Method, device, system and medium for continuous human identity recognition based on UWB biological radar detection of heart micro-motion

Abstract

Embodiments of the present application disclose a continuous human body identity recognition method, device and system based on UWB radar detection of heart micro-movement and a medium. The method comprises: generating a heart movement signal of a measured target according to an echo signal of an ultra-wideband (UWB) radar; analyzing the heart movement signal through double wavelet joint positioning to obtain a heartbeat signal of the measured target; extracting feature data from the heartbeat signal according to a two-dimensional principal component analysis (2D-PCA); and identifying the identity of the measured target through a classification strategy.

Description

Technical Field

[0001] The embodiments of the present invention relate to security authentication technology, and in particular to a continuous human identification method, device, system and medium based on UWB bio-radar detection of cardiac micro-movements. Background Technology

[0002] Identity verification technology has become a key line of defense for information security needs such as data security, identity security, and information privacy protection, and is widely used in various fields such as access control and electronic finance. Among the mainstream identity verification methods, biometric-based identity verification technology is increasingly being used in scenarios requiring identity verification and authentication due to its high security and convenience. Related biometric-based identity verification solutions include fingerprint recognition, facial recognition, iris recognition, and voiceprint recognition. However, these solutions cannot continuously or uninterruptedly identify and verify user identity over time.

[0003] Based on this, various continuous authentication solutions have emerged, such as continuous authentication mechanisms based on user biological behavioral characteristics (e.g., key kinetics, gaze patterns, etc.). While these mechanisms achieve continuous authentication to a certain extent, they still have several limitations: First, they require active user cooperation. Second, the authentication process requires direct or close contact between the user and the system (typing on a keyboard, pressing against the iris, or touching the fingerprint sensor), restricting user freedom. Finally, they are susceptible to environmental influences; for example, optical solutions are easily affected by low light, darkness, or clothing obstruction, while fingerprints are easily disrupted by sweat and foreign objects. Summary of the Invention

[0004] The present invention aims to provide a continuous human identification method, device, system, and medium based on UWB radar detection of cardiac micro-movements; which can improve the accuracy and stability of identification.

[0005] The technical solution of this invention is implemented as follows:

[0006] In a first aspect, embodiments of the present invention provide a continuous human identification method based on UWB radar detection of cardiac micro-movements, the method comprising:

[0007] The heart motion signal of the target is generated based on the echo signal of the ultra-wideband (UWB) radar; wherein the echo signal includes the echo signal of at least two range units for the heart of the target.

[0008] The cardiac motion signal is analyzed by dual wavelet joint localization to obtain the heartbeat signal of the target being tested;

[0009] Feature data were extracted from the heartbeat signal using two-dimensional principal component analysis (2D-PCA).

[0010] The feature data is used to identify the identity of the target being tested through a classification strategy.

[0011] Secondly, embodiments of the present invention provide a device for continuous human identification based on UWB radar detection of cardiac micro-movements, the device comprising: a generation part, a positioning part, an extraction part, and an identification part; wherein,

[0012] The generation section is configured to generate a cardiac motion signal of the target based on the echo signal of the UWB radar; wherein the echo signal includes an echo signal of a multi-range unit targeting the heart of the target.

[0013] The positioning section is configured to analyze the cardiac motion signal through dual wavelet joint positioning to obtain the heartbeat signal of the target being measured.

[0014] The extraction portion is configured to extract feature data from the heartbeat signal based on 2D-PCA;

[0015] The identification component is configured to identify the identity of the target being tested by using feature data through a classification strategy.

[0016] Thirdly, embodiments of the present invention provide a continuous human identification system based on UWB radar detection of cardiac micro-movements, the system comprising: UWB radar and computing device;

[0017] The UWB radar is used to send UWB radar waves to the heart location of the target being measured; and to receive the echo signal reflected by the UWB radar waves through the heart location of the target being measured.

[0018] The computing device includes a processor and a memory; the processor is used to execute instructions stored in the memory to implement the continuous human identification method based on UWB radar detection of cardiac micro-movements as described in the first aspect.

[0019] Fourthly, embodiments of the present invention provide a computer storage medium storing at least one instruction, which is executed by a processor to implement the continuous human identification method based on UWB radar detection of cardiac micro-movements as described in the first aspect.

[0020] This invention provides a continuous human identification method, device, system, and medium based on UWB radar detection of cardiac micromovements. It acquires multi-point echo signals from a UWB radar with range resolution to preserve the three-dimensional information of the heart, extracts heartbeat signals from the echo signals using combined wavelet processing to improve the accuracy of heartbeat signal segmentation, and extracts feature data from the heartbeat signals using 2D-PCA to preserve structural information in the cardiac motion data, thereby improving the accuracy and stability of identification. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of a system for identifying human identity provided in an embodiment of the present invention;

[0022] Figure 2 This is a schematic diagram of the echo signal provided in an embodiment of the present invention;

[0023] Figure 3 A schematic flowchart of a continuous human identification method based on UWB radar detection of cardiac micro-movements provided in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram illustrating the process of generating cardiac motion signals provided in an embodiment of the present invention;

[0025] Figure 5 This is a schematic diagram of the cardiac motion signal localization reference point provided in an embodiment of the present invention;

[0026] Figure 6 This is a schematic diagram illustrating the screening of heartbeat signals according to an embodiment of the present invention;

