CN115236749A

CN115236749A - Living body detection method, apparatus and storage medium

Info

Publication number: CN115236749A
Application number: CN202210802117.6A
Authority: CN
Inventors: 齐庆杰; 刘思昀; 张婧雯; 王月; 孙立峰; 程会锋; 刘英杰; 吴兵; 柴佳美
Original assignee: General Coal Research Institute Co Ltd
Current assignee: General Coal Research Institute Co Ltd
Priority date: 2022-07-08
Filing date: 2022-07-08
Publication date: 2022-10-25

Abstract

The application provides a method, a device and a storage medium for detecting a living body, wherein the method comprises the following steps: acquiring a first radar echo of a region to be detected, constructing a sampling window to traverse the first radar echo, and acquiring a second radar echo in the sampling window when the window is slid each time; acquiring a trained target life body detection model, and acquiring a life body detection result of a second radar echo based on the target life body detection model; and acquiring a target life body detection result of the region to be detected corresponding to the first radar echo according to the respective life body detection results of all the second radar echoes. In the method and the device, the calculated amount of the detection result of the life body of the radar echo is reduced, the detection efficiency of the life body is improved, the detection accuracy of the life body in the region to be detected is improved, and the detection algorithm of the life body is optimized.

Description

Living body detection method, apparatus and storage medium

Technical Field

The present application relates to the field of signal processing technologies, and in particular, to a method and an apparatus for detecting a living body, and a storage medium.

Background

With the development of the technology, people can detect the life information of the trapped people on the rescue site by adopting a life body detection technology, so that the trapped people can be rescued.

In the related technology, the extraction of the vital sign signals of the trapped people in the rescue scene can be realized through a related short-time Fourier transform algorithm, a wavelet transform and a processing algorithm of empirical mode decomposition. However, due to the limitation of the performance of the algorithm and the complexity of the environment of the rescue scene, the vital body detection technology in the related art may be affected by the complex environment of the rescue scene, thereby causing the accuracy of vital sign signal detection of the trapped person to be low.

Disclosure of Invention

The present application is directed to solving, to some extent, one of the technical problems in the related art. Therefore, an object of the present application is to provide a method, an apparatus and a storage medium for detecting a vital sign, so as to solve the problem of low accuracy of detecting a vital sign signal caused by a complex environment of a rescue scene in a vital sign detection technology in the related art to some extent. The technical scheme of the application is as follows:

in a first aspect, the present application provides a method for detecting a living body, including: acquiring a first radar echo of a region to be detected, constructing a sampling window to traverse the first radar echo, and acquiring a second radar echo in the sampling window when the window is slid each time; acquiring a trained target life body detection model, and acquiring a life body detection result of the second radar echo based on the target life body detection model; and acquiring a target life detection result of the area to be detected corresponding to the first radar echo according to respective life detection results of all the second radar echoes.

In addition, the living body detecting method provided by the first aspect of the present application further has the following additional features:

according to an embodiment of the present application, the obtaining, according to respective living body detection results of all second radar echoes, a target living body detection result of the region to be detected corresponding to the first radar echo includes: integrating respective life body detection results of all the second radar echoes to obtain candidate life body detection results of the first radar echoes; and determining and deleting the false recognition result in the candidate life body detection result, and determining the deleted candidate life body detection result as the target life body detection result of the region to be detected corresponding to the first radar echo.

According to an embodiment of the present application, the integrating the living body detection results of all the second radar echoes to obtain the candidate living body detection result of the first radar echo includes: and splicing the life body detection results of all the second radar echoes according to the positions of the second radar echoes in the first radar echoes, and acquiring the candidate life body detection result of the first radar echo according to the spliced life body detection result.

According to an embodiment of the present application, the determining and deleting a false recognition result in the candidate life body detection results, and determining the deleted candidate life body detection result as the target life body detection result of the region to be detected corresponding to the first radar echo includes: clustering the candidate life body detection results to obtain the maximum Euclidean distance and the average Euclidean distance of each cluster; determining the misrecognition result from the candidate life body detection result according to the maximum Euclidean distance and the average Euclidean distance; and deleting the false recognition result in the candidate life body detection result to obtain the target life body detection result.

According to an embodiment of the present application, after obtaining the target living body detection result, the method includes: acquiring a detection state value, a prediction state value and a Kalman gain of a life body in the region to be detected in a single slow time from the detection result of the target life body; determining a target state value of the living body at the single slow time according to the detection state value, the prediction state value and the Kalman gain; and acquiring the moving track of the living body in the region to be detected according to the target state values corresponding to all slow times in the first radar echo.

According to an embodiment of the present application, the acquiring a first radar echo of a region to be detected includes: acquiring an initial radar echo of the area to be detected according to the ultra-wideband radar; eliminating static background wave components and direct current component components in the initial radar echo to obtain a candidate radar echo of the area to be detected after elimination; and performing gain control on the candidate radar echo, and taking the radar echo obtained after the gain control as the first radar echo of the area to be detected.

According to an embodiment of the present application, the eliminating static background wave components and direct current component components in the initial radar echo to obtain the eliminated candidate radar echo of the region to be detected includes: removing the static background wave component in the initial radar echo based on a trace signal subtraction method; and eliminating the direct-current component in the initial radar echo based on the time average subtraction.

According to an embodiment of the present application, before acquiring the trained target living body detection model and extracting the living body detection result of the second radar echo based on the target living body detection model, the method includes: acquiring a to-be-trained life body detection model; acquiring a sample vital sign signal, a sample environment signal and a sample interference signal, and generating a training sample of the vital body detection model; training a to-be-trained life body detection model based on the training sample until the training is finished to obtain the trained target life body detection model.

The second aspect of the present application also provides a living body detecting apparatus, including: the system comprises a traversing module, a first radar echo acquisition module and a second radar echo acquisition module, wherein the traversing module is used for acquiring a first radar echo of a region to be detected, constructing a sampling window to traverse the first radar echo and acquiring a second radar echo in the sampling window when the window is slid each time; the identification module is used for acquiring a trained target life body detection model and acquiring a life body detection result of the second radar echo based on the target life body detection model; and the acquisition module is used for acquiring a target life body detection result of the area to be detected corresponding to the first radar echo according to respective life body detection results of all the second radar echoes.

A third aspect of the present application proposes a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the living-body detecting method as proposed in the first aspect above.

