A fall recognition method and device

Abstract

The present invention proposes a fall recognition method and device, which relates to the technical field of fall detection. The present invention obtains human motion data, including acceleration and heart rate data; then calculates the combined acceleration according to the acceleration in the human motion data; then determines the human motion state according to the combined acceleration and heart rate data; according to the human motion state, the human motion data is solved by using the quaternion posture solution method to obtain human posture parameters; finally, the human posture parameters and human motion data are input into a preset fall recognition model, and the preset fall recognition model includes a feature transformer constructed based on the DBN deep belief network algorithm and a classifier constructed based on the RF random forest algorithm. The feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix, and the feature matrix is ​​input into the classifier to obtain a fall recognition result. It enables it to obtain a more satisfactory accuracy and ensures the accuracy of fall recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of fall detection, and in particular to a fall identification method and device. Background Art

[0002] With the increasingly serious trend of population aging, the solution to the elderly care problem has become more urgent. Although the welfare system is constantly improving, sudden incidents of death among the elderly due to falls still occur frequently. Therefore, preventing the elderly from falling has become increasingly important in real life.

[0003] At present, with the increasing aging of the population in my country, health monitoring of the elderly has become a key issue of concern to the entire society. Fall detection equipment is also one of the hot topics of research. Many researchers and funds at home and abroad have devoted themselves to the research and development of fall detection equipment or algorithms. According to the data collection method and usage of the fall detection device, fall detection can be divided into three categories: environmental signal fall detection technology, video image fall detection technology, and wearable device fall detection technology.

[0004] However, these fall detection devices all require the design of algorithms for fall identification. The existing algorithms are difficult to identify, have low accuracy, and false alarms often occur. Summary of the invention

[0005] The purpose of the present invention is to provide a fall recognition method and device, so as to improve the problem that the differences in evaluation indicators of different areas are not considered in the prior art, resulting in inaccurate evaluation.

[0006] In a first aspect, an embodiment of the present application provides a fall recognition method, comprising the following steps:

[0007] Acquiring human motion data, wherein the human motion data includes at least acceleration and heart rate data;

[0008] Calculating a combined acceleration based on the acceleration in the human body motion data;

[0009] Determining a human body motion state according to the combined acceleration and the heart rate data;

[0010] According to the human body motion state, the human body motion data is solved by using a quaternion posture solution method to obtain human body posture parameters;

[0011] The human body posture parameters and the human body motion data are input into a preset fall recognition model, wherein the preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm. The feature transformer performs feature transformation on the human body posture parameters to obtain a feature matrix, and the classifier performs classification according to the feature matrix to obtain a fall recognition result.

[0012] Based on the first aspect, in some embodiments of the present invention, the acceleration includes an x-axis acceleration a x , y-axis acceleration a y and z-axis acceleration a z ;

[0013] The step of calculating the combined acceleration according to the acceleration in the human body motion data comprises the following steps:

[0014] The x-axis acceleration a x , y-axis acceleration a y and z-axis acceleration a z Substitute it into the calculation formula of the combined acceleration to obtain the combined acceleration, which is:

[0015]

[0016] Where: A is the total acceleration, A1 is the x-axis acceleration, A2 is the y-axis acceleration, and A3 is the z-axis acceleration.

[0017] Based on the first aspect, in some embodiments of the present invention, the step of determining the human body motion state according to the combined acceleration and the heart rate data comprises the following steps:

[0018] Comparing the combined acceleration with a preset acceleration threshold to obtain a first comparison result;

[0019] Comparing the heart rate data with a preset heart rate threshold to obtain a second comparison result;

[0020] The movement state of the human body is obtained according to the first comparison result and the second comparison result.

[0021] Based on the first aspect, in some embodiments of the present invention, according to the human body motion state, the human body motion data is solved using a quaternion posture solution method to obtain a human body posture parameter, comprising the following steps:

[0022] Determine whether the human body motion state is strenuous exercise. If so, solve the human body motion data using a quaternion posture solution method to obtain human body posture parameters; if not, generate normal activity reminder information.

[0023] Based on the first aspect, in some embodiments of the present invention, the following steps are further included:

[0024] Acquire a sample data set, where each sample in the sample data set includes a sample feature matrix and a sample category;

[0025] According to the sample data set, the sample feature matrix is ​​used as input and the sample category is used as output, and the initial classifier constructed by the RF random forest algorithm is trained to obtain a classifier constructed based on the RF random forest algorithm.