[0027] Figure 7 A schematic diagram of a device for identifying human identity provided in an embodiment of the present invention;

[0028] Figure 8 A schematic diagram of another device for identifying human identity provided in an embodiment of the present invention;

[0029] Figure 9 This is a schematic diagram of the structure of a computing device provided in an embodiment of the present invention. Detailed Implementation

[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0031] Regarding user biological behavioral characteristics, each person's physical attributes (anatomical structure, component size, spatial location, etc.) and physiological attributes (cardiac pattern, blood circulation pattern) are different, causing the heart's volume, surface shape, and 3D deformation during motion to exhibit specific and unique changes at different stages of the cardiac cycle. Therefore, radio frequency electromagnetic signals can be used to obtain continuous and unique cardiac motion-related characteristics of the human body in a non-contact, active, and penetrating manner, thereby achieving human identification.

[0032] In schemes utilizing cardiac motion characteristics for identity recognition, continuous wave (CW) radar is typically used to acquire radar echo signals from the human heart. Based on the heartbeat data or five morphological features of the radar echo signals, preliminary identity recognition has been achieved for a small number of individual samples. However, because CW radar lacks range resolution, the subtle movements of the chest wall caused by cardiac motion must be treated as a single point target, resulting in only one-dimensional echo signals. Furthermore, since the heart is a three-dimensional and continuously beating organ, acquiring only single-point echo information leads to limited information acquisition and susceptibility to interference, severely impacting the accuracy and stability of identity recognition.

[0033] To address the aforementioned deficiencies, see [link to relevant documentation]. Figure 1 The illustration shows a continuous human identification system 10 based on UWB radar detection of cardiac micro-movements provided by an embodiment of the present invention. The system may include: UWB radar 11 and computing device 12.

[0034] To avoid the shortcomings of the single-point echo information mentioned above, in this embodiment of the invention, the heart is regarded as a three-dimensional and continuously beating organ, and multi-point echo signals of the heart's motion are collected by a UWB radar with range resolution.

[0035] For example, Figure 1 The transmitting antenna of the UWB radar 11 shown transmits pulse signals at a preset pulse repetition frequency with a set time interval. After the pulse signals reach the target (such as the human heart) and are reflected by the target, the receiving antenna of the UWB radar 11 receives electromagnetic reflection echoes from different distance ranges on the surface of the human heart.

[0036] s i (t)=d i +d(t)=d i +d ri sin(2πf r t)+d hi sin(2πf h t)

[0037] Where, d id is the fixed distance between the UWB radar antenna 11 and the i-th range element of the curved surface of the human chest wall. r The displacement amplitude caused by breathing within this distance unit, d h f is the displacement amplitude caused by the heartbeat within this distance unit. r f represents respiratory rate h It represents heart rate.

[0038] If the normalized transmitted pulse is represented by δ(t), then the total impulse response is:

[0039] r(t, τ) = ∑ i＝1 A i δ(τ-τ i (t))+∑ k＝1 A k δ(τ-τ k )

[0040] Where t is the observation time and τ is the propagation time. A i δ(τ-τ i (t) represents the propagation time τ corresponding to the i-th distance unit. i (t) and amplitude A i The response of slight chest wall movements. ∑ k＝1 A k δ(τ-τ k ) represents the response of each static target, with a propagation time of τ. k Amplitude A k . τ i (t) represents the transmit-receive pulse time interval, derived from s i (t) is determined by the intrinsic propagation time τ. i Add the sum of the two sinusoidal delays related to respiratory and cardiac displacement:

[0041]

[0042] Where c is the speed of light, approximately 3 × 10⁻⁶. 8 m / s, τ i =2d i / c, τ ri =2d ri / c, τ h =2d hi / c.

[0043] In slow time t=nT s The N discrete-time sequences (n = 1, 2...N) obtained at discrete times are stored in matrix R, resulting in the echo matrix R (e.g., ...). Figure 2 (As shown) is represented as:

[0044] R[m,n]=r(τ=mT) f, t=nT s )

[0045] Where m and n represent the number of samples in the fast and slow times, respectively. T s It is the duration of the slow-time pulse, T f It is a fast sampling interval.

[0046] The echo matrix R can be understood as the two-dimensional radar detection echo of time-distance change of micro-motion signal region of the heart surface within the effective range. The row vector records the time-distance change of micro-motion of any distance unit on the surface of the heart as time changes, while the column vector records the micro-motion change of different distance units at any time.

[0047] Using the aforementioned echo matrix R as the echo signal of the UWB radar, the method for identifying human identity provided in this embodiment of the invention is executed by the computing device 12. In some examples, the computing device 12 can be at least one of devices such as a smartphone, smartwatch, desktop computer, laptop, virtual reality terminal, augmented reality terminal, wireless terminal, and laptop computer. The computing device 12 has communication capabilities and can access wired or wireless networks. The computing device 12 can refer to one of multiple terminals, and those skilled in the art will understand that the number of such terminals can be more or less. In some examples, the computing device 12 can receive the echo signal of the UWB radar based on the accessed wired or wireless network, or it can receive the echo signal of the UWB radar based on the input of the operator (or user). It is understood that the computing device 12 undertakes the calculation and processing work of the technical solution of the present invention, and this disclosure does not limit this aspect.

