Life body perception method and device

Abstract

The invention relates to the technical field of perception, and discloses a life body perception method and device, and the method comprises the steps: collecting a first perception signal corresponding to at least one perception path in a preset time period; for each first sensing signal, according to the first sensing signal, calculating a multi-dimensional environment characteristic value of a sensing path corresponding to the first sensing signal, the multi-dimensional environment characteristic value being used for representing a multi-dimensional environment change condition of the sensing path in a preset time period; determining at least one second sensing signal in all the first sensing signals according to all the multi-dimensional environment characteristic values; all the second sensing signals are input into a preset life body sensing model, life body sensing information of the sensing path corresponding to each second sensing signal is obtained, and the life body sensing information at least comprises life body type information and / or life body position information. Therefore, by implementing the method, the accuracy and comprehensiveness of life perception can be improved.

Description

Technical Field

[0001] The present invention relates to the field of perception technology, and in particular to a method and device for perceiving a living being. Background Art

[0002] With the continuous development of science and technology, more and more smart devices perceive the surrounding world through a variety of sensors, so as to improve the user experience and convenience of device use. Taking smart cars as an example, by installing radars, cameras and other sensors, users can sense whether there are living things around the vehicle.

[0003] However, practice has found that when smart cars are in unfavorable environmental conditions such as low light, bad weather, etc., the environmental adaptability of related sensors is relatively limited, and their perception and recognition accuracy of living things will decrease. They cannot provide users with accurate perception information, which reduces the user experience and is prone to safety hazards.

[0004] It can be seen that how to improve the accuracy of living body perception is particularly important. Summary of the invention

[0005] The present invention provides a method and device for sensing a living body, which can improve the accuracy of sensing a living body.

[0006] In order to solve the above technical problems, the first aspect of the present invention discloses a method for sensing a living being, the method comprising:

[0007] Collecting a first sensing signal corresponding to at least one sensing path in a preset time period;

[0008] For each of the first perception signals, calculating, according to the first perception signal, a multi-dimensional environmental feature value of the perception path corresponding to the first perception signal, wherein the multi-dimensional environmental feature value is used to represent a change in the multi-dimensional environment of the perception path in the preset time period;

[0009] Determine at least one second perception signal from among all the first perception signals according to all the multi-dimensional environmental feature values;

[0010] All the second perception signals are input into a preset life form perception model to obtain life form perception information of the perception path corresponding to each second perception signal, wherein the life form perception information at least includes life form type information and / or life form position information.

[0011] As an optional implementation manner, in the first aspect of the present invention, the first perception signal includes at least one perception sub-signal, each of the perception sub-signal corresponds to each moment in the preset time period, and for each of the first perception signals, the calculating, according to the first perception signal, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path includes:

[0012] For each of the perception sub-signals of the first perception signal, calculating a sub-signal characteristic value of at least one dimension of the perception sub-signal, the sub-signal characteristic value comprising at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, and a signal overlap rate, the signal overlap rate being used to indicate a transmission overlap between the perception sub-signal and other perception sub-signals of the first perception signal in the preset time period;

[0013] For each dimension, according to all the sub-signal feature values ​​of the first perception signal in the dimension, a signal feature change value of the first perception signal in the dimension is calculated, wherein the signal change feature value is used to represent a feature change of the first perception signal corresponding to the dimension in the preset time period;

[0014] According to the signal feature change values ​​in all the dimensions of the first perception signal, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path is calculated.

[0015] As an optional embodiment, in the first aspect of the present invention, the method further comprises:

[0016] Analyze current scene information according to all the perception paths;

[0017] Acquire historical scene information matching the current scene information;

[0018] Analyzing the historical life activity information of the current scene corresponding to the current scene in the preset time period according to the current scene information and the historical scene information;

[0019] Wherein, the historical life body activity information at least includes historical life body activity area distribution information and historical life body activity frequency information, and the historical life body activity area distribution information corresponds to the historical life body activity frequency information;

[0020] And, calculating the multi-dimensional environmental feature value of the first perception signal corresponding to the perception path according to the signal feature change values ​​in all the dimensions of the first perception signal includes:

[0021] According to the signal feature change values ​​in all the dimensions of the first perception signal, a target priority value of the signal feature change value in each dimension of the first perception signal is matched, and the correlation relationship between the target priority values ​​of the signal feature change values ​​in different dimensions is positively correlated with the correlation relationship between the signal feature change values ​​in the different dimensions;

[0022] According to the historical life activity information, updating each of the target priority values;

[0023] According to the signal characteristic change values ​​in all the dimensions of the first perception signal and all the corresponding updated target priority values, the multi-dimensional environmental characteristic value of the first perception signal corresponding to the perception path is calculated.

[0024] As an optional implementation manner, in the first aspect of the present invention, determining at least one second perception signal from all the first perception signals according to all the multi-dimensional environment feature values ​​includes:

[0025] For each of the first perception signals, a life path prediction value corresponding to the perception path of the first perception signal is calculated according to the multi-dimensional environmental feature value of the first perception signal, wherein the life path prediction value is used to represent the distribution of vital signs existing in the perception path;

[0026] Determine whether the life path prediction value of the first perception signal corresponding to the perception path is greater than or equal to a preset life path prediction threshold. When it is determined that the life path prediction value of the first perception signal corresponding to the perception path is greater than or equal to the preset life path prediction threshold, determine that the first perception signal is a second perception signal.

[0027] As an optional implementation, in the first aspect of the present invention, the preset life form perception model includes an input layer, a convolution layer, a hidden layer, an output layer, and a convolution activation function corresponding to the convolution layer, and all the second perception signals are input into the preset life form perception model to obtain the life form perception information of the perception path corresponding to each second perception signal, including:

[0028] Generate a perception model input matrix according to all the second perception signals and the input layer, wherein the perception model input matrix includes at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, a signal overlap rate, and a sub-signal identifier of each perception sub-signal in each of the second perception signals;

[0029] The perception model input matrix is ​​sequentially input into the convolutional layer, hidden layer and output layer in the preset life form perception model to obtain life form perception information of each second perception signal corresponding to the perception path.

[0030] As an optional embodiment, in the first aspect of the present invention, when the life form perception information includes at least the life form type information and the life form position information, the output layer includes a corresponding classification activation function and a position prediction activation function, the classification activation function is used to generate a predicted probability of the life form category, the life form type information is generated based on the predicted probability, and the position prediction activation function is used to predict the life form position information.

[0031] As an optional implementation, in the first aspect of the present invention, the preset life form perception model further includes a life form perception comprehensive loss function, the life form perception comprehensive loss function includes a classification loss function and a position prediction regression loss function, the classification loss function corresponds to the classification activation function, and the position prediction regression loss function corresponds to the position prediction activation function;

[0032] Furthermore, the preset life form perception model also includes a model training optimizer.