[0026] Based on the first aspect, in some embodiments of the present invention, the following steps are further included:

[0027] generating a control command according to the fall recognition result;

[0028] The control command is sent to the anti-fall device to achieve fall protection.

[0029] Based on the first aspect, in some embodiments of the present invention, the following steps are further included:

[0030] A fall reminder message is generated according to the fall recognition result and sent to the user end.

[0031] In a second aspect, an embodiment of the present application provides a fall recognition device, comprising:

[0032] A data acquisition module, used to acquire human motion data, wherein the human motion data at least includes acceleration and heart rate data;

[0033] A combined acceleration calculation module, used to calculate the combined acceleration according to the acceleration in the human body motion data;

[0034] A motion state determination module, used for determining the motion state of a human body according to the combined acceleration and the heart rate data;

[0035] A posture calculation module is used to calculate the human body motion data using a quaternion posture calculation method according to the human body motion state to obtain human body posture parameters;

[0036] The fall recognition model module is used to input the human posture parameters and the human motion data into a preset fall recognition model. The preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm. The feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix. The classifier performs classification according to the feature matrix to obtain a fall recognition result.

[0037] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory for storing one or more programs and a processor. When the one or more programs are executed by the processor, the method described in any one of the first aspects above is implemented.

[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method as described in any one of the above-mentioned first aspects.

[0039] The embodiments of the present invention have at least the following advantages or beneficial effects:

[0040] The embodiment of the present invention provides a fall recognition method and device, which obtains human motion data, wherein the human motion data includes at least acceleration and heart rate data; then calculates the combined acceleration according to the acceleration in the human motion data; then determines the human motion state according to the combined acceleration and the heart rate data; according to the human motion state, solves the human motion data using a quaternion posture solution method to obtain human posture parameters; finally, inputs the human posture parameters and the human motion data into a preset fall recognition model, wherein the preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm, wherein the feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix, and the classifier performs classification according to the feature matrix to obtain a fall recognition result. By determining the human body's motion state, the human body's state can be accurately judged to make a suspected fall judgment, and the human body posture parameters can be further calculated according to the human body's motion state. The human body posture parameters and human body motion data are then combined as the input of the fall recognition model to determine whether a fall has occurred. The fall recognition model uses a feature transformer built based on the DBN deep belief network algorithm and a classifier built based on the RF random forest algorithm. DBN is used for nonlinear feature transformation, thereby introducing better features, enabling it to obtain more satisfactory accuracy and ensuring the accuracy of fall recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A flow chart of a fall recognition method provided by an embodiment of the present invention;

[0043] Figure 2 A structural block diagram of a fall recognition device provided by an embodiment of the present invention;

[0044] Figure 3 A structural block diagram of an electronic device provided by an embodiment of the present invention.

[0045] Icons: 110 - data acquisition module; 120 - combined acceleration calculation module; 130 - motion state determination module; 140 - posture solution module; 150 - fall recognition model module; 101 - memory; 102 - processor; 103 - communication interface. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0047] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0048] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0049] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0050] In the description of the present application, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the product of the application is usually placed when in use. They are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0051] Example

[0052] In conjunction with the accompanying drawings, some implementation methods of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0053] Please see Figure 1 , Figure 1 A flow chart of a fall recognition method provided by an embodiment of the present invention. The fall recognition method comprises the following steps:

[0054] Step S110: Obtain human motion data, the human motion data at least includes acceleration and heart rate data; in this embodiment, the human motion data also includes angular velocity, the angular velocity and acceleration can be obtained through an anti-fall device worn on the human body, the anti-fall device can be a wearable vest, a sensor is provided inside, and data can be collected in real time, the sensor can be a nine-axis motion sensor of model MPU9250, the nine-axis motion sensor has a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer integrated inside, and can provide original nine-axis sensor data for subsequent human posture solution. The heart rate data can be obtained by real-time monitoring through a wearable bracelet, or by real-time monitoring through a heart rate detection device in the anti-fall device. It should be noted that the human motion data can be real-time data, or it can be obtained at a certain period. If it is obtained at a certain period, the human motion data also includes time.