[0048] See Figure 3 This invention illustrates a continuous human identification method based on UWB radar detection of cardiac micro-movements, provided by an embodiment of the present invention. The method includes:

[0049] S301: Generates the heart movement signal of the target based on the echo signal of the UWB radar;

[0050] The echo signal includes echo signals from at least two distance units to the target heart.

[0051] In this embodiment of the invention, the echo signal acquired by the UWB radar includes echo signals from multiple range units within the effective human detection range. For each range unit's echo signal, there is the heartbeat itself and the micro-movements within the chest cavity caused by the heartbeat. Furthermore, the echo signal from each range unit is also mixed with static clutter and background noise from the entire range space environment, chest wall micro-movements caused by the heartbeat and accompanying lung respiration, respiratory harmonics, and human motion signal waves. Therefore, the echo signal can be preprocessed to eliminate the aforementioned static clutter and background noise, chest wall micro-movements, respiratory harmonics, and human motion signal waves before generating an effective cardiac motion signal.

[0052] S302: The cardiac motion signal is analyzed by dual wavelet joint positioning to obtain the heartbeat signal of the target being measured;

[0053] In this embodiment of the invention, since the echo signal includes echo signals from multiple distance units, the cardiac motion signal also includes cardiac motion signals from multiple distance units. For the cardiac motion signal of each distance unit, a heartbeat signal segmentation scheme based on reference points can be adopted, that is, after accurately locating the reference points of the peaks and troughs, heartbeat segmentation is performed. For the cardiac motion signal of any distance unit, this embodiment of the invention uses a dual-wavelet joint positioning method combining Daubechies (dbN) wavelet and Mexican Hat (mexh) wavelet to locate the peak and trough reference points. This dual-wavelet joint positioning method comprehensively utilizes the periodic sensitivity of the mexh wavelet and the detail sensitivity of the db1 wavelet, improving the accuracy of reference point positioning, thereby contributing to the precise segmentation of the heartbeat signal.

[0054] S303: Extract feature data from the heartbeat signal using 2D-PCA;

[0055] In this embodiment of the invention, since the heartbeat signal obtained in the aforementioned steps retains the three-dimensional structural information of the heart, feature data is extracted from the heartbeat signal by 2D-PCA. Compared with one-dimensional PCA, it can more completely retain the structural information in the cardiac motion data, thereby improving the accuracy of identity recognition.

[0056] S304: Identify the identity of the target being tested by using the feature data through a classification strategy.

[0057] In this embodiment of the invention, a feature database of a target individual can be constructed using the feature data extracted by 2D-PCA, and a machine learning classifier neural network model can be trained using the feature data in the database, thereby using the trained classifier neural network model to identify the identity of the target being tested.

[0058] for Figure 3The technical solution shown in this embodiment of the invention treats the heart as a three-dimensional and continuously beating organ. It collects multi-point echo signals from the heart using a UWB radar with distance resolution to preserve the three-dimensional information of the heart. It obtains the heartbeat signal from the echo signal through joint wavelet processing to improve the accurate segmentation of the heartbeat signal. It extracts feature data from the heartbeat signal through 2D-PCA to preserve the structural information in the heart motion data, thereby improving the accuracy and stability of identity recognition.

[0059] for Figure 3 In some examples of the technical solutions shown, the step of generating the cardiac motion signal of the target based on the echo signal of the UWB radar includes:

[0060] Receive the echo signal reflected from the heart location of the target by the transmitted wave of the UWB radar;

[0061] Eliminate the background noise component of the echo signal to obtain the noise-reduced echo signal;

[0062] Obtain the signal energy value corresponding to the echo signal of each distance cell in the denoised echo signal;

[0063] Among the signal energy values ​​corresponding to the echo signals of all range cells, the echo signal of the range cell with the largest signal energy value is determined as the echo signal of the optimal range cell.

[0064] By eliminating the respiratory signal components from the echo signal of the optimal distance unit and the echo signal of the distance unit adjacent to the optimal distance unit, the cardiac motion signal of the target under test is obtained.

[0065] For the example above, combined Figure 4 Specifically Figure 4 (a) shows the echo signal of the transmitted wave received by the UWB radar reflected from the heart location of the target under test.

[0066] The detailed preprocessing procedure for this echo signal may include:

[0067] First, in a static environment, the generated clutter can be considered as a DC component in the slow time direction, and the background clutter is independent of the slow time t. Therefore, background clutter can be removed by filtering the echo signal. For example, this can be achieved by de-averaging the original signal; specifically, by subtracting the average value of all received waveforms from the original echo signal R, i.e., by subtracting the average value of all rows in R from each row to obtain a new matrix X. Subsequently, DC filtering is performed to remove baseline drift in the echo signal. Finally, a 31st-order Hamming window low-pass filter with a cutoff frequency of 5Hz is constructed to remove high-frequency components from the original signal. The post-processed signal of the denoised echo signal obtained after the above processing is as follows: Figure 4 As shown in (b), the background noise and static clutter have been largely removed, and a clear micro-motion response echo can be seen within the range of the measured target.