[0033] The second aspect of the present invention discloses a life sensing device, the device comprising:

[0034] A collection module, used to collect a first perception signal corresponding to at least one perception path in a preset time period;

[0035] a calculation module, configured to calculate, for each of the first perception signals, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path according to the first perception signal, wherein the multi-dimensional environmental feature value is used to represent a change in the multi-dimensional environment of the perception path in the preset time period;

[0036] a determination module, configured to determine at least one second perception signal from among all the first perception signals according to all the multi-dimensional environmental feature values;

[0037] The perception module is used to input all the second perception signals into a preset life form perception model to obtain life form perception information corresponding to the perception path of each second perception signal, wherein the life form perception information at least includes life form type information and / or life form position information.

[0038] As an optional implementation manner, in the second aspect of the present invention, the first perception signal includes at least one perception sub-signal, each of the perception sub-signal corresponds to each moment in the preset time period, and for each of the first perception signals, the specific manner in which the calculation module calculates the multi-dimensional environmental feature value of the first perception signal corresponding to the perception path according to the first perception signal includes:

[0039] For each of the perception sub-signals of the first perception signal, calculating a sub-signal characteristic value of at least one dimension of the perception sub-signal, the sub-signal characteristic value comprising at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, and a signal overlap rate, the signal overlap rate being used to indicate a transmission overlap between the perception sub-signal and other perception sub-signals of the first perception signal in the preset time period;

[0040] For each dimension, according to all the sub-signal feature values ​​of the first perception signal in the dimension, a signal feature change value of the first perception signal in the dimension is calculated, wherein the signal change feature value is used to represent a feature change of the first perception signal corresponding to the dimension in the preset time period;

[0041] According to the signal feature change values ​​in all the dimensions of the first perception signal, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path is calculated.

[0042] As an optional implementation, in the second aspect of the present invention, the device further includes:

[0043] An analysis module, used for analyzing current scene information according to all the perception paths;

[0044] A matching module, used to obtain historical scene information that matches the current scene information;

[0045] The analysis module is further used to analyze the historical life activity information of the current scene corresponding to the current scene in the preset time period according to the current scene information and the historical scene information;

[0046] Wherein, the historical life body activity information at least includes historical life body activity area distribution information and historical life body activity frequency information, and the historical life body activity area distribution information corresponds to the historical life body activity frequency information;

[0047] And, the specific manner in which the calculation module calculates the multi-dimensional environmental feature value of the first perception signal corresponding to the perception path according to the signal feature change values ​​in all the dimensions of the first perception signal includes:

[0048] According to the signal feature change values ​​in all the dimensions of the first perception signal, a target priority value of the signal feature change value in each dimension of the first perception signal is matched, and the correlation relationship between the target priority values ​​of the signal feature change values ​​in different dimensions is positively correlated with the correlation relationship between the signal feature change values ​​in the different dimensions;

[0049] According to the historical life activity information, updating each of the target priority values;

[0050] According to the signal characteristic change values ​​in all the dimensions of the first perception signal and all the corresponding updated target priority values, the multi-dimensional environmental characteristic value of the first perception signal corresponding to the perception path is calculated.

[0051] As an optional implementation manner, in the second aspect of the present invention, the specific manner in which the determination module determines at least one second perception signal from all the first perception signals according to all the multi-dimensional environmental feature values ​​includes:

[0052] For each of the first perception signals, a life path prediction value corresponding to the perception path of the first perception signal is calculated according to the multi-dimensional environmental feature value of the first perception signal, wherein the life path prediction value is used to represent the distribution of vital signs existing in the perception path;

[0053] Determine whether the life path prediction value of the first perception signal corresponding to the perception path is greater than or equal to a preset life path prediction threshold. When it is determined that the life path prediction value of the first perception signal corresponding to the perception path is greater than or equal to the preset life path prediction threshold, determine that the first perception signal is a second perception signal.

[0054] As an optional implementation, in the second aspect of the present invention, the preset life form perception model includes an input layer, a convolution layer, a hidden layer, an output layer, and a convolution activation function corresponding to the convolution layer, and the perception module inputs all the second perception signals into the preset life form perception model, and the specific method of obtaining the life form perception information of the perception path corresponding to each second perception signal includes:

[0055] Generate a perception model input matrix according to all the second perception signals and the input layer, wherein the perception model input matrix includes at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, a signal overlap rate, and a sub-signal identifier of each perception sub-signal in each of the second perception signals;

[0056] The perception model input matrix is ​​sequentially input into the convolutional layer, hidden layer and output layer in the preset life form perception model to obtain life form perception information of each second perception signal corresponding to the perception path.

[0057] As an optional embodiment, in the second aspect of the present invention, when the life form perception information includes at least the life form type information and the life form position information, the output layer includes a corresponding classification activation function and a position prediction activation function, the classification activation function is used to generate a predicted probability of the life form category, the life form type information is generated based on the predicted probability, and the position prediction activation function is used to predict the life form position information.

[0058] As an optional implementation, in the second aspect of the present invention, the preset life form perception model also includes a life form perception comprehensive loss function, the life form perception comprehensive loss function includes a classification loss function and a position prediction regression loss function, the classification loss function corresponds to the classification activation function, and the position prediction regression loss function corresponds to the position prediction activation function;

[0059] Furthermore, the preset life form perception model also includes a model training optimizer.

[0060] The third aspect of the present invention discloses another life sensing device, the device comprising:

[0061] A memory storing executable program code;

[0062] a processor coupled to the memory;

[0063] The processor calls the executable program code stored in the memory to execute the living body perception method disclosed in the first aspect of the present invention.

[0064] The fourth aspect of the present invention discloses a computer storage medium, wherein the computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the life form perception method disclosed in the first aspect of the present invention.

[0065] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0066] In an embodiment of the present invention, a first perception signal corresponding to at least one perception path in a preset time period is collected; for each first perception signal, a multi-dimensional environmental characteristic value of the perception path corresponding to the first perception signal is calculated based on the first perception signal, and the multi-dimensional environmental characteristic value is used to represent the multi-dimensional environmental change of the perception path in the preset time period; based on all the multi-dimensional environmental characteristic values, at least one second perception signal is determined from all the first perception signals; all the second perception signals are input into a preset life form perception model to obtain life form perception information of the perception path corresponding to each second perception signal, and the life form perception information includes at least life form type information and / or life form position information. It can be seen that the implementation of the present invention can calculate the multi-dimensional environmental characteristic value of each first perception signal for representing the multi-dimensional environmental change of its corresponding perception path in the preset time period according to the first perception signal corresponding to at least one perception path collected in the preset time period, and can perceive life forms on multiple perception paths and improve the accuracy of life form perception. Through a comprehensive analysis of multi-dimensional environmental changes, the accuracy, comprehensiveness and scientificity of life form perception can be further improved, and the environmental adaptability of life form perception can be improved. According to all multi-dimensional environmental characteristic values, at least one second perception signal can be determined from all first perception signals, which can further help reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and all second The perception signal is input into the preset life form perception model to obtain life form perception information of at least life form type information and / or life form position information corresponding to the perception path of each second perception signal. This can further improve the accuracy, scientificity and operability of life form perception based on the actual preset life form perception model. By generating life form perception information of at least life form type information and / or life form position information corresponding to the perception path of each second perception signal, the accuracy and comprehensiveness of life form perception can be further improved. Specifically, the analysis accuracy of life form type information and the prediction accuracy of life form position information can be improved, which is beneficial to improving the user's application equipment usage safety while improving the user's application equipment usage experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0068] Figure 1 It is a flow chart of a method for sensing a living being disclosed in an embodiment of the present invention;