According to the method and the device for detecting the living body, the first radar echo of the area to be detected is obtained, the first radar echo is traversed through the sampling window which is constructed in a sliding mode, and the second radar echo framed in the sampling window is obtained. And further, inputting the second radar echo into the trained target life body detection model, and acquiring a life body detection result of the second radar echo according to an output result of the model. And acquiring the life body detection result of the first radar echo according to the respective life body detection results of all the second radar echoes, and determining the life body detection result as the target life body detection result of the region to be detected. In the method, the first radar echo is subjected to traversal sampling to obtain the second radar echo, the life body detection result of the first radar echo is obtained through the life body detection result of the second radar echo, the calculation amount of the life body detection result of the radar echo is reduced, the life body detection efficiency is improved, the life body detection result is determined based on a trained target life body detection model, the accuracy of life body detection in a region to be detected is improved, and a life body detection algorithm is optimized.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application and are not to be construed as limiting the application.

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

FIG. 2 is a diagram illustrating radar returns according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of a method for detecting a living body according to another embodiment of the present application;

FIG. 4 is a schematic illustration of a radar echo according to another embodiment of the present application;

FIG. 5 is a schematic diagram of a method for detecting a living body according to another embodiment of the present application;

FIG. 6 is a schematic illustration of a radar echo according to another embodiment of the present application;

FIG. 7 is a schematic diagram of a method for detecting a living body according to another embodiment of the present application;

FIG. 8 is a schematic diagram of a method for detecting a living body according to another embodiment of the present application;

FIG. 9 is a schematic view of a life detection model according to an embodiment of the present application;

FIG. 10 is a schematic illustration of a radar echo according to another embodiment of the present application;

FIG. 11 is a schematic diagram illustrating a variation of a loss value in training of a living body detection model according to an embodiment of the present application;

FIG. 12 is a schematic diagram illustrating recognition accuracy of a life detection model according to an embodiment of the present application;

FIG. 13 is a schematic output diagram of a life detection model according to an embodiment of the present application;

FIG. 14 is a schematic view of a method for detecting a living body according to another embodiment of the present application;

FIG. 15 is a schematic flow chart of a living body detecting device according to an embodiment of the present application;

fig. 16 is a block diagram of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make the technical solutions of the present application better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functionality throughout. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application and should not be construed as limiting the present application.

Fig. 1 is a schematic diagram of a living body detection method according to an embodiment of the present application, as shown in fig. 1, the method includes:

s101, acquiring a first radar echo of a region to be detected, constructing a sampling window to traverse the first radar echo, and acquiring a second radar echo in the sampling window when the window is slid each time.

In some scenarios, a life may exist in an area covered by an entity, in which the life information in the area cannot be directly acquired through observation, and the life information in the area can be acquired through radar detection of the area.

The region to be detected may be identified as a region to be detected.

Optionally, a pulse signal may be sent to the region to be detected by a radar detection instrument, and a radar echo of the region to be detected is collected as a first radar echo of the region to be detected, where the first radar echo may be a radar echo matrix.

Furthermore, the first radar echo returned by the area to be detected comprises a radar echo returned by the life in the area to be detected and a radar echo returned by other non-life objects in the area to be detected, so that the life information in the area to be detected can be determined by analyzing and identifying the information carried by the first radar echo.

As shown in fig. 2, the acquired first radar echo of the region to be detected is set as shown in the image with the number 1 in fig. 2, where the image is framed by a dashed frame in the figure, that is, a part of radar echo returned by a living body in the region to be detected, which is obtained based on analysis in the first radar echo of the region to be detected, where as can be seen from the mapping position information of the part of radar echo in the fast time dimension and the slow time dimension shown in the image with the number 1 in fig. 2, the living body is located at a distance 4m from the radar detection instrument and does not move to other positions.

For another example, the acquired first radar echo of the region to be detected is set as shown in the image with the number 2 in fig. 2, where the dashed frame in the figure is defined as a part of radar echo returned by the living body in the region to be detected, which is obtained based on analysis in the first radar echo of the region to be detected, and it can be known from the mapping position information of the part of radar echo in the fast time dimension and the slow time dimension shown in the image with the number 2 in fig. 2 that the living body is located at the position 3m to 12m away from the radar detection instrument in a reciprocating manner.

Further, in order to more accurately detect the information of the living body from the first radar echo, the first radar echo may be segmented and sampled, and the vital sign information in a part of the radar echoes obtained after sampling may be analyzed.

Optionally, a corresponding sampling window may be constructed according to the first radar echo, and the first radar echo is traversed by sliding of the sampling window, wherein a part of the radar echo framed in the sampling window when the sampling window is slid may be determined as the second radar echo.

And S102, acquiring the trained target life body detection model, and acquiring a life body detection result of the second radar echo based on the target life body detection model.

In the embodiment of the application, the life body in the second radar echo can be identified through the trained target life body detection model, wherein the second radar echo obtained by sampling can be input into the trained target life body detection model, and the second radar echo is analyzed and processed through the target life body detection model, so that the life body detection result corresponding to the second radar echo is obtained.

Optionally, it may be determined whether the second radar echo carries vital sign data of a living body in the region to be measured through an output result of the target living body detection model, so as to identify and detect the living body in the region to be measured.

And S103, acquiring a target life detection result of the to-be-detected area corresponding to the first radar echo according to the life detection results of all the second radar echoes.

In the embodiment of the application, the second radar echo is a part of radar echoes in the first radar obtained by traversing the first radar echo, so that the first radar echo can be covered by all the second radar echoes.

In this scenario, a living body detection result corresponding to the first radar echo may be obtained according to a living body detection result of each of all the second radar echoes.

Optionally, there is a corresponding transmission time for a pulse signal sent by the radar detection instrument, and there is a corresponding return time for a first radar echo corresponding to the pulse signal, so that there is a corresponding return time for a second radar echo obtained by performing traversal sampling on the first radar echo, and all the life detection results of the second radar echoes may be spliced in a time sequence dimension based on the respective corresponding return times of all the second radar echoes, and according to the life detection result obtained after splicing, a life detection result in the first radar echo is obtained, and the life detection result is used as a target life detection result of the region to be detected.

According to the method for detecting the living body, the first radar echo of the area to be detected is obtained, the first radar echo is traversed through the sampling window which is constructed in a sliding mode, and the second radar echo framed in the sampling window is obtained. And further, inputting the second radar echo into the trained target life body detection model, and acquiring a life body detection result of the second radar echo according to an output result of the model. And acquiring the life body detection result of the first radar echo according to the respective life body detection results of all the second radar echoes, and determining the life body detection result as the target life body detection result of the region to be detected. In the application, the first radar echo is subjected to traversal sampling to obtain the second radar echo, the life body detection result of the first radar echo is obtained through the life body detection result of the second radar echo, the calculated amount of the life body detection result of the radar echo is reduced, the life body detection efficiency is improved, the life body detection result is determined based on a trained target life body detection model, the accuracy of life body detection in a region to be detected is improved, and a life body detection algorithm is optimized.