[0055] Step S120: Calculating a combined acceleration according to the acceleration in the human body motion data;

[0056] Among them, for the acceleration collected by the three-axis accelerometer, the acceleration includes the x-axis acceleration a x , y-axis acceleration a y and z-axis acceleration a z The combined acceleration can be calculated using the combined acceleration calculation formula. Specifically, the following steps are included:

[0057] The x-axis acceleration a x , y-axis acceleration a y and z-axis acceleration a zSubstitute it into the calculation formula of the combined acceleration to obtain the combined acceleration, which is:

[0058]

[0059] Where: A is the total acceleration, A1 is the x-axis acceleration, A2 is the y-axis acceleration, and A3 is the z-axis acceleration.

[0060] According to the above calculation formula, the total acceleration can be obtained as

[0061] Step S130: Determine the human body's motion state based on the combined acceleration and the heart rate data; in this embodiment, the combined acceleration can be used to distinguish the human body's motion state. The greater the combined acceleration, the more intense the motion, and vice versa, the more stable the human body's motion. In order to further improve the accuracy of determining the human body's motion state, heart rate data can also be added for determination. Specifically, the following steps are included:

[0062] First, the combined acceleration is compared with a preset acceleration threshold to obtain a first comparison result; the above acceleration threshold can be set according to actual conditions.

[0063] Then, the heart rate data is compared with the preset heart rate threshold to obtain a second comparison result; the above-mentioned preset heart rate threshold can be set by the user according to his or her physical condition to improve the judgment rate, or a preset default value can be adopted, and the above-mentioned default value can be set based on experience.

[0064] Finally, the human body motion state is obtained according to the first comparison result and the second comparison result. In this embodiment, if the first comparison result is greater than the acceleration threshold and the second comparison result is greater than the heart rate threshold, it means that vigorous exercise may occur, and the human body motion state can be obtained as vigorous exercise; in other cases, it means that it is not vigorous exercise, and the human body motion state can be obtained as normal activity.

[0065] By combining the acceleration and heart rate data to determine the human body's motion state, the accuracy of determining the human body's motion state can be improved, and deviations in the determination of the human body's motion state due to data collection errors can be avoided.

[0066] Step S140: according to the human body motion state, the human body motion data is solved by using the quaternion posture solution method to obtain human body posture parameters; in the case of intense exercise, the probability of falling is relatively high, therefore, after determining the human body motion state, a suspected fall judgment can be made according to the human body motion state, specifically:

[0067] Determine whether the human body motion state is strenuous exercise. If so, the human body motion data is solved using the quaternion posture solution method to obtain human body posture parameters; if not, normal activity reminder information is generated. If the human body motion state is not strenuous exercise and belongs to normal activity, normal activity reminder information can be generated and sent to the user so that the user can check the motion status. The above-mentioned user can be a remote user. By uploading the normal activity reminder information to the server, the user accesses the server to obtain the reminder information, so that the user can understand the motion status of the relevant person in a timely manner.

[0068] In this embodiment, when the Euler angle represents the posture change, it represents three rotations around the three coordinate axes i, j, and k. According to Euler's theorem, these three rotations can be equivalent to one rotation around a certain axis. Using the quaternion posture solution method to solve the problem means setting the quaternion to represent the rotation in three-dimensional space. The basic representation of the quaternion is:

[0069] Q(q0,q1,q2,q3)=q0+q1i+q2j+q3k, where i rotation represents the rotation from the positive direction of the X axis to the positive direction of the Y axis in the plane where the X axis intersects with the Y axis, j rotation represents the rotation from the positive direction of the Z axis to the positive direction of the X axis in the plane where the Z axis intersects with the X axis, and k rotation represents the rotation from the positive direction of the Y axis to the positive direction of the Z axis in the plane where the Y axis intersects with the Z axis. When used for attitude representation, q0=cos(θ / 2), q n =u n *sin(θ / 2), n=1, 2, 3, u n represents a certain axis, and θ is the rotation angle, which can be obtained according to the acceleration in the human motion data. Then the quaternion is solved to obtain the human posture parameters, which include the angle parameter roll, the angle parameter pitch and the angle parameter yaw, which can be expressed as:

[0070]

[0071] Step S150: Input the human body posture parameters and the human body motion data into a preset fall recognition model, wherein the preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm, wherein the feature transformer performs feature transformation on the human body posture parameters to obtain a feature matrix, and the classifier performs classification according to the feature matrix to obtain a fall recognition result.

[0072] In this embodiment, the human body posture parameters and the human body motion data can be constructed into an initial feature matrix, which includes the angular parameters roll, pitch, and yaw in the human body posture parameters and the acceleration, heart rate data, and angular velocity in the human body motion data.