[0069] Figure 2is a flow chart of another method for sensing a living being disclosed in an embodiment of the present invention;

[0070] Figure 3 It is a structural schematic diagram of a life sensing device disclosed in an embodiment of the present invention;

[0071] Figure 4 is a schematic diagram of the structure of another life sensing device disclosed in an embodiment of the present invention;

[0072] Figure 5 It is a structural schematic diagram of another life sensing device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0073] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0075] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0076] The present invention discloses a method and device for sensing a living being, which can calculate a multi-dimensional environmental characteristic value of each first sensing signal for indicating a multi-dimensional environmental change of the corresponding sensing path in a preset time period according to a first sensing signal corresponding to at least one sensing path collected in a preset time period, can sense a living being in multiple sensing paths, improve the accuracy of the living being's perception, and further improve the accuracy, comprehensiveness and scientificity of the living being's perception through a comprehensive analysis of the multi-dimensional environmental change, and improve the environmental adaptability of the living being's perception, and can determine at least one second sensing signal from all the first sensing signals according to all the multi-dimensional environmental characteristic values, which can further help reduce the multipath effect and improve the accuracy and scientificity of determining the second sensing signal. Input all second perception signals into the preset life form perception model to obtain life form perception information that at least includes life form type information and / or life form position information for the perception path corresponding to each second perception signal. This can further improve the accuracy, scientificity and operability of life form perception based on the actual preset life form perception model. By generating life form perception information that at least includes life form type information and / or life form position information for the perception path corresponding to each second perception signal, the accuracy and comprehensiveness of life form perception can be further improved. Specifically, the analysis accuracy of life form type information and the prediction accuracy of life form position information can be improved, which is beneficial to improving the user's application equipment usage experience while improving the user's application equipment usage safety. The following are detailed descriptions.

[0077] Embodiment 1

[0078] See also Figure 1 , Figure 1 : is a flow chart of a method for sensing a living being disclosed in an embodiment of the present invention. Figure 1 The described method for sensing life forms can be applied to sensing devices, and can also be applied to intelligent devices related to sensing devices, such as but not limited to one or more of cloud devices, edge computing devices, relay devices, base station devices, urban management devices, intelligent network devices, and smart home devices, which are not limited in the embodiments of the present invention. Figure 1 As shown, the method for sensing a living being may include the following operations:

[0079] 101. Collect a first sensing signal corresponding to at least one sensing path in a preset time period;

[0080] In an embodiment of the present invention, optionally, the above-mentioned first perception signal may include but is not limited to electromagnetic wave signals such as Wifi signals and CSI signals. Further optionally, the above-mentioned first perception signal may also include a communication perception fusion frame structure, and the above-mentioned communication perception fusion frame structure is related to the above-mentioned perception path and a specific moment in the preset time period.

[0081] 102. For each first perception signal, calculate, according to the first perception signal, a multi-dimensional environmental feature value of a perception path corresponding to the first perception signal, where the multi-dimensional environmental feature value is used to represent a multi-dimensional environmental change of the perception path in a preset time period;

[0082] In an embodiment of the present invention, as an optional implementation manner, the first perception signal includes at least one perception sub-signal, each perception sub-signal corresponds to each moment in a preset time period, and for each first perception signal, calculating, according to the first perception signal, a multi-dimensional environmental feature value of a perception path corresponding to the first perception signal, including:

[0083] For each perception sub-signal of the first perception signal, calculating a sub-signal characteristic value of at least one dimension of the perception sub-signal, where the sub-signal characteristic value includes at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, and a signal overlap rate, where the signal overlap rate is used to indicate a transmission overlap between the perception sub-signal and other perception sub-signals of other first perception signals in a preset time period;

[0084] For each dimension, according to all sub-signal feature values ​​of the first perception signal in the dimension, a signal feature change value of the first perception signal in the dimension is calculated, where the signal change feature value is used to indicate a feature change of the first perception signal in the corresponding dimension in a preset time period;

[0085] According to the signal feature change values ​​in all dimensions of the first perception signal, a multi-dimensional environmental feature value of a perception path corresponding to the first perception signal is calculated.

[0086] In this optional embodiment, optionally, the above-mentioned moment usually refers to a specific time point, which may include different time precisions, that is, the time unit is not specifically limited, such as seconds, microseconds, milliseconds and other time units.

[0087] Further optionally, the above-mentioned perception sub-signal can be specifically in a complex form, such as z=a+bi, where a is used to represent the real part and b is used to represent the imaginary part. Optionally, the amplitude of the above-mentioned sub-signal can be calculated by a first formula, and the first formula is:

[0088]

[0089] Further optionally, the above sub-signal phase value can be calculated by a second formula, and the second formula is:

[0090] θ=arctan2(b,a)

[0091] Further optionally, the above-mentioned sub-signal frequency value, sub-signal wavelength value, and sub-signal radiation value can be obtained by converting the above-mentioned perceived sub-signal complex form into trigonometric functions and the like. It should be noted that the above-mentioned sub-signal characteristic values ​​are not calculated during the actual application process and can be empty and do not participate in subsequent calculations.

[0092] It can be seen that the implementation of this optional embodiment can further improve the analysis granularity for the first perception signal corresponding to each perception path in the preset time period, and calculate the sub-signal characteristic value of at least one dimension of each perception sub-signal from the level of the perception sub-signal at each moment in the corresponding preset time period included in each first perception signal, specifically including at least one of the sub-signal amplitude, sub-signal phase value, sub-signal frequency value, sub-signal radiation value, sub-signal wavelength value, and signal overlap rate, so as to further improve the dimensional richness and comprehensiveness of the analysis of the life body perception data on the basis of improving the granularity of the analysis of the life body perception data, so that for each dimension, according to all the sub-signal characteristic values ​​of each first perception signal in the dimension, the signal characteristic change value of each first perception signal in the dimension used to represent the characteristic change of the corresponding dimension of the first perception signal in the preset time period is calculated, and the comprehensiveness of the analysis of the signal characteristic change of the first perception signal in the preset time period is improved, thereby improving the calculation accuracy, scientificity and feasibility of the multi-dimensional environmental characteristic value of each first perception signal used to represent the multi-dimensional environmental change of its corresponding perception path in the preset time period, which is conducive to improving the adaptability of environmental perception and the accuracy of life body perception.

[0093] In this optional embodiment, as an optional implementation, the method may further include:

[0094] Analyze the current scene information based on all perception paths;

[0095] Obtain historical scene information that matches the current scene information;

[0096] According to the current scene information and the historical scene information, the historical life activity information of the current scene corresponding to the current scene in the preset time period is analyzed;

[0097] The historical life form activity information at least includes the historical life form activity regional distribution information and the historical life form activity frequency information, and the historical life form activity regional distribution information corresponds to the historical life form activity frequency information;

[0098] Optionally, the above-mentioned current scene information may include but is not limited to at least one of a reading scene, a sleeping scene, a cooking scene, a sports scene, a washing scene, a driving scene, and a working scene.