In the above embodiment, regarding the acquisition of the target living body detection result of the region to be detected, it can be further understood by referring to fig. 3, fig. 3 is a schematic flow chart of a living body detection method according to another embodiment of the present application, and as shown in fig. 3, the method includes:

and S301, integrating respective life body detection results of all second radar echoes to obtain candidate life body detection results of the first radar echoes.

In this embodiment, the respective life body detection results of all the second radar echoes may be integrated based on a data integration method in the related art, and the integrated life body detection result may be determined as a candidate life body detection result of the first radar echo.

Optionally, the second radar echo has a corresponding position in the first radar echo, so that all the life detection results of the second radar echo are spliced according to the position of the second radar echo in the first radar echo, and the candidate life detection result of the first radar echo is obtained according to the spliced life detection result.

In the implementation, the first radar echo is a radar echo signal generated in a region to be detected by a pulse signal sequence which is sent by a radar detection instrument and has a duration set time, and under the scene that the first radar echo is traversed by sliding a sampling window, a second radar echo obtained by sampling has a corresponding position in the first radar echo.

Therefore, the position relationship between the life detection results of all the second radar echoes can be determined based on the positions of all the second radar echoes in the first radar echo, further, the life detection results of all the second radar echoes are spliced based on the position relationship, and the candidate life detection result of the first radar echo is obtained according to the spliced life detection result.

In some implementations, there may be overlap of partial radar echoes between adjacent second radar echoes in the first radar echo, so that the spliced living body detection results may be de-overlapped and candidate living body detection results of the first radar echo may be obtained.

And S302, determining and deleting the false recognition result in the candidate life body detection result, and determining the deleted candidate life body detection result as the target life body detection result of the region to be detected corresponding to the first radar echo.

In implementation, due to the complexity of the environment in the region to be detected, there is a possibility that a false recognition result may occur in the candidate living body detection result of the first radar echo, and therefore, it is necessary to determine the false recognition result in the candidate living body detection result and delete the false recognition result from the candidate living body detection result.

Optionally, the candidate life body detection results may be clustered, and a maximum euclidean distance and an average euclidean distance of each cluster may be obtained.

In the embodiment of the present application, each of the candidate life detection results may be clustered, where the candidate life detection results may be subjected to distance according to a K-means clustering algorithm (K-means clustering algorithm), and a misrecognition result in the candidate life detection results may be recognized according to a cluster obtained after clustering.

The candidate life body detection results may be clustered according to euclidean distances between all the life body detection results included in the candidate life body detection results.

Setting a candidate living body detection result x _i All the living body detection results included are x _i ＝{x ₁ ,x ₂ ,x ₃ ,…,x _m Get x according to the Euclidean distance calculation formula _i Corresponding k neighbor distance matrices, wherein the Euclidean distance d _xy The calculation formula of (a) is as follows:

in the above formula, n represents the data space dimension of each of the candidate living body detection results, and k is x _i The number of corresponding close-range matrices, in { x } according to the above formula ₁ ,x ₂ ,x ₃ ,…,x _m K cluster sample points determined in (1), and according to the k cluster sample point pairs x _i Clustering is performed, wherein a sample point set composed of k cluster sample points can be identified as N _k (x)。

Further, based on the clustering rule in the related art, N is _k (x) In determining x _i The cluster y to which it belongs _i The formula is as follows:

in the above formula, K is a cluster y _i Further, the total number of the candidate life body detection results x _i All the living body detection results included are x _i ＝{x ₁ ,x ₂ ,x ₃ ,…,x _m Constructing a corresponding cluster structure, such as a K-MST structure, so as to realize the detection result x of the candidate life body _i And obtaining the clustered class y _i ，c _j And detecting a result set for the life bodies in each clustered class.

In the implementation, the coverage range of the class cluster in the candidate life body detection result can be obtained through the maximum Euclidean distance and the average Euclidean distance of the class cluster, so that the maximum Euclidean distance and the average Euclidean distance of all the class clusters obtained after clustering can be calculated, and the misrecognition result is determined from the candidate life body detection result according to the maximum Euclidean distance and the average Euclidean distance.

The detection result which is not covered by the class cluster in the candidate life body detection result can be determined through the maximum Euclidean distance and the average Euclidean distance of the class cluster, and the part of detection results can be understood as the outlier which does not belong to the class cluster, so that the detection result corresponding to the outlier can be determined as the false recognition result in the candidate life body detection result.

Further, the false recognition result in the candidate life body detection result is deleted, and the target life body detection result is obtained.

The determined false recognition result may be deleted from the candidate life detection result, and the deleted candidate life detection result is a target life detection result corresponding to the to-be-detected region carried in the first radar echo.

As shown in fig. 4, in the image numbered 1 in fig. 4, if the point 1 and the point 2 are outliers in the image, the living body detection results corresponding to the point 1 and the point 2 can be deleted from the candidate living body detection results, and the target living body detection result of the region to be measured to which the corresponding first radar echo belongs in the image numbered 1 can be obtained.

In the image numbered 2 in fig. 4, if the point 3 is an outlier in the image, the living body detection result corresponding to the point 3 may be deleted from the candidate living body detection results, and a target living body detection result of the region to be measured to which the corresponding first radar echo belongs in the image numbered 2 may be obtained.

In the image numbered 3 in fig. 4, if the point 4, the point 5, and the point 6 are outliers in the image, the living body detection results corresponding to the point 4, the point 5, and the point 6 may be deleted from the candidate living body detection results, and the target living body detection result of the region to be measured to which the corresponding first radar echo belongs in the image numbered 3 may be obtained.

In the image numbered 4 in fig. 4, if the points 7 and 8 are outliers in the image, the detection results of the living bodies corresponding to the points 7 and 8 can be deleted from the detection results of the candidate living bodies, so as to obtain the detection result of the target living body in the region to be measured to which the corresponding first radar echo belongs in the image numbered 4.

According to the method for detecting the life body, the respective life body detection results of all the second radar echoes are integrated, so that the candidate life body detection result of the first radar echo is obtained, further, the false recognition result in the candidate life body detection result is recognized and deleted, and therefore the target life body recognition result of the region to be detected carried by the first radar echo is obtained. According to the method and the device, the accuracy of the target life body detection result is improved and the life body detection method is optimized through determining and deleting the error recognition result.

Further, a state trajectory of a living body in the region to be detected may be obtained according to the target living body detection result, which can be further understood by referring to fig. 5, where fig. 5 is a schematic flow diagram of a living body detection method according to another embodiment of the present application, and as shown in fig. 5, the method includes:

s501, acquiring a detection state value, a prediction state value and a Kalman gain of a life body in a region to be detected in a single slow time from a target life body detection result.

In the embodiment of the application, the track extraction can be performed on the detection result of the target life body based on the Kalman filtering algorithm, so that the moving track of the life body in the area to be detected is obtained.