[0073] The feature transformer constructed based on the DBN deep belief network algorithm can be to set K restricted Boltzmann machines in the DBN network, and perform feature transformation on the feature matrix in the original feature domain to obtain the corresponding feature matrix in these K regions, and use the corresponding feature matrix in these K regions as the new feature matrix for training or testing. Among them, the output feature matrix of the mth restricted Boltzmann machine is expressed as: F m =F m-1 W m +B m ,m=1,2,…,K, where W m is the weight matrix of the mth restricted Boltzmann machine, B m is the deviation matrix of the mth restricted Boltzmann machine. The specific value of K can be determined according to the recognition accuracy. When DBN is used for feature conversion, its performance can be optimized by adjusting its iteration cycle, momentum coefficient, learning rate and network structure.

[0074] After feature transformation by the feature transformer based on DBN deep belief network, classification and recognition are then performed using a classifier based on RF random forest algorithm. The above classifier can be trained through sample data, specifically including the following steps:

[0075] First, a sample data set is obtained, wherein each sample in the sample data set includes a sample feature matrix and a sample category; the sample data set may be a historical data set, including a data set of falls and a data set of no falls. The sample feature matrix is ​​a feature matrix converted by a feature transformer constructed by a DBN deep belief network algorithm, and the sample category may be falls and normal.

[0076] Then, according to the sample data set, the sample feature matrix is ​​used as input and the sample category is used as output, and the initial classifier constructed by the RF random forest algorithm is trained to obtain a classifier constructed based on the RF random forest algorithm.

[0077] In this embodiment, the classifier includes multiple decision trees and a voting selection. The decision tree adopts a classification and regression tree based on the CART algorithm. The optimal number of decision trees is set to 500. The decision results are defined as two categories: fall and normal. The voting selection is based on the 500 decision results. The LOO (leave one out) method can be used to cross-validate the training samples. At the same time, based on the LOO algorithm, the sample test results are statistically analyzed based on the LOSO (leave-one-subject-out) method. If most decision trees are identified as a fall, the voting selection considers that the recognition result is a fall; otherwise, the recognition result is normal.

[0078] In the above implementation process, human motion data is obtained, and the human motion data includes at least acceleration and heart rate data; then, the combined acceleration is calculated according to the acceleration in the human motion data; then, the human motion state is determined according to the combined acceleration and the heart rate data; according to the human motion state, the human motion data is solved by using the quaternion posture solution method to obtain human posture parameters; finally, the human posture parameters and the human motion data are input into a preset fall recognition model, and the preset fall recognition model includes a feature transformer constructed based on the DBN deep belief network algorithm and a classifier constructed based on the RF random forest algorithm, the feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix, and the classifier performs classification according to the feature matrix to obtain a fall recognition result. By determining the human body's motion state, the human body's state can be accurately judged to make a suspected fall judgment, and the human body posture parameters can be further calculated according to the human body's motion state. The human body posture parameters and human body motion data are then combined as the input of the fall recognition model to determine whether a fall has occurred. The fall recognition model uses a feature transformer built based on the DBN deep belief network algorithm and a classifier built based on the RF random forest algorithm. DBN is used for nonlinear feature transformation, thereby introducing better features, enabling it to obtain more satisfactory accuracy and ensuring the accuracy of fall recognition.

[0079] After the fall recognition result is obtained, a fall protection operation can also be performed, which specifically includes the following steps:

[0080] First, a control command is generated according to the fall recognition result; in this embodiment, when the fall recognition result is a fall, a control command for turning on the anti-fall device is generated.

[0081] Then, the control command is sent to the anti-fall device to achieve fall protection. After receiving the control command, the anti-fall device opens the anti-fall device to perform fall protection. The anti-fall device belongs to the prior art. For example, the anti-fall device is equipped with a high-pressure gas cylinder connected to a plurality of first airbags through a first air pipe. During the implementation process, the anti-fall device is also provided with an inner layer, a flexible buffer layer and an outer layer from the inside to the outside. The airbag is installed on the outer layer. The airbag can be opened quickly and automatically. The flexible buffer layer is used to buffer the pressure generated by the instantaneous inflation of the first airbag to prevent the airbag from causing accidental injury to the human body during the instantaneous inflation process. After receiving the control command, the anti-fall device opens the high-pressure gas cylinder to inflate the first airbag to achieve fall protection.