[0068] Next, in the echo signal, the micro-motion information of the cardiac chest wall region is only distributed within multiple distance units over a certain distance range. For the denoised echo signal, the signal S of each distance channel is calculated along the distance direction. i The energy value of (t) is used to obtain the total energy matrix E = [E1, E2, ..., E] under different distance units. i ,...E M Due to the signal energy E of any distance unit i The signal amplitude is proportional to the corresponding distance interval. By finding the maximum energy value max(E), the echo signal corresponding to the optimal distance unit within the effective chest wall micromotion distance range can be located as S. B (t). After the optimal distance unit is located, it is used as a reference to extend along both sides of the distance axis, and the distance units adjacent to the optimal distance unit are located by using the energy threshold. The optimal distance unit and the distance units adjacent to the optimal distance unit obtained above form the effective distance range of cardiac micromotion information. For example, the echo signal of M′ distance units (such as...) Figure 4 (c) shown). In this embodiment of the invention, the effective distance range S E =[S B-1 (t), S B (t), S B+1 (t), S B+2 [(t)], that is, M′=4.

[0069] Finally, since normal human breathing inevitably affects cardiac motion signals, it is necessary to filter the echo signals to remove interference from breathing and extract the desired heartbeat signals. The normal human respiratory rate is 0.27–0.33 Hz, while the heartbeat frequency ranges above 1 Hz. For the echo signal of any distance unit within the aforementioned effective distance range, this embodiment of the invention uses a bandpass filter with a cutoff frequency of 1 to 5 Hz to eliminate the breathing signal component, thereby extracting the cardiac motion signals corresponding to multiple distance units within the effective distance range of the target. For example… Figure 4 As shown in (d), compared with the mixed heartbeat and respiratory signal before the respiratory signal component was eliminated, the cardiac motion signal after the respiratory signal component was eliminated showed a clear heartbeat rhythm, and the amplitude was much smaller than that of the mixed signal. This also verifies that the larger respiratory component has been effectively filtered out.

[0070] for Figure 3 In some examples of the technical solutions shown, the analysis of the cardiac motion signal using dual wavelet joint localization to obtain the heartbeat signal of the target being measured includes:

[0071] The cardiac motion signal was divided by combining the Dobesy dbN wavelet and the Mexican mexh wavelet to obtain the heartbeat data of the target being measured.

[0072] Abnormal heartbeat data is removed from the heartbeat data according to the heartbeat template to obtain the heartbeat signal of the target being tested.

[0073] In this embodiment of the invention, for the cardiac motion signal corresponding to each distance unit within the effective distance range, it is necessary to perform heartbeat segmentation and effective heartbeat screening to obtain the heartbeat signal of the target being tested.

[0074] For beat segmentation, this embodiment of the invention employs a benchmark-based beat segmentation method, with peaks and troughs being suitable benchmark point types. A beat length is defined as the data length of one peak and two troughs. After accurately locating the peak and trough benchmarks, multiple beat data points can be obtained by segmenting according to the aforementioned beat length.

[0075] Specifically, in this embodiment of the invention, wavelet transform is preferably used to locate the peaks and troughs of the cardiac motion signal. However, since the heartbeat signal is a time-varying, non-stationary signal, using a single mother wavelet for wavelet transform to locate the reference points is inaccurate. For example, for the cardiac motion signal corresponding to any distance unit within the effective distance range, such as... Figure 5As shown in (a), while the method of using Mexh wavelets to locate peaks and troughs has good localization effects in both the time and frequency domains, it is prone to missing detections of periodically fluctuating cardiac motion signals because Mexh wavelets place greater emphasis on the periodicity of the signal. Figure 5 (b) shows the missed data points highlighted in the circle. For example, while using a first-order dbN wavelet (db1) to locate the peak and trough reference points for the cardiac motion signal corresponding to any distance unit within the effective range is more sensitive to details, it can also easily lead to over-detection. Figure 5 As shown in (c), although all peaks and troughs were detected, the following also appeared: Figure 5 (c) shows the multi-detection data points indicated by the circled boxes.

[0076] Based on the shortcomings of the single mother wavelet positioning reference point, this embodiment of the invention utilizes the characteristics of the mexh wavelet's sensitivity to periodicity and the db1 wavelet's sensitivity to detail, combining the two wavelet transform methods to comprehensively utilize the periodic sensitivity of the mexh wavelet and the detail sensitivity of the db1 wavelet.

[0077] Based on the above description, in some examples, the analysis of the cardiac motion signal by combining the Dobessie dbN wavelet and the Mexican mexh wavelet to obtain the heartbeat data of the target being measured includes:

[0078] The cardiac motion signal was located using the dbN wavelet to obtain the first location data;

[0079] The cardiac motion signal was located using the mexh wavelet to obtain second location data;

[0080] The peak and trough data in the first positioning data are superimposed with the peak and trough data in the second positioning data to obtain the superimposed data;

[0081] After removing duplicate data points and discarding multiple detection data from the superimposed data, multiple heartbeat data are obtained by dividing the data according to the set heartbeat length.