In the implementation, the first radar echo has a plurality of corresponding slow times, wherein the moving track of the living body in the area to be detected can be determined according to the position, the moving speed and other related information of the living body in the area to be detected in each slow time.

Alternatively, from the target life object detection result, the detection state value, the prediction state value, and the corresponding kalman gain of the life object at each single slow time may be acquired.

The detection state value may include a position and a moving speed of the living body in the region to be detected, which are acquired through the first radar echo, and the prediction state value may include a position and a moving speed of the living body in the region to be detected, which are predicted according to a kalman filter algorithm.

Optionally, the predicted state value X of the living body in the region to be detected in the slow time k _k The calculation formula of (c) is as follows:

X _k ＝A _k X _k +B _k U _k +w _k

in the above formula, U _k For the amount of control applied to the radar detection system over a slow time k, A _k For the state transition matrix acting on the slow time k-1, B _k To act on U _k A control matrix of (a), wherein B _k Can be used to describe the influence of the control quantity on the state of the radar detection system, w _k Process noise detected for radar, wherein w _k Obey mean of 0, covariance of Q _k Is independently multivariate normally distributed.

Optionally, the detection state value Z of the living body in the region to be detected in the slow time k _k The calculation formula of (a) is as follows:

Z _k ＝H _k X _k +v _k

in the above formula, X _k Is a predicted state value H of a living body in the region to be measured at a slow time k _k To detect a matrix, in which H _k Can be used to describe the mapping of the actual state space to the probe space, v _k To detect noise, wherein v _k Obedience mean of 0 and covariance of R _k Is normally distributed.

Wherein the state value Z is detected _k Including the position of the living body detected by the radar detection instrument in the region to be detected. In the implementation, the radar detection instrument has a corresponding set coordinate system, the position of the living body in the region to be detected has a mapping position in the set coordinate system.

Further, the abscissa and the ordinate corresponding to the mapping position may be obtained by the following formula:

in the above-mentioned formula,

the target detection result detected by the radar detection instrument in the slow time k comprises the mean value of all horizontal coordinates of the position of the living body in the region to be detected, which are mapped in the set coordinate system,

the method is characterized in that the target detection result detected by a radar detection instrument in slow time k comprises the mean value of all vertical coordinates of the position of a living body in a region to be detected, which are mapped in a set coordinate system, and t is time.

In some implementations, the predicted state value of the living body in the region to be measured at the current slow time can be determined by the predicted state value of the living body in the region to be measured at the previous slow time of the current slow time, and further, the target state value of the living body in the region to be measured at the current slow time can be obtained according to the detection state value and the predicted state value of the living body in the region to be measured at the current slow time.

Alternatively, if the current slow time is set as the slow time k, the formula according to the predicted state value of the living body in the region to be measured at the previous slow time passing through the current slow time may be as follows:

in the above-mentioned formula,

to obtain the a priori predicted value at slow time k from the predicted state value at slow time k-1,

for the life in the region to be measured at the slow time k-1Is set to the target state value of (a),

for an estimated error value, Q, between a predicted state value and a detected state value of the radar detection system over a slow time k _k Detection process noise w for radar detection systems _k Covariance matrix of (2), P _k-1 Is a covariance matrix of posterior errors, U, over a slow time k-1 _k For the amount of control applied to the radar detection system over a slow time k, A _k To act on the state transition matrix at slow time k-1,

is A _k Transposed matrix of (A), B _k To act on U _k A control matrix of (a), wherein B _k Can be used to describe the influence of the control quantity on the state of the radar detection system, w _k Process noise detected for radar, wherein w _k Obedience mean 0 and covariance Q _k Is independently multivariate normally distributed.

Further, a Kalman gain K over a slow time K is obtained _k The formula can be as follows:

in the above formula, R _k Detecting noise v for radar detection systems _k The covariance matrix of (a) is determined,

is an estimated error value between a predicted state value and a detected state value of the radar detection system over a slow time k, H _k To detect a matrix, in which H _k Can be used to describe the mapping of the actual state space to the probe space,

is H _k The transposed matrix of (2).

Further, determining the incidence relation between the prediction state value and the detection state value in the slow time k according to the Kalman gain, thereby obtaining the target state value of the living body in the region to be detected in the slow time k.

And S502, determining a target state value of the living body in a single slow time according to the detection state value, the prediction state value and the Kalman gain.

In the embodiment of the application, the detection state value, the prediction state value and the Kalman gain can be calculated based on a set algorithm, so that the target state value of the living body in a single slow time is obtained.

Wherein, if the current slow time is set as the slow time k, the target state value

The calculation formula of (a) is as follows:

in the above-mentioned formula,

is a priori predicted value K on slow time K acquired according to a predicted state value on slow time K-1 _k Is the Kalman gain, Z, over a slow time, k _k Is the detection state value H of the living body in the region to be detected in the slow time k _k To detect a matrix, in which H _k Can be used to describe the mapping of the actual state space to the probe space.

It should be noted that the target state value of the living body at the slow time k includes a target position and a target moving speed of the living body in the region to be measured at the slow time k, wherein the target position and the target moving speed can be understood as a position parameter and a moving speed parameter which are as close as possible to an actual position and an actual moving speed of the living body in the region to be measured.

And S503, acquiring the moving track of the living body in the area to be detected according to the target state values corresponding to all slow times in the first radar echo.

In the embodiment of the application, according to the target position and the target speed included in the target state value of each of the living bodies in all the single slow time periods, the moving position and the moving speed of the living body in the region to be detected can be obtained, so that the moving track of the living body in the region to be detected is generated.

For example, it is assumed that target state values corresponding to all slow times included in the first radar echo of the living body respectively indicate that three living bodies exist in the region to be detected, where the first living body is located at the position 5m away from the radar detection instrument and does not move, the second living body is located at the position 8m away from the radar detection instrument and does not move, and the third living body is located at the position 11m away from the radar detection instrument and does not move, in this scenario, moving trajectories of the three living bodies in the region to be detected may be as shown in the image numbered 1 in fig. 6.

For another example, if it is set that target state values corresponding to all slow times included in the first radar echo of the living body respectively indicate that there is a living body in the region to be detected, where the living body reciprocates between the position of the range radar detection instrument 3m and the position of the range radar detection instrument 7.5m at a set speed, then the moving trajectory of the living body in the region to be detected in the scene may be as shown in the image numbered 2 in fig. 6.

For another example, it is assumed that target state values corresponding to all slow times included in the first radar echo of the living body respectively indicate that three living bodies exist in the region to be detected, where the three living bodies reciprocate based on the set speed between the position of the range radar detection instrument 4m and the position of the range radar detection instrument 12m at the same time, and then the moving trajectories of the three living bodies in the region to be detected in the scene may be as shown in the image numbered 3 in fig. 6.