[0099] Optionally, the calculating, according to the signal feature change values ​​in all dimensions of the first perception signal, the multi-dimensional environmental feature value of the perception path corresponding to the first perception signal includes:

[0100] According to the signal feature change values ​​in all dimensions of the first perception signal, a target priority value of the signal feature change value in each dimension of the first perception signal is matched, and the correlation relationship between the target priority values ​​of the signal feature change values ​​in different dimensions is positively correlated with the correlation relationship between the signal feature change values ​​in different dimensions;

[0101] Update the priority value of each target based on historical life activity information;

[0102] According to the signal characteristic change values ​​in all dimensions of the first perception signal and all corresponding updated target priority values, the multi-dimensional environmental characteristic value of the perception path corresponding to the first perception signal is calculated.

[0103] It should be noted that, in this optional embodiment, the above-mentioned "matching the target priority value of the signal feature change value in each dimension of the first perception signal according to the signal feature change value in all dimensions of the first perception signal, and the correlation between the target priority values ​​of the signal feature change values ​​in different dimensions is positively correlated with the correlation between the signal feature change values ​​in different dimensions;" can be performed separately. After that, the corresponding operation of "calculating the multi-dimensional environmental feature value of the perception path corresponding to the first perception signal according to the signal feature change values ​​in all dimensions of the first perception signal and all corresponding target priority values" can be directly executed. At this time, it is used to represent the scenario when the historical life activity information is empty, that is, the multi-dimensional environmental feature value can be generated without the above-mentioned historical life activity information. The specific operation shall be based on the actual application scenario needs and capabilities.

[0104] In this optional embodiment, it should be noted that the correlation relationship between the target priority values ​​of the signal feature change values ​​under different dimensions is positively correlated with the correlation relationship between the signal feature change values ​​under different dimensions, which is used to indicate that if the sub-signal feature value under the corresponding dimension reflected by a certain signal feature change value is larger than the sub-signal feature values ​​corresponding to other signal feature change values, then the target priority value corresponding to the certain signal feature change value is higher than the target priority values ​​of other signal feature change values. For example, when the above-mentioned sub-signal feature value only includes the sub-signal amplitude and the sub-signal phase value, if the target priority value of the signal feature change value corresponding to the sub-signal amplitude is It is used to represent the square of the signal characteristic change value corresponding to the sub-signal amplitude. When the target priority value of the signal characteristic change value corresponding to the sub-signal phase value is 1, it means that by comparing the signal characteristic change value of the first perception signal in the dimension of sub-signal amplitude and the signal characteristic change value in the dimension of sub-signal phase value, it is reflected that the sub-signal amplitude change of the first perception signal in the preset time period is smaller than the sub-signal phase value.

[0105] It can be seen that the implementation of this optional embodiment can match the target priority value of the signal feature change value of each first perception signal in each dimension by comparing the correlation between the signal feature change values ​​in different dimensions after obtaining the signal feature change values ​​in all dimensions of each first perception signal, which can be beneficial to improve the calculation accuracy and scientificity of the multi-dimensional environmental feature value of the perception path corresponding to each first perception signal, thereby improving the accuracy of life perception. On this basis, by analyzing the current scene information according to all perception paths, obtaining historical scene information matching the current scene information, and analyzing the historical life activity information corresponding to the current scene in a preset time period according to the current scene information and the historical scene information, the historical life activity information including at least the historical life activity area distribution information and the historical life activity frequency information can be further combined with the historical activity situation of the corresponding application scene to improve the matching accuracy of the target priority value, thereby further improving the calculation accuracy and scientificity of the multi-dimensional environmental feature value of the perception path corresponding to each first perception signal, improving the accuracy of life perception, fitting the application reality, and reducing the possibility of incorrect analysis of multi-dimensional environmental feature values ​​due to unexpected conditions.

[0106] 103. Determine at least one second perception signal from all first perception signals according to all multi-dimensional environmental feature values;

[0107] In the embodiment of the present invention, as another optional implementation manner, the above-mentioned determining at least one second perception signal from all first perception signals according to all multi-dimensional environmental feature values ​​includes:

[0108] For each first perception signal, a life path prediction value of a perception path corresponding to the first perception signal is calculated according to the multi-dimensional environmental feature value of the first perception signal, where the life path prediction value is used to represent the distribution of vital signs existing in the perception path;

[0109] Determine whether the life path prediction value of the perception path corresponding to the first perception signal is greater than or equal to a preset life path prediction threshold. When it is determined that the life path prediction value of the perception path corresponding to the first perception signal is greater than or equal to the preset life path prediction threshold, determine that the first perception signal is a second perception signal.

[0110] In this optional embodiment, optionally, the above-mentioned life path prediction value can be calculated by a third formula, and the third formula is:

[0111]

[0112] Wherein, δ is used to represent the preset first weight parameter, φ is used to represent the preset second weight parameter, and χ is used to represent the multi-dimensional environmental characteristic value. Used to indicate the preset first bias parameter.

[0113] It can be seen that the implementation of this optional embodiment can, after calculating all the multi-dimensional environmental feature values, for each first perception signal, calculate the life path prediction value of the first perception signal based on the multi-dimensional environmental feature value of the first perception signal, which is used to represent the distribution of vital signs of the life form in the corresponding perception path, and when it is judged that the life path prediction value of the perception path corresponding to the first perception signal is greater than or equal to the preset life path prediction threshold, the first perception signal is determined to be the second perception signal, which can further help reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and improve the accuracy and scientificity of life perception.

[0114] 104. Input all second perception signals into a preset life form perception model to obtain life form perception information of a perception path corresponding to each second perception signal, where the life form perception information at least includes life form type information and / or life form position information.

[0115] It can be seen that the implementation of the embodiment of the present invention can calculate the multi-dimensional environmental feature value of each first perception signal for representing the multi-dimensional environmental change of its corresponding perception path in the preset time period according to the first perception signal corresponding to at least one perception path collected in the preset time period, and can perceive life forms on multiple perception paths and improve the accuracy of life form perception. Through a comprehensive analysis of multi-dimensional environmental changes, the accuracy, comprehensiveness and scientificity of life form perception can be further improved, and the environmental adaptability of life form perception can be improved. According to all multi-dimensional environmental feature values, at least one second perception signal can be determined from all first perception signals, which can further help reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and can further reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and improve all the first perception signals. The two perception signals are input into the preset life form perception model to obtain life form perception information of at least life form type information and / or life form position information corresponding to the perception path of each second perception signal. This can further improve the accuracy, scientificity and operability of life form perception based on the actual preset life form perception model. By generating life form perception information of at least life form type information and / or life form position information corresponding to the perception path of each second perception signal, the accuracy and comprehensiveness of life form perception can be further improved. Specifically, the analysis accuracy of life form type information and the prediction accuracy of life form position information can be improved, which is conducive to improving the user's application equipment usage safety while improving the user's application equipment usage experience.