[0082] In the example above, the superimposed data simultaneously reflects the location results of the peak and trough reference points by both the db1 wavelet and the mexh wavelet. To address this, duplicate data points can be removed. For example, for the i-th maximum point (a potential peak data point) in the superimposed data, the maximum value within the range of 8 sampling points before and after the i-th maximum point can be used as the peak data point. After locating this peak data point, if the lateral distance between the two peaks before and after this peak data point is less than 6 sampling points, it indicates a duplicate data point that needs to be removed. Therefore, the smaller of the two peaks before and after this peak data point is removed. Similarly, for the j-th minimum point (a potential trough data point) in the superimposed data, the minimum value within the range of 8 sampling points before and after this j-th minimum point can be used as the trough data point. After locating this trough data point, if the lateral distance between the two troughs before and after this trough data point is less than 6 sampling points, the larger of the two troughs before and after this trough data point is removed.

[0083] The peaks and troughs of the cardiac motion signal for each distance unit were accurately located (e.g., Figure 5 (d) After that, based on the previously set heartbeat length, a reliable heartbeat can be obtained by dividing the trough-trough data segment into segments with only one peak data point between consecutive trough data points and a length ≤ 1.5 times the average heartbeat length. That is, after determining the peak data point and the trough data point, if there are still extra sampling points between adjacent peak points and trough points, then these extra sampling points are removed, thereby completing the process of removing extra data.

[0084] Based on the above division method, the cardiac motion signal of each distance unit can be divided into multiple reliable heartbeat data, such as... Figure 6 (a) shows the heart rate data.

[0085] For effective heartbeat screening, optionally, the step of removing abnormal heartbeat data from the heartbeat data according to the heartbeat template to obtain the heartbeat signal of the target being tested includes:

[0086] Use the average value of all heartbeat data as the heartbeat template;

[0087] For each heartbeat data point, obtain the Pearson correlation coefficient and Euclidean distance between it and the heartbeat template;

[0088] The Pearson correlation coefficient and Euclidean distance are compared with their corresponding significance level thresholds. Abnormal heartbeat data exceeding the significance level thresholds are removed to obtain valid heartbeat signals. Figure 6 (b) shows the heart rate data after removal.

[0089] for Figure 3In some examples of the technical solutions shown, the extraction of feature data from the heartbeat signal based on 2D-PCA includes:

[0090] The heartbeat signals of the optimal distance unit and the distance units adjacent to the optimal distance unit are used to form the heartbeat matrix of the target being measured. Where M′ represents the number of distance units, n′ represents the heartbeat signal index of each distance unit, N′ represents the number of heartbeat signals in each distance unit, and L=50 represents the heartbeat sampling points;

[0091] According to the heartbeat matrix The covariance matrix is ​​obtained from the heartbeat signal matrix of all distance units in the matrix.

[0092] In the eigenvectors and eigenvalues ​​of the covariance matrix, the eigenvectors corresponding to the first k largest eigenvalues ​​are used to form a projection matrix;

[0093] The principal component vector is obtained from the heartbeat matrix based on the projection matrix;

[0094] A feature matrix is ​​generated based on the principal component vectors; wherein the feature matrix is ​​used to represent the feature data of the heart motion of the target being measured.

[0095] For the above example, it should be noted that for the multi-channel signal S within the effective distance range E After heartbeat segmentation and filtering, each heartbeat matrix can be represented as: Where M′ = 4 represents 4 effective distance units, n′ is the effective heartbeat sequence index, and L = 50 represents the heartbeat sampling points. Each heartbeat matrix (of size M′ × L) can be viewed as the projection of the three-dimensional heart motion onto the distance direction at the current heartbeat moment. Multiple distance units are interconnected, implicitly containing certain structural relationships within the heart. Based on this, this embodiment of the invention uses 2D-PCA to extract feature data, effectively preserving the structural information of the heart motion data while extracting the intrinsic features of each heartbeat matrix, and also achieving dimensionality reduction.

[0096] for Figure 3 In some examples of the technical solutions shown, the step of identifying the identity of the target being tested using the feature data through a classification strategy includes:

[0097] The feature data is input into a trained neural network model, and the identity of the target under test is identified based on the output of the neural network model.

[0098] Regarding the above example, it should be noted that after extracting feature data using 2D-PCA, further construction of an individual identity feature database can be achieved by combining a neural network model of a machine learning classifier for training and identity recognition. This neural network model can include a quadratic SVM model or a three-layer neural network model.

[0099] Based on the foregoing technical solution, embodiments of the present invention verify the technical effects of the above technical solution through experiments, under the following experimental conditions:

[0100] First, the UWB radar parameters are shown in Table 1:

[0101] System parameters symbol Parameter value / unit Detection range R 0.4-5.0m frequency band △B 7.25-10.20GHz Baseband sampling rate f B ]]> 2.916GHz Sampling frequency f s ]]> 17Hz Average output power P ≤44dBm / MHz

[0102] Table 1

[0103] Secondly, to simulate radar detection of cardiac micromovements under ideal conditions, subjects lay flat on the ground facing the radar, with laser positioning used to keep the radar directly above the heart. The distance between the radar and the heart was adjusted to approximately 55 cm. A total of 18 subjects (aged 19 to 30, 9 males and 9 females) participated in the experiment, and data was collected from each subject according to standardized instructions. On the other hand, to compare the impact of respiratory behavior on cardiac micromovement identification, data was collected from each subject under two basic states: normal breathing and breath-holding. During the experiment, 12 sets of data were collected from each subject. Each set of data was collected for 25 seconds under normal breathing and 20 seconds under breath-holding. Sufficient time was allocated before and after each set of data collection to allow subjects to calm down.