For another example, it is set that target state values corresponding to all slow times included in the first radar echo of the living body indicate that one living body exists in the region to be measured, where the living body randomly moves in the region to be measured, and then a moving track of the living body in the region to be measured in the scene may be as shown in an image numbered 4 in fig. 6.

It should be noted that after the target state value of the living entity at the current slow time is obtained, the posterior error covariance matrix corresponding to the current slow time may be updated based on the target state value at the current slow time, so that the target state value of the living entity at the next slow time in the region to be measured may be obtained according to the updated posterior error covariance matrix.

Setting the current slow time as k, and setting the covariance matrix P of the posterior error in the slow time k _k The calculation formula of (2) is as follows:

in the above formula, K _k Is the Kalman gain on slow time k, H _k To detect a matrix, in which H _k Can be used to describe the mapping of the actual state space to the probe space,

is an estimated error value between a predicted state value and a detected state value of the radar detection system in a slow time k, and I is an identity matrix.

According to the method for detecting the living body, the target state value of the living body in a single slow time is obtained according to the detection result of the target living body, and further, the moving track of the living body in the area to be detected is obtained according to the respective target state values in all slow times. According to the method and the device, the moving track of the life body in the area to be detected is obtained through the target life body detection result, the accuracy of the moving track is improved, and the precision of the moving track is optimized.

In the above embodiment, regarding the acquisition of the first radar echo, it can be further understood with reference to fig. 7, fig. 7 is a schematic flowchart of a living body detection method according to another embodiment of the present application, and as shown in fig. 7, the method includes:

and S701, acquiring an initial radar echo of the area to be detected according to the ultra-wideband radar.

In the embodiment of the application, the radar detection device provided with the ultra-wideband radar can send a continuous pulse signal sequence to the area to be detected and receive an echo signal returned by the area to be detected.

And further, taking the received echo signal returned by the area to be detected as an initial radar echo of the area to be detected.

And S702, eliminating static background wave components and direct current component components in the initial radar echo to obtain the eliminated candidate radar echo of the area to be detected.

In the embodiment of the present application, due to the complexity of the environment of the region to be detected, there may be some echoes in the initial radar echo that may affect the detection result of the living body, and therefore, the echo may need to be eliminated.

In some implementations, the initial radar echo may include a constituent as shown in the following equation:

R[m,n]＝r[m,n]+c[n]+w[m,n]+d[m]+l[m,n]+t[m,n]

in the above formula, R [ m, n ] is the initial radar echo, R [ m, n ] is the vital sign signal of the living body in the region to be measured, c [ n ] is the static background wave, w [ m, n ] is the additive white noise, d [ m ] is the direct current component in the unstable fast time, l [ m, n ] is the linear trend generated in the slow time due to the stable amplitude value in the detection process of the radar detection system, t [ m, n ] is the harmonic wave and signal distortion caused by the moving body in the region to be measured.

In addition, the static background wave and the dc component in the unstable fast time in the initial radar echo may affect the result of detecting the living body, and therefore, it needs to be eliminated from the initial radar echo.

Optionally, the static background wave component in the initial radar echo may be eliminated based on a trace signal subtraction, wherein the computation formula of the trace signal subtraction is as follows:

R'[m,1]＝R[m,1]

R'[m,n]＝R[m,n]-R[m,n-1](m＝1,2,...,M)(n＝2,3,...,N)

in the above formula, R'm, n represents an initial radar echo obtained by eliminating a static background wave component from the initial radar echo.

Alternatively, the dc component in the initial radar echo may be removed based on time-averaged subtraction, wherein the calculation formula of the time-averaged subtraction is as follows:

in the above formula, R ″ [ m, n ] is an initial radar echo in which a dc component is eliminated from an initial radar echo in which a static background wave component is eliminated.

Further, the initial radar echo for eliminating the static background wave component and the direct current component in the unstable fast time can be determined as a candidate radar echo of the area to be detected.

And S703, performing gain control on the candidate radar echo, and taking the radar echo obtained after the gain control as a first radar echo of the area to be detected.

In the embodiment of the application, in order to improve the signal-to-noise ratio of the vital sign signals carried in the radar echo, gain control can be performed on the candidate radar echo, so that the vital sign signals are enhanced in the fast time direction.

Alternatively, the algorithm for gain control of the candidate radar returns may be as follows:

in the above formula, d is the window length of the time window, g _max Is the maximum gain threshold, g _mask Is a gain factor.

And further, using the candidate radar echo obtained after gain control as a first radar echo of the area to be detected.

According to the method for detecting the living body, static background clutter and direct current components are eliminated from an initial radar echo of a region to be detected, a candidate radar echo after elimination is obtained, and further gain control is carried out on the candidate radar echo to obtain a first radar echo of the region to be detected. By preprocessing the signal of the initial radar echo, the interference factor in the first radar echo is eliminated, meanwhile, the vital sign signal in the first radar echo is strengthened, the accuracy of detecting the vital body is improved, and the detection method of the vital body is optimized.

In the above embodiment, regarding the obtaining of the target living body detection model, it can be further understood with reference to fig. 8, fig. 8 is a schematic flowchart of a living body detection method according to another embodiment of the present application, and as shown in fig. 8, the method includes:

s801, obtaining a to-be-trained life body detection model.

In the embodiment of the present application, a to-be-trained living body detection model may be constructed based on a CNN model structure, as shown in fig. 9, the to-be-trained living body detection model may include a 5-layer neural network shown in fig. 9, which includes 4 feature extraction layers and 1 full connection layer.

As shown in fig. 9, the 4 feature extraction layers include a feature extraction layer 1, a feature extraction layer 2, a feature extraction layer 3, and a feature extraction layer 4, where the feature extraction layer 1, the feature extraction layer 2, the feature extraction layer 3, and the feature extraction layer 4 respectively include 1 convolutional layer, 1 batch normalization layer, and 1 nonlinear function activation layer (relu).

Further, as shown in fig. 9, the living body exploration model to be trained further includes 1 maximum pooling layer and 1 residual connection, where the residual connection includes 1 convolutional layer and 1 batch normalization layer, and the feature map output by the feature extraction layer 2 can be transmitted to the non-linear function activation layer (relu) of the feature extraction layer 4 through the residual connection.