[0116] Embodiment 2

[0117] See also Figure 2 , Figure 2 is a flow chart of another method for sensing a living being disclosed in an embodiment of the present invention. Figure 2 The described method for sensing life forms can be applied to sensing devices, and can also be applied to intelligent devices related to sensing devices, such as but not limited to one or more of cloud devices, edge computing devices, relay devices, base station devices, urban management devices, intelligent network devices, and smart home devices, which are not limited in the embodiments of the present invention. Figure 2 As shown, the method for sensing a living being may include the following operations:

[0118] 201. Collect a first sensing signal corresponding to at least one sensing path in a preset time period;

[0119] 202. For each first perception signal, calculate, according to the first perception signal, a multi-dimensional environmental feature value of a perception path corresponding to the first perception signal, where the multi-dimensional environmental feature value is used to represent a change in the multi-dimensional environment of the perception path in a preset time period;

[0120] 203. Determine at least one second perception signal from all first perception signals according to all multi-dimensional environment feature values;

[0121] In the embodiment of the present invention, for the supplementary description of step 201 to step 203, please refer to the supplementary description of step 101 to step 103 in the first embodiment, and the embodiment of the present invention will not elaborate on this.

[0122] 204. Generate a perception model input matrix according to all second perception signals and the input layer included in the preset life form perception model, wherein the perception model input matrix includes at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, a signal overlap rate, and a sub-signal identifier of each perception sub-signal in each second perception signal;

[0123] 205. Input the perception model input matrix into the convolution layer, hidden layer and output layer of the preset life form perception model in sequence to obtain life form perception information of the perception path corresponding to each second perception signal, where the life form perception information includes at least life form type information and / or life form position information.

[0124] It can be seen that the implementation of the embodiment of the present invention discloses a specific structure of a preset life form perception model, so that after determining all the second perception signals, based on all the second perception signals and the input layer included in the preset life form perception model, a perception model input matrix is ​​generated, which includes at least one of the sub-signal amplitude, sub-signal phase value, sub-signal frequency value, sub-signal radiation value, sub-signal wavelength value, signal overlap rate, and sub-signal identification of each perception sub-signal in each second perception signal, so as to further improve the perception analysis accuracy and comprehensiveness of the preset life form perception model, and then the matrix is ​​sequentially input into the convolutional layer, hidden layer and output layer also included in the preset life form perception model to obtain the life form perception information of the perception path corresponding to each second perception signal, which includes at least life form type information and / or life form position information, further improving the accuracy and comprehensiveness of life form perception, specifically, improving the analysis accuracy of life form type information and the prediction accuracy of life form position information, which is beneficial to improving the user's application device usage experience while improving the user's application device usage safety.

[0125] In an embodiment of the present invention, as an optional implementation, the above-mentioned preset life form perception model further includes a convolution activation function, the convolution activation function corresponds to the convolution layer, and the convolution activation function is:

[0126] X l =ReLU(W l *X l-1 +b l ), l∈N *

[0127] Among them, l is used to represent the number of convolutional layers, X l It is used to represent the output result of the convolution activation function of the lth convolutional layer, W l Used to represent the weight of the convolution kernel in the lth convolution layer, X l-1 It is used to represent the result of the convolution activation function output of the l-1th convolutional layer. When l = 1, X l-1 Used to represent the perception model input matrix, b l Used to represent the bias of the convolution kernel in the lth convolution layer.

[0128] It can be seen that the implementation of this optional embodiment also discloses the structure of the convolution layer and the content of the convolution activation function included in the preset life form perception model to improve the feature extraction accuracy of the perception model input matrix and the life form perception accuracy, feasibility and scientificity of the preset life form perception model.

[0129] In an embodiment of the present invention, as an optional implementation, when the above-mentioned life body perception information includes at least life body type information and life body position information, the output layer includes a corresponding classification activation function and a position prediction activation function, and the classification activation function is:

[0130]

[0131] in, It is used to represent the predicted probability of the cth category. The life type information is generated based on the predicted probability. Z is used to represent the result of the hidden layer output;

[0132] And, the position prediction activation function is:

[0133]

[0134] in, Used to represent the predicted location information of the living being, W xyz Used to represent preset linear weights, b xyz Used to represent the preset bias vector.

[0135] In this optional embodiment, optionally, a plurality of fully connected hidden layers may be included between the above-mentioned convolutional layer and the output layer to improve the nonlinear expression capability of the preset life form perception model.

[0136] It can be seen that the implementation of this optional embodiment further discloses the output layer structure of the preset life form perception model, specifically including the specific contents of the classification activation function for analyzing the life form type information and the position prediction activation function for analyzing the life form position information. While improving the accuracy and comprehensiveness of life form perception, specifically, improving the analysis accuracy of life form type information and the prediction accuracy of life form position information, it also improves the feasibility and scientificity of the analysis and prediction of life form type information and life form position information, which is beneficial to improving the user's application device usage experience while improving the user's application device usage safety.

[0137] In this optional embodiment, as an optional implementation, the above-mentioned preset life body perception model also includes a life body perception comprehensive loss function, which includes a classification loss function and a position prediction regression loss function. The classification loss function corresponds to the classification activation function, and the position prediction regression loss function corresponds to the position prediction activation function, wherein the life body perception comprehensive loss function is:

[0138] L 综合 =L1+λL2

[0139] Among them, λ is used to represent the preset weight coefficient;

[0140] Optionally, the classification loss function is:

[0141]

[0142] in, Used to represent the predicted probability of the cth category, y c Used to indicate The corresponding true label value;

[0143] Optionally, the position prediction regression loss function is:

[0144]

[0145] Among them, (x, y, z) is used to represent The corresponding real location information;

[0146] Optionally, the preset life form perception model also includes a model training optimizer, and the update rule of the model training optimizer is:

[0147] m t =β1m t-1 +(1-β1)g t

[0148]

[0149] Among them, m tUsed to represent the first-order moment estimate at time t, m t-1 Used to represent the first-order moment estimate at time t-1, g t It is used to represent the gradient at time t, β1 is used to represent the decay rate of the first-order moment estimate, and v t Used to represent the second-order moment estimate at time t, v t-1 It is used to represent the second-order moment estimate at time t-1. It is used to represent the square of the gradient at time t, β2 is used to represent the decay rate of the second-order moment estimate, and θ t+1 Used to represent the updated parameters, θ t It is used to represent the parameters before updating, η is used to represent the learning rate, and ε is used to represent the preset constant.

[0150] It can be seen that the implementation of this optional embodiment further discloses the preset life form perception model including a comprehensive loss function of life form perception, which is a comprehensive loss function that integrates the classification loss function and the position prediction regression loss function, and a model training optimizer, which can improve the feasibility and scientificity of the training of the preset life form perception model and improve the life form perception accuracy of the preset life form perception model.