[0104] Next, comparative schemes were designed. Comparative Scheme 1 involves obtaining the cardiac motion signal from the optimal distance unit and the heartbeat signal according to S302, then inputting these signals into a neural network model for training, and finally performing identity recognition based on the trained neural network model. Comparative Scheme 2 involves obtaining the cardiac motion signals from the optimal distance unit and its neighboring distance units, obtaining the heartbeat signal according to S302, then inputting these signals into a neural network model for training, and finally performing identity recognition based on the trained neural network model. The comparison metrics are accuracy and F1 score. The identity recognition results based on the quadratic SVM model of the technical solution proposed in this embodiment of the invention (represented as 2D-PCA in the experiment), Comparative Scheme 1, and Comparative Scheme 2 are shown in Table 2. As shown in Table 2, under normal breathing conditions, the identification accuracy rate based on Comparison Scheme 1 was 81.01%, the identification accuracy rate based on Comparison Scheme 2 was 56.81%, while the identification accuracy rate of 2D-PCA was 89.21%, representing relative improvements of 8.2% and 32.4%, respectively. Meanwhile, under breath-holding conditions, the identification accuracy rate of 2D-PCA was 90.25%, representing improvements of 5.95% and 30.55% compared to Comparison Scheme 1 and Comparison Scheme 2, respectively.

[0105]

[0106] Table 2

[0107] Table 3 shows the identity recognition results of 2D-PCA, Comparison Scheme 1, and Comparison Scheme 2 based on a three-layer neural network. As can be seen from Table 3, the identity recognition accuracy of 2D-PCA under normal breathing conditions is 84.21% (an improvement of 5.48% and 24.86% compared to Comparison Scheme 1 and Comparison Scheme 2, respectively), and 85.75% under breath-holding conditions (an improvement of 5.01% and 21.86% compared to Comparison Scheme 1 and Comparison Scheme 2, respectively).

[0108]

[0109] Table 3

[0110] The results of identity recognition based on the KNN model for 2D-PCA, Comparison Scheme 1, and Comparison Scheme 2 are shown in Table 4. Table 4 shows that the identity recognition accuracy of 2D-PCA under normal breathing conditions is 73.42% (an improvement of 0.66% and 14.31% compared to Comparison Scheme 1 and Comparison Scheme 2, respectively), and 80.25% under breath-holding conditions (an improvement of 1.75% and 13.68% compared to Comparison Scheme 1 and Comparison Scheme 2, respectively).

[0111]

[0112] Table 4

[0113] Summarizing Tables 2 to 4, 2D-PCA demonstrates the best recognition performance. When using two classifiers, quadratic SVM and cubic neural network, the overall average recognition accuracy is improved by 6.16% (6.84% for normal breathing and 5.48% for breath-holding) compared to the first comparison scheme, and by 27.42% (28.63% for normal breathing and 26.21% for breath-holding) compared to the second comparison scheme. These results demonstrate that using 2D-PCA to extract features from the distance channel-sampling point two-dimensional matrix exhibits higher individual specificity and stability.

[0114] Secondly, compared to the first scheme, the second scheme has an absolute advantage in the amount of feature data, with nearly four times the number of features achieved through multi-channel expansion of the effective distance range (M′=4). However, the actual recognition performance deteriorated sharply (the average recognition accuracy of the three classifiers decreased by 19.08% and 17.80% in normal breathing and breath-holding states, respectively). This may be because, although the amount of feature data increased after multi-channel expansion, there were still significant differences between adjacent multi-channel heartbeat data. The increased data variability, while expanding the data, had a dominant negative impact on identity recognition, severely affecting the classifier's learning effect on individual features.

[0115] Based on the same inventive concept as the aforementioned technical solution, see [link to inventive concept]. Figure 7 This illustration shows a continuous human identification device 70 based on UWB radar detection of cardiac micro-movements according to an embodiment of the present invention. The device 70 may include: a generation section 710, a positioning section 720, an extraction section 730, and an identification section 740; wherein,

[0116] The generation section 710 is configured to generate a cardiac motion signal of the target under test based on the echo signal of the ultra-wideband UWB radar; wherein the echo signal includes an echo signal of a multi-range unit for the heart of the target under test.

[0117] The positioning section 720 is configured to analyze the cardiac motion signal through dual wavelet joint positioning to obtain the heartbeat signal of the target being measured.

[0118] The extraction portion 730 is configured to extract feature data from the heartbeat signal according to the two-dimensional principal component analysis (2D-PCA) method.

[0119] The identification section 740 is configured to identify the identity of the target being tested by using the feature data through a classification strategy.

[0120] In some examples, the generation section 710 is configured to:

[0121] Receive the echo signal reflected from the heart location of the target by the transmitted wave of the UWB radar;

[0122] Eliminate the background noise component of the echo signal to obtain the noise-reduced echo signal;

[0123] Obtain the signal energy value corresponding to the echo signal of each distance unit in the denoised echo signal;

[0124] Among the signal energy values ​​corresponding to the echo signals of all range cells, the echo signal of the range cell with the largest signal energy value is determined as the echo signal of the optimal range cell.

[0125] By eliminating the respiratory signal components from the echo signal of the optimal distance unit and the echo signal of the distance unit adjacent to the optimal distance unit, the cardiac motion signal of the target under test is obtained.