In implementation, the neural network layer of the living body detection model has set model parameters, and optionally, the model parameter setting of the living body detection model in the embodiment of the present application may be as shown in fig. 9, where the number N of convolution kernels in the convolution layer of the feature extraction layer 1 is N _C Is 8, the convolution kernel size FH XFW is 3X 3, stepLong SL of 1, fill P _n Is 1. Number N of convolution kernels in convolution layer of feature extraction layer 2 _C 16, the convolution kernel size FH XFW is 3X 3, the step SL is 1, and the padding P is _n Is 1. Number of convolution kernels N in convolution layer of feature extraction layer 3 _C At 8, the convolution kernel size FH × FW is 3 × 3, the step SL is 1, and the padding P is _n Is 1. Number N of convolution kernels in convolution layer of feature extraction layer 4 _C At 32, the convolution kernel size FH × FW is 3 × 3, the step SL is 1, and the padding P is _n Is 1. Number of convolution kernels N in maximum pooling layer _C At 32, the convolution kernel size FH × FW is 2 × 2, the step SL is 2, and the padding P is _n Is 0. Number of convolution kernels N in convolutional layer in residual concatenation _C At 32, the convolution kernel size FH × FW is 1 × 1, the step SL is 1, and the padding P is _n Is 0.

Accordingly, there are set size parameters for the input size and the output size of the neural network layer of the living body exploration model, as shown in fig. 9, the input size in the feature extraction layer 1 is (1,10,20), the output size is (8,10,20), the input size in the feature extraction layer 2 is (8,10,20), the output size is (16,10,20), the input size in the feature extraction layer 3 is (16,10,20), the output size is (16,10,20), the input size in the feature extraction layer 4 is (16,10,20), the output size is (32,10,20), the input size of the maximum pooling layer is (32,20,20), the output size is (32,5,10), the input size of the fully connected layer is 32 × 5 × 10=1600, and the output size is 3=1600.

It should be noted that, normalization of the weight parameters can be realized by the living body detection model through the batch normalization layer and the nonlinear function activation layer relu in each feature extraction layer.

S802, acquiring a sample vital sign signal, a sample environment signal and a sample interference signal, and generating a training sample of the vital body detection model.

In order to realize effective training of the detection model of the living body to be trained, a corresponding training sample can be constructed based on the environment of an actual rescue scene.

In some implementations, the radar echo signals acquired at the rescue site may include vital sign signals of trapped persons, environment signals of the rescue site, and interference signals in the environment of the rescue site caused by movement of the trapped persons, so that corresponding sample vital sign signals, sample environment signals, and sample interference signals may be acquired respectively, and the acquired sample vital sign signals, sample environment signals, and sample interference signals are mixed, thereby acquiring training samples required by training of the living body detection model.

As shown in fig. 10, setting the image numbered 1 in fig. 10 as a corresponding image of the sample vital sign signal, the image numbered 2 as a corresponding image of the sample environment signal, and the image numbered 3 as a corresponding image of the sample interference signal, may randomly mix the images numbered 1, 2, and 3, and determine the mixed signal obtained by random mixing as a training sample of the living body detection model to be trained.

And S803, training the to-be-trained life body detection model based on the training sample until the training is finished to obtain the trained target life body detection model.

In the embodiment of the application, the batch division can be performed on the mixed signal of the acquired sample vital sign signal, the sample environment signal and the sample interference signal, the training sample is divided into a plurality of batches of sub-samples, and the living body detection model is trained through each sub-sample.

Optionally, in the training process of the life detection model, the training output result of the model may include the labels of the subsamples input into the life detection model in the current round and the corresponding confidences of the labels.

Further, the output result of the life detection model may be a K-dimensional vector including at least one label to which the sub-sample input to the life detection model for training is recognized by the model, and a confidence corresponding to the label, where the confidence of the at least one label to which the sub-sample output by the model belongs may be identified as

Has a value interval of [ q ] _i,0 ,q _i,1 ...,q _i,K-1 ]。

The dimensionality of the vector output by the model can be determined according to the number of labels of the training samples, for example, the training samples input into the life body detection model for training are set to include 3 labels, which are respectively corresponding labels of the vital sign signals, corresponding labels of the environmental signals and corresponding labels of the interference signals, and under the scene, the dimensionality K of the vector output by the model training can be 3.

Optionally, the adjustment and optimization of the model parameters of the to-be-trained life body detection model may be performed according to the training output result of each training round of the life body detection model in the model training process and the loss value between the training samples input into the model in the corresponding training round until the condition that the model training is finished is satisfied.

The loss value between the training output result of each training turn of the life detection model in the model training process and the training sample of the corresponding training turn input model can be obtained based on the following calculation formula of a loss function loss:

in the above-mentioned formula,

the cross entropy corresponding to the output result of the model,

for a regularization penalty term in the model training process, N is the number of samples included in the subsamples for training the current round input model, K is the number of classification labels, q is _i,k Representing the confidence with which the ith sample was identified as the kth label, λ is the regularization parameter,

for the weight parameter of each layer in the life body detection model, l represents the number of layers of the life body detection modelAnd n is the weight number of the corresponding layer.

Alternatively, the variation of the loss value in the training process of the living body detection model may be as shown in fig. 11, and the loss value of the living body detection model shows a decreasing trend with the increase of the number of iterations and the optimization of the model parameters.

In the embodiment of the present application, the mixed signal of the sample vital sign signal, the sample environment signal, and the sample interference signal may be divided in proportion to obtain the divided training sample and the test sample, for example, 80% of the mixed signal in the mixed signal may be used as the training sample of the living body detection model, and the remaining 20% of the mixed signal may be used as the test sample of the trained living body detection model.

Optionally, the number of training iterations of the life detection model may be set, for example, if the number of iterations of each training round of the life detection model is set to 80, the number of iterations in the model training process may be monitored and recorded, and if the 80 th iteration optimization is performed on the model parameters after the current model training is finished, it may be determined that the current round of model training is finished.

Further, in the process of model training, model parameters can be adjusted and optimized based on a set model optimization algorithm.

For example, the Adam algorithm may be selected to adjust and optimize the model parameters, and set the initial learning rate of the model to 0.003, and further set the learning rate decay period to 10 training rounds and the learning rate decay factor to 0.8. After 10 rounds of model training and weight parameter adjustment and optimization of the model are performed, the initial learning rate can be screened based on an attenuation factor of 0.8, so that the update gradient of the weight parameter is slowed down, and further, the refined model optimization of the life body detection model in the training process is realized.

Further, after each round of model training is finished, the recognition accuracy of the trained life body detection model can be obtained through the test sample, and the end condition of the model training can be set based on the recognition accuracy of the life body detection model to the test sample.

If the condition for ending the model training is set to be that the identification accuracy of the test sample is greater than or equal to the accuracy threshold, the test sample can be input into the life body detection model of which the training of the current round is ended after the model training of each round is ended. If the recognition success rate of the life body detection model after the current round of training on the test sample is greater than or equal to the accuracy threshold, it can be determined that the life body detection model after the current round of training meets the model training ending condition, the model training on the life body detection model can be ended, and the life body detection model obtained after the current round of training is ended is used as the trained target life body detection model.