[0151] Embodiment 3

[0152] See also Figure 3 , Figure 3 : is a schematic diagram of the structure of a life-sensing device disclosed in an embodiment of the present invention. The life-sensing device can be applied to a sensing device, and can also be applied to an intelligent device related to the sensing device, which includes but is not limited to one or more of cloud devices, edge computing devices, relay devices, base station devices, urban management devices, intelligent network devices, and smart home devices, which are not limited in the embodiment of the present invention. Figure 3 As shown, the life sensing device may include:

[0153] The collection module 301 is used to collect a first sensing signal corresponding to at least one sensing path in a preset time period;

[0154] A calculation module 302 is used to calculate, for each first perception signal, a multi-dimensional environmental feature value of a perception path corresponding to the first perception signal according to the first perception signal, where the multi-dimensional environmental feature value is used to represent a multi-dimensional environmental change of the perception path in a preset time period;

[0155] A determination module 303 is used to determine at least one second perception signal from all first perception signals according to all multi-dimensional environmental feature values;

[0156] The perception module 304 is used to input all second perception signals into a preset life form perception model to obtain life form perception information of a perception path corresponding to each second perception signal, where the life form perception information at least includes life form type information and / or life form position information.

[0157] It can be seen that the implementation of the embodiment of the present invention can calculate the multi-dimensional environmental feature value of each first perception signal for representing the multi-dimensional environmental change of its corresponding perception path in the preset time period according to the first perception signal corresponding to at least one perception path collected in the preset time period, and can perceive life forms on multiple perception paths and improve the accuracy of life form perception. Through a comprehensive analysis of multi-dimensional environmental changes, the accuracy, comprehensiveness and scientificity of life form perception can be further improved, and the environmental adaptability of life form perception can be improved. According to all multi-dimensional environmental feature values, at least one second perception signal can be determined from all first perception signals, which can further help reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and can further reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and improve all the first perception signals. The two perception signals are input into the preset life form perception model to obtain life form perception information of at least life form type information and / or life form position information corresponding to the perception path of each second perception signal. This can further improve the accuracy, scientificity and operability of life form perception based on the actual preset life form perception model. By generating life form perception information of at least life form type information and / or life form position information corresponding to the perception path of each second perception signal, the accuracy and comprehensiveness of life form perception can be further improved. Specifically, the analysis accuracy of life form type information and the prediction accuracy of life form position information can be improved, which is conducive to improving the user's application equipment usage safety while improving the user's application equipment usage experience.

[0158] In the embodiment of the present invention, as an optional implementation manner, the first perception signal includes at least one perception sub-signal, each perception sub-signal corresponds to each moment in a preset time period, and for each first perception signal, the calculation module 302 calculates the multi-dimensional environmental feature value of the perception path corresponding to the first perception signal according to the first perception signal. The specific manner includes:

[0159] For each perception sub-signal of the first perception signal, calculating a sub-signal characteristic value of at least one dimension of the perception sub-signal, where the sub-signal characteristic value includes at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, and a signal overlap rate, where the signal overlap rate is used to indicate a transmission overlap between the perception sub-signal and other perception sub-signals of other first perception signals in a preset time period;

[0160] For each dimension, according to all sub-signal feature values ​​of the first perception signal in the dimension, a signal feature change value of the first perception signal in the dimension is calculated, where the signal change feature value is used to indicate a feature change of the first perception signal in the corresponding dimension in a preset time period;

[0161] According to the signal feature change values ​​in all dimensions of the first perception signal, a multi-dimensional environmental feature value of a perception path corresponding to the first perception signal is calculated.

[0162] It can be seen that the implementation of this optional embodiment can further improve the analysis granularity for the first perception signal corresponding to each perception path in the preset time period, and calculate the sub-signal characteristic value of at least one dimension of each perception sub-signal from the level of the perception sub-signal at each moment in the corresponding preset time period included in each first perception signal, specifically including at least one of the sub-signal amplitude, sub-signal phase value, sub-signal frequency value, sub-signal radiation value, sub-signal wavelength value, and signal overlap rate, so as to further improve the dimensional richness and comprehensiveness of the analysis of the life body perception data on the basis of improving the granularity of the analysis of the life body perception data, so that for each dimension, according to all the sub-signal characteristic values ​​of each first perception signal in the dimension, the signal characteristic change value of each first perception signal in the dimension used to represent the characteristic change of the corresponding dimension of the first perception signal in the preset time period is calculated, and the comprehensiveness of the analysis of the signal characteristic change of the first perception signal in the preset time period is improved, thereby improving the calculation accuracy, scientificity and feasibility of the multi-dimensional environmental characteristic value of each first perception signal used to represent the multi-dimensional environmental change of its corresponding perception path in the preset time period, which is conducive to improving the adaptability of environmental perception and the accuracy of life body perception.

[0163] In the embodiment of the present invention, as another optional implementation, Figure 4 As shown, the device may also include:

[0164] An analysis module 305 is used to analyze current scene information according to all perception paths;

[0165] A matching module 306 is used to obtain historical scene information that matches the current scene information;

[0166] The analysis module 305 is further used to analyze the historical life activity information of the current scene information corresponding to the current scene in a preset time period according to the current scene information and the historical scene information;

[0167] The historical life form activity information at least includes the historical life form activity regional distribution information and the historical life form activity frequency information, and the historical life form activity regional distribution information corresponds to the historical life form activity frequency information;

[0168] Optionally, the specific manner in which the calculation module 302 calculates the multi-dimensional environmental feature value of the perception path corresponding to the first perception signal according to the signal feature change values ​​in all dimensions of the first perception signal includes:

[0169] According to the signal feature change values ​​in all dimensions of the first perception signal, a target priority value of the signal feature change value in each dimension of the first perception signal is matched, and the correlation relationship between the target priority values ​​of the signal feature change values ​​in different dimensions is positively correlated with the correlation relationship between the signal feature change values ​​in different dimensions;

[0170] Update the priority value of each target based on historical life activity information;

[0171] According to the signal characteristic change values ​​in all dimensions of the first perception signal and all corresponding updated target priority values, the multi-dimensional environmental characteristic value of the perception path corresponding to the first perception signal is calculated.

[0172] It can be seen that the implementation of this optional embodiment can match the target priority value of the signal feature change value of each first perception signal in each dimension by comparing the correlation between the signal feature change values ​​in different dimensions after obtaining the signal feature change values ​​in all dimensions of each first perception signal, which can be beneficial to improve the calculation accuracy and scientificity of the multi-dimensional environmental feature value of the perception path corresponding to each first perception signal, thereby improving the accuracy of life perception. On this basis, by analyzing the current scene information according to all perception paths, obtaining historical scene information matching the current scene information, and analyzing the historical life activity information corresponding to the current scene in a preset time period according to the current scene information and the historical scene information, the historical life activity information including at least the historical life activity area distribution information and the historical life activity frequency information can be further combined with the historical activity situation of the corresponding application scene to improve the matching accuracy of the target priority value, thereby further improving the calculation accuracy and scientificity of the multi-dimensional environmental feature value of the perception path corresponding to each first perception signal, improving the accuracy of life perception, fitting the application reality, and reducing the possibility of incorrect analysis of multi-dimensional environmental feature values ​​due to unexpected conditions.