[0082] The following steps are also included:

[0083] Generate fall reminder information according to the fall recognition result and send it to the user end. In this embodiment, when the fall recognition result is a fall, a fall reminder information can be generated and sent to the user end through the application of the wireless communication function. The user end can be an emergency contact or a medical institution, etc. The fall reminder information can be used to timely understand the fall accident, and then timely take measures to reduce the subsequent damage caused by the fall.

[0084] It should be noted that if a false alarm occurs, the fall reminder information can be manually canceled within a certain period of time to reduce unnecessary panic caused by the false alarm.

[0085] Based on the same inventive concept, the present invention also proposes a fall recognition device, see Figure 2 , Figure 2 This is a structural block diagram of a fall recognition device provided by an embodiment of the present invention. The fall recognition device includes:

[0086] A data acquisition module 110 is used to acquire human motion data, wherein the human motion data at least includes acceleration and heart rate data;

[0087] A combined acceleration calculation module 120, configured to calculate a combined acceleration based on the acceleration in the human body motion data;

[0088] A motion state determination module 130, configured to determine a human motion state according to the combined acceleration and the heart rate data;

[0089] The posture calculation module 140 is used to calculate the human body motion data using a quaternion posture calculation method according to the human body motion state to obtain human body posture parameters;

[0090] The fall recognition model module 150 is used to input the human posture parameters into a preset fall recognition model. The preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm. The feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix. The classifier performs classification according to the feature matrix to obtain a fall recognition result.

[0091] In the above implementation process, human motion data is acquired through the data acquisition module 110, and the human motion data includes at least acceleration and heart rate data; the combined acceleration calculation module 120 calculates the combined acceleration according to the acceleration in the human motion data; the motion state determination module 130 determines the human motion state according to the combined acceleration and the heart rate data; the posture solution module 140 solves the human motion data using the quaternion posture solution method according to the human motion state to obtain human posture parameters; the fall recognition model module 150 inputs the human posture parameters and the human motion data into a preset fall recognition model, and the preset fall recognition model includes a feature transformer constructed based on the DBN deep belief network algorithm and a classifier constructed based on the RF random forest algorithm, the feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix, and the classifier performs classification according to the feature matrix to obtain a fall recognition result. By determining the human body's motion state, the human body's state can be accurately judged to make a suspected fall judgment, and the human body posture parameters can be further calculated according to the human body's motion state. The human body posture parameters and human body motion data are then combined as the input of the fall recognition model to determine whether a fall has occurred. The fall recognition model uses a feature transformer built based on the DBN deep belief network algorithm and a classifier built based on the RF random forest algorithm. DBN is used for nonlinear feature transformation, thereby introducing better features, enabling it to obtain more satisfactory accuracy and ensuring the accuracy of fall recognition.

[0092] See also Figure 3 , Figure 3 A schematic structural block diagram of an electronic device provided in an embodiment of the present application. The electronic device includes a memory 101, a processor 102 and a communication interface 103, and the memory 101, the processor 102 and the communication interface 103 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines. The memory 101 can be used to store software programs and modules, such as program instructions / modules corresponding to the fall recognition device 100 provided in an embodiment of the present application. The processor 102 executes various functional applications and data processing by executing the software programs and modules stored in the memory 101. The communication interface 103 can be used to communicate signaling or data with other node devices.