Optionally, the recognition accuracy of the life detection model for the training sample and the recognition accuracy of the test sample after each round of training are finished may be as shown in fig. 12, and as can be seen from fig. 12, along with the increase of the number of iterations, the recognition accuracy of the life detection model for the training sample and the test sample presents an increasing trend.

Further, the radar echo to be recognized may be input into a trained target life body detection model for recognition, where the target life body detection model may output the recognition result of the radar echo to be recognized based on a model output in the form of a confusion matrix, and optionally, the confusion matrix output by the target life body detection model may be as shown in fig. 13.

As can be seen from fig. 13, the radar echo to be recognized, which is input into the target life body detection model, includes a vital sign signal, an environmental signal and an interference signal, where line 1 is a recognition result of the target life body detection model on a component of the vital sign signal in the radar echo to be recognized, line 2 is a recognition result of the target life body detection model on a component of the environmental signal in the radar echo to be recognized, and line 3 is a recognition result of the target life body detection model on a component of the interference signal in the radar echo to be recognized.

According to the content in the row 1, the identification result of the target vital sign signal to the vital sign signal component in the radar echo to be identified is the vital sign signal with the confidence coefficient of 0.95, the environmental signal with the confidence coefficient of 0.02 and the interference signal with the confidence coefficient of 0.03, so that the identification result of the target vital sign signal to the vital sign signal in the radar echo to be identified is accurate.

Correspondingly, as can be seen from the content in row 2, the recognition result of the target vital sign signal on the environmental signal component in the radar echo to be recognized is the vital sign signal with the confidence coefficient of 0.01, the environmental signal with the confidence coefficient of 0.92, and the interference signal with the confidence coefficient of 0.07, so that the recognition result of the target vital sign signal on the environmental signal in the radar echo to be recognized is accurate.

Further, as can be seen from the content in row 3, the recognition result of the target vital sign signal for the interference signal component in the radar echo to be recognized is the vital sign signal with the confidence coefficient of 0.06, the ambient signal with the confidence coefficient of 0.07, and the interference signal with the confidence coefficient of 0.87, so that it can be seen that the recognition result of the target vital sign signal for the ambient signal in the radar echo to be recognized is accurate.

According to the method for detecting the life body, the life body detection model to be trained is obtained, the life body detection model is trained based on the training sample composed of the mixed signal of the sample vital sign signal, the sample environment signal and the sample interference signal, and then the trained target life body detection model is obtained. In the process of detecting the life body, the identification and extraction of the vital sign signals in the first radar echo can be realized based on the trained target life body detection model, and the accuracy of detecting the life body in the region to be detected is improved.

To better understand the above embodiments, fig. 14 can be combined with fig. 14, and fig. 14 is a schematic flowchart of a living body detecting method according to another embodiment of the present application, as shown in fig. 14:

after the initial radar echo of the area to be detected is obtained, preprocessing is carried out on the initial radar echo, interference components in the initial radar echo are removed through a channel signal subtraction method and a time averaging method, gain control is carried out on the initial radar echo after the interference components are removed, and a first radar echo of the area to be detected is obtained. And constructing a sampling window, traversing the first radar echo by sliding the sampling window to obtain a second radar echo, inputting the second radar echo into a trained target life detection model, and obtaining a life detection result of the second radar echo.

And acquiring candidate life body detection results of the region to be detected according to respective life body detection results of all the second radar echoes, clustering the candidate life body detection results to acquire false recognition results therein, and deleting the false recognition results from the candidate life body detection results to obtain target life body detection results of the region to be detected carried in the first radar echoes.

Further, the moving track of the living body in the region to be detected is obtained according to the target living body detection result, so that the detection of the living body in the region to be detected is realized.

According to the method for detecting the life body, the first radar echo is subjected to traversal sampling to obtain the second radar echo, the life body detection result of the first radar echo is obtained through the life body detection result of the second radar echo, the calculation amount of the life body detection result of the radar echo is reduced, the efficiency of life body detection is improved, the accuracy of the target life body detection result is improved through determining and deleting the error recognition result, the life body detection result is determined based on the trained target life body detection model, the life body detection algorithm is optimized, the accuracy of life body detection in the region to be detected is improved, the moving track of the life body in the region to be detected is obtained through the target life body detection result, the accuracy of the moving track is improved, and the precision of the moving track is optimized.

In accordance with the methods for detecting a living body proposed by the above embodiments, an embodiment of the present application further proposes a device for detecting a living body, and since the device for detecting a living body proposed by the embodiment of the present application corresponds to the methods for detecting a living body proposed by the above embodiments, the embodiments of the method for detecting a living body are also applicable to the device for detecting a living body proposed by the embodiment of the present application, and will not be described in detail in the following embodiments.

Fig. 15 is a schematic structural diagram of a living body detecting apparatus according to an embodiment of the present application, and as shown in fig. 15, the living body detecting apparatus 1500 includes a traversal module 151, a recognition module 152, an acquisition module 153, a trajectory extraction module 154, and a training module 155, where:

the traversing module 151 is configured to acquire a first radar echo of the area to be measured, construct a sampling window to traverse the first radar echo, and acquire a second radar echo in the sampling window when the window slides each time;

an identification module 152, configured to obtain the trained target living body detection model, and obtain a living body detection result of the second radar echo based on the target living body detection model;

and the obtaining module 153 is configured to obtain a target life detection result of the to-be-detected region corresponding to the first radar echo according to respective life detection results of all the second radar echoes.

In this embodiment, the obtaining module 153 is further configured to: integrating respective life body detection results of all second radar echoes to obtain candidate life body detection results of the first radar echoes; and determining and deleting false recognition results in the candidate life body detection results, and determining the deleted candidate life body detection results as target life body detection results of the region to be detected corresponding to the first radar echo.

In this embodiment of the application, the obtaining module 153 is further configured to: and splicing the life body detection results of all the second radar echoes according to the positions of the second radar echoes in the first radar echoes, and acquiring the candidate life body detection result of the first radar echo according to the spliced life body detection result.

In this embodiment of the application, the obtaining module 153 is further configured to: clustering the candidate life body detection results to obtain the maximum Euclidean distance and the average Euclidean distance of each cluster; determining a false recognition result from the candidate life body detection result according to the maximum Euclidean distance and the average Euclidean distance; and deleting the false recognition result in the candidate life body detection result to obtain the target life body detection result.

In this embodiment, the living body detecting apparatus 1500 further includes a track extracting module 154, configured to: acquiring a detection state value, a prediction state value and a Kalman gain of a life body in a region to be detected in a single slow time from a target life body detection result; determining a target state value of the living body in a single slow time according to the detection state value, the prediction state value and the Kalman gain; and acquiring the moving track of the living body in the area to be detected according to the target state values corresponding to all the slow times in the first radar echo.