[0173] In the embodiment of the present invention, as another optional implementation manner, the specific manner in which the determination module 303 determines at least one second perception signal from among all first perception signals according to all multi-dimensional environmental feature values ​​includes:

[0174] For each first perception signal, a life path prediction value of a perception path corresponding to the first perception signal is calculated according to the multi-dimensional environmental feature value of the first perception signal, where the life path prediction value is used to represent the distribution of vital signs existing in the perception path;

[0175] Determine whether the life path prediction value of the perception path corresponding to the first perception signal is greater than or equal to a preset life path prediction threshold. When it is determined that the life path prediction value of the perception path corresponding to the first perception signal is greater than or equal to the preset life path prediction threshold, determine that the first perception signal is a second perception signal.

[0176] It can be seen that the implementation of this optional embodiment can, after calculating all the multi-dimensional environmental feature values, for each first perception signal, calculate the life path prediction value of the first perception signal based on the multi-dimensional environmental feature value of the first perception signal, which is used to represent the distribution of vital signs of the life form in the corresponding perception path, and when it is judged that the life path prediction value of the perception path corresponding to the first perception signal is greater than or equal to the preset life path prediction threshold, the first perception signal is determined to be the second perception signal, which can further help reduce the multipath effect, improve the accuracy and scientificity of determining the second perception signal, and improve the accuracy and scientificity of life perception.

[0177] In the embodiment of the present invention, as an optional implementation, the preset life form perception model includes an input layer, a convolution layer, a hidden layer and an output layer. The perception module 304 inputs all second perception signals into the preset life form perception model, and the specific method of obtaining the life form perception information of the perception path corresponding to each second perception signal includes:

[0178] Generate a perception model input matrix according to all second perception signals and the input layer, wherein the perception model input matrix includes at least one of a sub-signal amplitude value, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, a signal overlap rate, and a sub-signal identifier of each perception sub-signal in each second perception signal;

[0179] Inputting the perception model input matrix into the convolution layer, hidden layer and output layer of the preset life form perception model in sequence, obtaining life form perception information of the perception path corresponding to each second perception signal;

[0180] It can be seen that the implementation of the embodiment of the present invention discloses a specific structure of a preset life form perception model, so that after determining all the second perception signals, based on all the second perception signals and the input layer included in the preset life form perception model, a perception model input matrix is ​​generated, which includes at least one of the sub-signal amplitude, sub-signal phase value, sub-signal frequency value, sub-signal radiation value, sub-signal wavelength value, signal overlap rate, and sub-signal identification of each perception sub-signal in each second perception signal, so as to further improve the perception analysis accuracy and comprehensiveness of the preset life form perception model, and then the matrix is ​​sequentially input into the convolutional layer, hidden layer and output layer also included in the preset life form perception model to obtain the life form perception information of the perception path corresponding to each second perception signal, which includes at least life form type information and / or life form position information, further improving the accuracy and comprehensiveness of life form perception, specifically, improving the analysis accuracy of life form type information and the prediction accuracy of life form position information, which is beneficial to improving the user's application device usage experience while improving the user's application device usage safety.

[0181] Optionally, the preset life form perception model further includes a convolution activation function, which corresponds to the convolution layer and is:

[0182] X l =ReLU(W l *X l-1 +b l ), l∈N *

[0183] Among them, l is used to represent the number of convolutional layers, X l It is used to represent the output result of the convolution activation function of the lth convolutional layer, W l Used to represent the weight of the convolution kernel in the lth convolution layer, X l-1 It is used to represent the result of the convolution activation function output of the l-1th convolutional layer. When l = 1, X l-1 Used to represent the perception model input matrix, b l Used to represent the bias of the convolution kernel in the lth convolution layer.

[0184] It can be seen that the implementation of this optional embodiment also discloses the structure of the convolution layer and the content of the convolution activation function included in the preset life form perception model to improve the feature extraction accuracy of the perception model input matrix and the life form perception accuracy, feasibility and scientificity of the preset life form perception model.

[0185] In this optional embodiment, as an optional implementation, when the life body perception information includes at least life body type information and life body position information, the output layer includes a corresponding classification activation function and a position prediction activation function, and the classification activation function is:

[0186]

[0187] in, It is used to represent the predicted probability of the cth category. The life type information is generated based on the predicted probability. Z is used to represent the result of the hidden layer output;

[0188] Optionally, the position prediction activation function is:

[0189]

[0190] in, Used to represent the predicted location information of the living being, W xyz Used to represent preset linear weights, b xyz Used to represent the preset bias vector.

[0191] It can be seen that the implementation of this optional embodiment further discloses the output layer structure of the preset life form perception model, specifically including the specific contents of the classification activation function for analyzing the life form type information and the position prediction activation function for analyzing the life form position information. While improving the accuracy and comprehensiveness of life form perception, specifically, improving the analysis accuracy of life form type information and the prediction accuracy of life form position information, it also improves the feasibility and scientificity of the analysis and prediction of life form type information and life form position information, which is beneficial to improving the user's application device usage experience while improving the user's application device usage safety.

[0192] In this optional embodiment, as another optional implementation, the above-mentioned preset life body perception model also includes a life body perception comprehensive loss function, and the life body perception comprehensive loss function includes a classification loss function and a position prediction regression loss function. The classification loss function corresponds to the classification activation function, and the position prediction regression loss function corresponds to the position prediction activation function. Optionally, the life body perception comprehensive loss function is:

[0193] L 综合 =L1+λL2

[0194] Among them, λ is used to represent the preset weight coefficient;

[0195] Optionally, the classification loss function is:

[0196]

[0197] in, Used to represent the predicted probability of the cth category, y c Used to indicate The corresponding true label value;

[0198] Optionally, the position prediction regression loss function is:

[0199]

[0200] Among them, (x, y, z) is used to represent The corresponding real location information;

[0201] Optionally, the preset life form perception model also includes a model training optimizer, and the update rule of the model training optimizer is:

[0202] m t =β1m t-1 +(1-β1)g t

[0203]

[0204] Among them, m t Used to represent the first-order moment estimate at time t, m t-1 Used to represent the first-order moment estimate at time t-1, g t It is used to represent the gradient at time t, β1 is used to represent the decay rate of the first-order moment estimate, and v t Used to represent the second-order moment estimate at time t, v t-1 It is used to represent the second-order moment estimate at time t-1. It is used to represent the square of the gradient at time t, β2 is used to represent the decay rate of the second-order moment estimate, and θ t+1 Used to represent the updated parameters, θ t It is used to represent the parameters before updating, η is used to represent the learning rate, and ε is used to represent the preset constant.

[0205] It can be seen that the implementation of this optional embodiment further discloses the preset life form perception model including a comprehensive loss function of life form perception, which is a comprehensive loss function that integrates the classification loss function and the position prediction regression loss function, and a model training optimizer, which can improve the feasibility and scientificity of the training of the preset life form perception model and improve the life form perception accuracy of the preset life form perception model.