[0126] In some examples, such as Figure 8 As shown, the positioning part 720 includes a joint wavelet partitioning module 721 and a filtering module 722: wherein,

[0127] The joint wavelet segmentation module 721 is configured to segment the cardiac motion signal by combining the Dobesie dbN wavelet and the Mexican mexh wavelet to obtain the heartbeat data of the target being measured.

[0128] The filtering module 722 is configured to remove abnormal heartbeat data from the heartbeat data according to the heartbeat template, and obtain the heartbeat signal of the target being tested.

[0129] In some examples, the joint wavelet partitioning module 721 is configured to

[0130] The cardiac motion signal was located using the dbN wavelet to obtain the first location data;

[0131] The cardiac motion signal was located using the mexh wavelet to obtain second location data;

[0132] The peak and trough data in the first positioning data are superimposed with the peak and trough data in the second positioning data to obtain the superimposed data;

[0133] After removing duplicate data points and discarding multiple detection data from the superimposed data, multiple heartbeat signals are obtained by dividing the data according to the heartbeat length.

[0134] In some examples, the filtering module 722 is configured to:

[0135] Use the average value of all heartbeat data as the heartbeat template;

[0136] For each heartbeat data point, obtain the Pearson correlation coefficient and Euclidean distance between it and the heartbeat template;

[0137] The Pearson correlation coefficient and Euclidean distance are compared with their corresponding significance level thresholds, and abnormal heartbeat data exceeding the significance level thresholds are removed to obtain the valid heartbeat signal of the target being tested.

[0138] In some examples, the extraction portion 730 is configured to:

[0139] The heartbeat signals of the optimal distance unit and the distance units adjacent to the optimal distance unit are used to form the heartbeat matrix of the target being measured. Where M′ represents the number of distance units, n′ represents the heartbeat signal index of each distance unit, N′ represents the number of heartbeat signals in each distance unit, and L=50 represents the heartbeat sampling points;

[0140] According to the heartbeat matrix The covariance matrix is ​​obtained from the heartbeat signal matrix of all distance units in the matrix.

[0141] In the eigenvectors and eigenvalues ​​of the covariance matrix, the eigenvectors corresponding to the first k largest eigenvalues ​​are used to form a projection matrix;

[0142] The principal component vector is obtained from the heartbeat matrix based on the projection matrix;

[0143] A feature matrix is ​​generated based on the principal component vectors; wherein the feature matrix is ​​used to represent the feature data of the heart motion of the target being measured.

[0144] In some examples, the identification portion 740 is configured to:

[0145] The feature data is input into a trained neural network model, and the identity of the target under test is identified based on the output of the neural network model.

[0146] See Figure 9 This diagram illustrates a structural block diagram of a computing device 12 provided in an exemplary embodiment of this application. The computing device 12 may include one or more components such as a processor 910 and a memory 920.

[0147] Optionally, the processor 910 connects various parts within the computing device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 920, and by calling data stored in the memory 920. Optionally, the processor 910 can be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 910 can integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), Neural-network Processing Unit (NPU), and baseband chip. Specifically, the CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display on the touch screen; the NPU is used to implement Artificial Intelligence (AI) functions; and the baseband chip is used for wireless communication. It is understandable that the aforementioned baseband chip may not be integrated into the processor 910, but may be implemented using a separate chip.

[0148] The memory 920 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 920 may include a non-transitory computer-readable storage medium. The memory 920 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 920 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data created according to the use of the computing device, etc.

[0149] In addition, those skilled in the art will understand that the structure of the computing device shown in the above figures does not constitute a limitation on the computing device. The computing device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the computing device may also include a display screen, camera assembly, microphone, speaker, radio frequency circuit, input unit, sensors (such as accelerometer, angular velocity sensor, light sensor, etc.), audio circuit, WiFi module, power supply, Bluetooth module, etc., which will not be described in detail here.

[0150] This application also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor to implement the continuous human identification method based on UWB radar detection of cardiac micro-movements as described in the above embodiments.

[0151] This application also provides a computer program product, which includes computer instructions stored in a computer-readable storage medium; a processor of a computing device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computing device to perform the continuous human identification method based on UWB radar detection of cardiac micro-movements described in the above embodiments.

[0152] Those skilled in the art will recognize that the functions described in the embodiments of this application in one or more of the above examples can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of a computer program from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0153] It should be noted that the technical solutions described in the embodiments of the present invention can be combined arbitrarily without conflict.

[0154] 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 variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for continuous human identification based on UWB radar detection of cardiac micro-movements, characterized in that, The method includes: The heart motion signal of the target is generated based on the echo signal of the ultra-wideband (UWB) radar; wherein the echo signal includes echo signals of at least two range units for the heart of the target. The cardiac motion signal is analyzed by dual wavelet joint localization to obtain the heartbeat signal of the target being tested; Feature data were extracted from the heartbeat signal using two-dimensional principal component analysis (2D-PCA). The feature data is used to identify the identity of the target being tested through a classification strategy; The step of analyzing the cardiac motion signal using dual wavelet joint localization to obtain the heartbeat signal of the target being measured includes: The cardiac motion signal was divided by combining the Dobesy dbN wavelet and the Mexican mexh wavelet to obtain the heartbeat data of the target being measured. Abnormal heartbeat data are removed from the heartbeat data according to the heartbeat template to obtain the heartbeat signal of the target being tested; The analysis of the cardiac motion signal by combining the Dobesied dbN wavelet and the Mexican mexh wavelet to obtain the heartbeat signal of the target being measured includes: The cardiac motion signal was located using the dbN wavelet to obtain the first location data; The cardiac motion signal was located using the mexh wavelet to obtain second location data; The peak and trough data in the first positioning data are superimposed with the peak and trough data in the second positioning data to obtain the superimposed data; After removing duplicate data points and discarding multiple detection data from the superimposed data, multiple heartbeat signals are obtained by dividing the data according to the heartbeat length.