In this embodiment of the application, the traversing module 151 is further configured to: acquiring an initial radar echo of a region to be detected according to the ultra-wideband radar; eliminating static background wave components and direct current component components in the initial radar echo to obtain a candidate radar echo of the eliminated to-be-detected area; and performing gain control on the candidate radar echoes, and taking the radar echoes obtained after the gain control as first radar echoes of the area to be detected.

In this embodiment of the application, the traversing module 151 is further configured to: based on the channel signal subtraction method, eliminating static background wave components in the initial radar echo; based on the time-averaged subtraction, the dc component in the initial radar echo is removed.

In this embodiment, the life detection apparatus 1500 further includes a training module 155, configured to: acquiring a to-be-trained life body detection model; acquiring a sample vital sign signal, a sample environment signal and a sample interference signal to generate a training sample of a vital body detection model; training the to-be-trained life body detection model based on the training sample until the training is finished to obtain the trained target life body detection model.

The life body detection device obtains a first radar echo of a region to be detected, and traverses the first radar echo through a sampling window constructed in a sliding mode to obtain a second radar echo framed in the sampling window. And further, inputting the second radar echo into the trained target life body detection model, and acquiring a life body detection result of the second radar echo according to an output result of the model. And acquiring the life body detection result of the first radar echo according to the respective life body detection results of all the second radar echoes, and determining the life body detection result as the target life body detection result of the region to be detected. In the method, the first radar echo is subjected to traversal sampling to obtain the second radar echo, the life body detection result of the first radar echo is obtained through the life body detection result of the second radar echo, the calculation amount of the life body detection result of the radar echo is reduced, the life body detection efficiency is improved, the life body detection result is determined based on a trained target life body detection model, the accuracy of life body detection in a region to be detected is improved, and a life body detection algorithm is optimized.

To achieve the above embodiments, the present application also provides an electronic device, a computer readable storage medium and a computer program product.

Fig. 16 is a block diagram of an electronic device according to an embodiment of the present application, and the electronic device shown in fig. 16 may implement the method for detecting a living body according to the embodiments of fig. 1 to 14.

In order to achieve the above-described embodiments, the present application also provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the living body detecting method of the embodiment of fig. 1 to 14.

In order to implement the above embodiments, the present application further provides a computer program product, which when executed by an instruction processor in the computer program product, executes the living body detecting method of the embodiments of fig. 1 to 14.

In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.

Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, "plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.

Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present application in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present application.

The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.

It should be understood that portions of the present application may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.

It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.

In addition, functional units in the embodiments of the present application may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a separate product, may also be stored in a computer-readable storage medium.

The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application, and that variations, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present application.

Claims

1. A method of detecting a living body, the method comprising:

acquiring a first radar echo of a region to be detected, constructing a sampling window to traverse the first radar echo, and acquiring a second radar echo in the sampling window when the window is slid each time;

acquiring a trained target life body detection model, and acquiring a life body detection result of the second radar echo based on the target life body detection model;

and acquiring a target life body detection result of the area to be detected corresponding to the first radar echo according to respective life body detection results of all the second radar echoes.

2. The method according to claim 1, wherein the obtaining, according to the respective living body detection results of all the second radar echoes, the target living body detection result of the region to be detected corresponding to the first radar echo comprises:

integrating respective life body detection results of all the second radar echoes to obtain candidate life body detection results of the first radar echoes;

and determining and deleting the false recognition result in the candidate life body detection result, and determining the deleted candidate life body detection result as the target life body detection result of the region to be detected corresponding to the first radar echo.

3. The method of claim 2, wherein the integrating the life body detection results of all the second radar echoes to obtain the candidate life body detection result of the first radar echo comprises:

and splicing the life body detection results of all the second radar echoes according to the positions of the second radar echoes in the first radar echoes, and acquiring the candidate life body detection result of the first radar echo according to the spliced life body detection result.

4. The method according to claim 2, wherein the determining and deleting false recognition results in the candidate life body detection results, and determining the deleted candidate life body detection results as the target life body detection results of the region to be detected corresponding to the first radar echo comprises:

clustering the candidate life body detection results to obtain the maximum Euclidean distance and the average Euclidean distance of each cluster;

determining the false recognition result from the candidate life body detection result according to the maximum Euclidean distance and the average Euclidean distance;

and deleting the false recognition result in the candidate life body detection result to obtain the target life body detection result.

5. The method of claim 4, wherein after obtaining the target living body detection result, the method comprises:

acquiring a detection state value, a prediction state value and a Kalman gain of a life body in the region to be detected in a single slow time from the detection result of the target life body;

determining a target state value of the living body at the single slow time according to the detection state value, the prediction state value and the Kalman gain;

and acquiring the moving track of the living body in the region to be detected according to the target state values corresponding to all slow times in the first radar echo.

6. The method of claim 1, wherein the obtaining a first radar echo of the region under test comprises:

acquiring an initial radar echo of the area to be detected according to the ultra-wideband radar;

eliminating static background wave components and direct current component components in the initial radar echo to obtain a candidate radar echo of the area to be detected after elimination;

and performing gain control on the candidate radar echo, and taking the radar echo obtained after the gain control as the first radar echo of the area to be detected.

7. The method according to claim 6, wherein the eliminating static background wave components and direct current component components in the initial radar echo to obtain the eliminated candidate radar echo of the region to be measured includes:

eliminating the static background wave component in the initial radar echo based on a track signal subtraction method;

and eliminating the direct-current component in the initial radar echo based on the time-average subtraction.

8. The method of claim 1, wherein prior to obtaining the trained target life detection model and extracting the life detection result of the second radar echo based on the target life detection model, comprising:

acquiring a to-be-trained life body detection model;

acquiring a sample vital sign signal, a sample environment signal and a sample interference signal, and generating a training sample of the vital body detection model;

and training the life body detection model to be trained based on the training sample until the training is finished to obtain the trained target life body detection model.

9. A living body detecting apparatus, characterized in that the apparatus comprises:

the system comprises a traversing module, a first radar echo acquisition module and a second radar echo acquisition module, wherein the traversing module is used for acquiring a first radar echo of a region to be detected, constructing a sampling window to traverse the first radar echo and acquiring a second radar echo in the sampling window when the window is slid each time;

the identification module is used for acquiring a trained target life body detection model and acquiring a life body detection result of the second radar echo based on the target life body detection model;

and the acquisition module is used for acquiring a target life detection result of the area to be detected corresponding to the first radar echo according to respective life detection results of all the second radar echoes.

10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method of claims 1-8.