[0206] Embodiment 4

[0207] See also Figure 5 , Figure 5 This is a schematic diagram of the structure of another life sensing device disclosed in an embodiment of the present invention. The life sensing device can be applied to sensing devices, and can also be applied to smart devices related to sensing devices, including but not limited to cloud devices, edge computing devices, relay devices, base station devices, city management devices, smart network devices, smart home devices, etc., which are not limited in the embodiment of the present invention. Figure 5 As shown, the life sensing device may include:

[0208] The memory 401 stores executable program codes.

[0209] A processor 402 is coupled to the memory 401 .

[0210] The processor 402 calls the executable program code stored in the memory 401 to execute the steps of the life body perception method described in the first embodiment of the present invention or the second embodiment of the present invention.

[0211] Embodiment 5

[0212] An embodiment of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the steps in the life form perception method described in Embodiment 1 or Embodiment 2 of the present invention.

[0213] Embodiment 6

[0214] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps in the life form perception method described in Embodiment 1 or Embodiment 2.

[0215] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0216] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0217] Finally, it should be noted that the method and device for sensing a living being disclosed in the embodiments of the present invention disclose only the preferred embodiments of the present invention, and are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for sensing a living being, characterized in that: The method comprises: Collecting a first sensing signal corresponding to at least one sensing path in a preset time period; For each of the first perception signals, calculating, according to the first perception signal, a multi-dimensional environmental feature value of the perception path corresponding to the first perception signal, wherein the multi-dimensional environmental feature value is used to represent a change in the multi-dimensional environment of the perception path in the preset time period; Determine at least one second perception signal from among all the first perception signals according to all the multi-dimensional environmental feature values; All the second perception signals are input into a preset life form perception model to obtain life form perception information of the perception path corresponding to each second perception signal, wherein the life form perception information at least includes life form type information and / or life form position information.

2. The method for sensing a living being according to claim 1, characterized in that: The first perception signal includes at least one perception sub-signal, each of the perception sub-signals corresponds to each moment in the preset time period, and for each of the first perception signals, calculating, according to the first perception signal, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path, includes: For each of the perception sub-signals of the first perception signal, calculating a sub-signal characteristic value of at least one dimension of the perception sub-signal, the sub-signal characteristic value comprising at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, and a signal overlap rate, the signal overlap rate being used to indicate a transmission overlap between the perception sub-signal and other perception sub-signals of the first perception signal in the preset time period; For each dimension, according to all the sub-signal feature values ​​of the first perception signal in the dimension, a signal feature change value of the first perception signal in the dimension is calculated, wherein the signal change feature value is used to represent a feature change of the first perception signal corresponding to the dimension in the preset time period; According to the signal feature change values ​​in all the dimensions of the first perception signal, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path is calculated.

3. The method for sensing life forms according to claim 2, characterized in that: The method further comprises: Analyze current scene information according to all the perception paths; Acquire historical scene information matching the current scene information; Analyzing the historical life activity information of the current scene corresponding to the current scene in the preset time period according to the current scene information and the historical scene information; Wherein, the historical life body activity information at least includes historical life body activity area distribution information and historical life body activity frequency information, and the historical life body activity area distribution information corresponds to the historical life body activity frequency information; And, calculating the multi-dimensional environmental feature value of the first perception signal corresponding to the perception path according to the signal feature change values ​​in all the dimensions of the first perception signal includes: According to the signal feature change values ​​in all the dimensions of the first perception signal, a target priority value of the signal feature change value in each dimension of the first perception signal is matched, and the correlation relationship between the target priority values ​​of the signal feature change values ​​in different dimensions is positively correlated with the correlation relationship between the signal feature change values ​​in the different dimensions; According to the historical life activity information, updating each of the target priority values; According to the signal characteristic change values ​​in all the dimensions of the first perception signal and all the corresponding updated target priority values, the multi-dimensional environmental characteristic value of the first perception signal corresponding to the perception path is calculated.

4. The method for sensing life forms according to claim 1, characterized in that: The determining, according to all the multi-dimensional environment feature values, at least one second perception signal from all the first perception signals comprises: For each of the first perception signals, a life path prediction value corresponding to the perception path of the first perception signal is calculated according to the multi-dimensional environmental feature value of the first perception signal, wherein the life path prediction value is used to represent the distribution of vital signs existing in the perception path; Determine whether the life path prediction value of the first perception signal corresponding to the perception path is greater than or equal to a preset life path prediction threshold. When it is determined that the life path prediction value of the first perception signal corresponding to the perception path is greater than or equal to the preset life path prediction threshold, determine that the first perception signal is a second perception signal.

5. The method for sensing a living being according to any one of claims 1 to 4, characterized in that: The preset life form perception model includes an input layer, a convolution layer, a hidden layer, an output layer, and a convolution activation function corresponding to the convolution layer. All the second perception signals are input into the preset life form perception model to obtain life form perception information of the perception path corresponding to each second perception signal, including: Generate a perception model input matrix according to all the second perception signals and the input layer, wherein the perception model input matrix includes at least one of a sub-signal amplitude, a sub-signal phase value, a sub-signal frequency value, a sub-signal radiation value, a sub-signal wavelength value, a signal overlap rate, and a sub-signal identifier of each perception sub-signal in each of the second perception signals; The perception model input matrix is ​​sequentially input into the convolutional layer, hidden layer and output layer in the preset life form perception model to obtain life form perception information of each second perception signal corresponding to the perception path.

6. The method for sensing a living being according to claim 5, characterized in that: When the life form perception information includes at least the life form type information and the life form position information, the output layer includes a corresponding classification activation function and a position prediction activation function, the classification activation function is used to generate a predicted probability of the life form category, the life form type information is generated based on the predicted probability, and the position prediction activation function is used to predict the life form position information.

7. The method for sensing a living being according to claim 6, characterized in that: The preset life body perception model also includes a life body perception comprehensive loss function, and the life body perception comprehensive loss function includes a classification loss function and a position prediction regression loss function, the classification loss function corresponds to the classification activation function, and the position prediction regression loss function corresponds to the position prediction activation function; Furthermore, the preset life form perception model also includes a model training optimizer.

8. A life sensing device, characterized in that: The device comprises: A collection module, used to collect a first perception signal corresponding to at least one perception path in a preset time period; a calculation module, configured to calculate, for each of the first perception signals, a multi-dimensional environmental feature value of the first perception signal corresponding to the perception path according to the first perception signal, wherein the multi-dimensional environmental feature value is used to represent a change in the multi-dimensional environment of the perception path in the preset time period; a determination module, configured to determine at least one second perception signal from among all the first perception signals according to all the multi-dimensional environmental feature values; The perception module is used to input all the second perception signals into a preset life form perception model to obtain life form perception information corresponding to the perception path of each second perception signal, wherein the life form perception information at least includes life form type information and / or life form position information.

9. A life sensing device, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the living body perception method as described in any one of claims 1-7.

10. A computer storage medium, characterized in that: The computer storage medium stores computer instructions, which, when called, are used to execute the living body perception method as described in any one of claims 1-7.

Images