2. The method according to claim 1, characterized in that, The process of generating the cardiac motion signal of the target based on the echo signal from the ultra-wideband (UWB) radar includes: Receive the echo signal reflected from the heart location of the target by the transmitted wave of the UWB radar; Eliminate the background noise component of the echo signal to obtain the noise-reduced echo signal; Obtain the signal energy value corresponding to the echo signal of each distance unit in the denoised echo signal; Among the signal energy values ​​corresponding to the echo signals of all range cells, the echo signal of the range cell with the largest signal energy value is determined as the echo signal of the optimal range cell. By eliminating the respiratory signal components from the echo signal of the optimal distance unit and the echo signal of the distance unit adjacent to the optimal distance unit, the cardiac motion signal of the target under test is obtained.

3. The method according to claim 1, characterized in that, The step of removing abnormal heartbeat data from the heartbeat data according to the heartbeat template to obtain the heartbeat signal of the target being tested includes: Use the average value of all heartbeat data as the heartbeat template; For each heartbeat data point, obtain the Pearson correlation coefficient and Euclidean distance between it and the heartbeat template; The Pearson correlation coefficient and Euclidean distance are compared with their corresponding significance level thresholds, and abnormal heartbeat data exceeding the significance level thresholds are removed to obtain the valid heartbeat signal of the target being tested.

4. The method according to claim 1, characterized in that, The extraction of feature data from the heartbeat signal using two-dimensional principal component analysis (2D-PCA) includes: The heartbeat signals of the optimal distance unit and the distance units adjacent to the optimal distance unit are used to form the heartbeat matrix of the target being measured. Where M′ represents the number of distance units, n′ represents the heartbeat signal index of each distance unit, N′ represents the number of heartbeat signals in each distance unit, and L=50 represents the heartbeat sampling points; According to the heartbeat matrix The covariance matrix is ​​obtained from the heartbeat signal matrix of all distance units in the matrix. In the eigenvectors and eigenvalues ​​of the covariance matrix, the eigenvectors corresponding to the first k largest eigenvalues ​​are used to form a projection matrix; The principal component vector is obtained from the heartbeat matrix based on the projection matrix; A feature matrix is ​​generated based on the principal component vectors; wherein the feature matrix is ​​used to represent the feature data of the heart motion of the target being measured.

5. The method according to claim 1, characterized in that, The step of identifying the identity of the target being tested using the feature data through a classification strategy includes: The feature data is input into a trained neural network model, and the identity of the target under test is identified based on the output of the neural network model.

6. A device for continuous human identification based on UWB radar detection of cardiac micro-movements, characterized in that, The device includes: a generation section, a positioning section, an extraction section, and an identification section; wherein... The generation section is configured to generate a cardiac motion signal of the target under test based on the echo signal of the ultra-wideband (UWB) radar; wherein the echo signal includes echo signals of a multi-range unit targeting the heart of the target under test. The positioning section is configured to analyze the cardiac motion signal through dual wavelet joint positioning to obtain the heartbeat signal of the target being measured. The extraction portion is configured to extract feature data from the heartbeat signal using two-dimensional principal component analysis (2D-PCA). The identification section is configured to identify the identity of the target being tested by using the feature data through a classification strategy; The positioning component is further configured to include a joint wavelet partitioning module and a filtering module, wherein, The joint wavelet segmentation module is configured to segment the cardiac motion signal by combining the Dobesie dbN wavelet and the Mexican mexh wavelet to obtain the heartbeat data of the target being measured. The filtering module is configured to remove abnormal heartbeat data from the heartbeat data according to the heartbeat template, and obtain the heartbeat signal of the target being tested. The joint wavelet partitioning module is further configured as follows: The cardiac motion signal is located using the dbN wavelet to obtain first location data; the cardiac motion signal is located using the mexh wavelet to obtain second location data; the peak and trough data in the first location data are superimposed with the peak and trough data in the second location data to obtain superimposed data; after removing duplicate data points and discarding multiple detection data in the superimposed data, multiple heartbeat signals are obtained according to the heartbeat length.

7. A system for continuous human identification based on UWB radar detection of cardiac micro-movements, characterized in that, The system includes: UWB radar and computing equipment; The UWB radar is used to send UWB radar waves to the heart location of the target being measured; and to receive the echo signal reflected by the UWB radar waves through the heart location of the target being measured. The computing device includes a processor and a memory; the processor is used to execute instructions stored in the memory to implement the continuous human identification method based on UWB radar detection of cardiac micro-movements as described in any one of claims 1 to 5.

8. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is executed by a processor to implement the continuous human identification method based on UWB radar detection of cardiac micro-movements as described in any one of claims 1 to 5.

Images