phase change caused by the heart movement in Step 1 includes: If the chest moves a distance ∆R, the phase change between consecutive measurements is: ； Depend on , , The relationship between the distance R between the human body and the frequency f of the intermediate frequency signal is derived as follows: ； Among them, the obtaining of the heart movement signal includes: using the FMCW millimeter-wave radar to collect the human heart movement signal, and using the fast Fourier transform to estimate the distance of the target; extracting the phase signal caused by the heart movement at the position where the human body is located, and finally removing the noise from the phase signal.

3. The contactless identity recognition method as described in claim 1, characterized in that, The feature extraction in step two includes: (1) Constructing a cardiac motion model Choosing simulated cardiac motion signals, and equating the vital signal model to a superposition of sinusoidal signals, then: ； In the formula, Represents respiratory signal frequency. Represents the frequency of the heartbeat signal. Represents the amplitude of heartbeat. Represents respiratory amplitude; (2) Feature extraction The baseline points of cardiac motion signals are located using the extreme value method; the distances and amplitudes between each baseline point are calculated to obtain time and amplitude characteristics, which are then combined to form a feature vector.

4. A contactless identity recognition system applying the contactless identity recognition method as described in any one of claims 1 to 3, characterized in that, Contactless identification systems include: The vital signal acquisition module is used to extract the phase changes caused by the human body's position and heart movement from the signals collected by the FMCW millimeter-wave radar, and to acquire the heart movement signal based on the phase information. The Wild Dog Algorithm Optimization Module is used to optimize the Wild Dog Algorithm by introducing adaptive weight adjustment and adding a random perturbation mechanism after the individual wild dogs update their positions during the prey encirclement phase. The feature extraction module is used to construct a cardiac motion model. It locates the reference points of cardiac motion signals using the extreme value method and obtains feature vectors by calculating the distance and amplitude between each reference point.

Images