[0093] Among them, the memory 101 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0094] The processor 102 may be an integrated circuit chip with signal processing capability. The processor 102 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0095] Understandably, Figure 3 The structure shown is for illustration only. The electronic device may also include Figure 3 More or fewer components as shown, or with Figure 3 Different configurations shown. Figure 3 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0096] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0097] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0098] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0099] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0100] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A fall recognition method, characterized in that: The following steps are involved: Acquiring human motion data, wherein the human motion data includes at least acceleration and heart rate data; Calculating a combined acceleration based on the acceleration in the human body motion data; Determining a human body motion state according to the combined acceleration and the heart rate data; According to the human body motion state, the acceleration data is solved by using a quaternion posture solution method to obtain human body posture parameters; Inputting the human posture parameters and the human motion data into a preset fall recognition model, wherein the preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm, wherein the feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix, and the classifier performs classification according to the feature matrix to obtain a fall recognition result; The step of combining the acceleration and the heart rate data to determine the motion state of the human body comprises the following steps: Comparing the combined acceleration with a preset acceleration threshold to obtain a first comparison result; Comparing the heart rate data with a preset heart rate threshold to obtain a second comparison result; According to the first comparison result and the second comparison result, a human body movement state is obtained, wherein if the first comparison result is greater than an acceleration threshold and the second comparison result is greater than a heart rate threshold, then vigorous exercise occurs, and the human body movement state is recorded as vigorous exercise; By combining acceleration and heart rate data, the human body's motion state can be determined; Among them, the feature transformer built based on the DBN deep belief network algorithm is to set K restricted Boltzmann machines in the DBN network, and perform feature transformation on the feature matrix in the original feature domain to obtain the corresponding feature matrices in these K regions, and use the corresponding feature matrices in these K regions as new feature matrices for training or testing. Among them, the output feature matrix of the mth restricted Boltzmann machine is expressed as: ,in, is the weight matrix of the mth restricted Boltzmann machine, is the deviation matrix of the mth restricted Boltzmann machine, where the specific value of K is determined according to the recognition accuracy; After feature transformation by the feature transformer based on DBN deep belief network, classification and recognition are performed using a classifier based on RF random forest algorithm, where the classifier is trained by sample data, specifically including the following steps: Acquire a sample data set, wherein each sample in the sample data set includes a sample feature matrix and a sample category; wherein the sample data set is a historical data set, including a data set of falls and a data set of no falls, wherein the sample feature matrix is ​​a feature matrix converted by a feature transformer constructed by a DBN deep belief network algorithm, wherein the sample categories include falls and normal; According to the sample data set, the sample feature matrix is ​​used as input and the sample category is used as output, and the initial classifier constructed by the RF random forest algorithm is trained to obtain a classifier constructed based on the RF random forest algorithm; wherein the classifier includes multiple decision trees and a voting selection, the decision tree adopts a classification regression tree based on the CART algorithm, the number of decision trees is set to 500, and the decision results are defined as two categories: fall and normal. The voting selection is based on the 500 decision results. If most decision trees identify it as a fall, the voting selection considers that the recognition result is a fall; otherwise, the recognition result is normal.

2. The fall recognition method according to claim 1, characterized in that: The acceleration includes the x-axis acceleration , y-axis acceleration and z-axis acceleration ; The step of calculating the combined acceleration according to the acceleration in the human body motion data comprises the following steps: The x-axis acceleration , y-axis acceleration and z-axis acceleration Substitute it into the calculation formula of the combined acceleration to obtain the combined acceleration, which is: Where: A is the total acceleration, is the x-axis acceleration, is the y-axis acceleration, is the z-axis acceleration.

3. The fall recognition method according to claim 1, characterized in that: According to the human body motion state, the human body motion data is solved by using a quaternion posture solving method to obtain human body posture parameters, comprising the following steps: Determine whether the human body motion state is strenuous exercise. If so, solve the human body motion data using a quaternion posture solution method to obtain human body posture parameters; if not, generate normal activity reminder information.

4. The fall recognition method according to claim 1, characterized in that: The following steps are also included: generating a control command according to the fall recognition result; The control command is sent to the anti-fall device to achieve fall protection.

5. The fall recognition method according to claim 4, characterized in that: The following steps are also included: A fall reminder message is generated according to the fall recognition result and sent to the user end.

6. A fall recognition device, based on the fall recognition method according to any one of claims 1 to 5, characterized in that: include: A data acquisition module, used to acquire human motion data, wherein the human motion data at least includes acceleration and heart rate data; A combined acceleration calculation module, used to calculate the combined acceleration according to the acceleration in the human body motion data; A motion state determination module, used for determining the motion state of a human body according to the combined acceleration and the heart rate data; A posture calculation module is used to calculate the human body motion data using a quaternion posture calculation method according to the human body motion state to obtain human body posture parameters; The fall recognition model module is used to input the human posture parameters and the human motion data into a preset fall recognition model. The preset fall recognition model includes a feature transformer constructed based on a DBN deep belief network algorithm and a classifier constructed based on an RF random forest algorithm. The feature transformer performs feature transformation on the human posture parameters to obtain a feature matrix. The classifier performs classification according to the feature matrix to obtain a fall recognition result.

7. An electronic device, characterized in that: include: A memory for storing one or more programs; processor; When the one or more programs are executed by the processor, the fall recognition method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fall recognition method according to any one of claims 1 to 5 is implemented.

Images