A scalable inertial sensor fall risk instant warning method and system

Abstract

This invention proposes a scalable inertial sensor-based real-time fall risk warning method and system, belonging to the field of human fall prediction technology. The method includes: setting up a main inertial sensor and sub-inertial sensors and establishing a network communication connection; collecting acceleration and angular velocity data of human behavior based on the main and sub-inertial sensors and preprocessing them, then obtaining human posture features through a Kalman filter and complementary filter fusion algorithm; based on the obtained human posture features, transforming the acceleration data to a reference coordinate system to obtain acceleration features in the reference system, removing gravitational acceleration, and integrating to obtain velocity features; inputting the acceleration features, velocity features, and human posture features into a trained fall prediction model for fall prediction. Through automatic networking between devices, data linkage between multiple inertial sensors can be achieved, improving the accuracy of fall prediction and providing sufficient preparation time for fall protection.

Description

Technical Field

[0001] The present invention belongs to the technical field of human fall prediction, and in particular relates to an expandable inertial sensor fall risk instant warning method and system. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous development of science and technology in the past few decades, high-tech products have been integrated into the daily lives of citizens. The acceptance and control of high-tech products by the elderly have significantly improved, and their demand for smart elderly care and smart elderly care products is also growing. These products can not only improve the quality of life of the elderly, but also provide necessary safety protection in emergency situations. In the daily lives of the elderly, accidental falls often lead to serious physical injuries, and because the elderly may not be able to get up or call for help on their own, delays in treatment after a fall may have more serious consequences.

[0004] To address this problem, the smart elderly care sector needs to develop products that can predict falls in advance and provide immediate protection. These products need to be highly intelligent and automated, able to predict the elderly's fall before they hit the ground, take protective measures to reduce the damage caused by ground impact to the human body, and notify family members or emergency rescue personnel in time after a fall so that medical treatment can be provided as soon as possible.

[0005] At present, wearable smart devices based on inertial sensors have been tried and applied in many fields, but the penetration rate is not high. The fundamental reason is that the detection accuracy of a single inertial sensor is difficult to achieve accurate decision-making, and the balance between missed recognition and false recognition is difficult to control. The patent with announcement number CN 116189382 A discloses that inertial sensors are set at 15 positions on the human body. Although there are many inertial sensors installed, their installation positions and numbers are fixed and cannot be split. Each inertial sensor operates independently, and cannot effectively realize synchronous sampling of data, and cannot perform real-time fall prediction. At the same time, it also increases the cost of equipment; it mainly realizes fall detection by classifying user behavior, and cannot provide preparation time for fall protection. Summary of the invention

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides an expandable inertial sensor fall risk immediate warning method and system, which is expanded on the basis of a single sensor and freely integrates multiple inertial sensors to provide recognition accuracy and convenience of fall risk immediate warning.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0008] A first aspect of the present invention provides an expandable inertial sensor fall risk instant warning method;

[0009] A scalable inertial sensor fall risk instant warning method, comprising:

[0010] Set up the main inertial sensor and sub-inertial sensor, and connect them to the network for communication;

[0011] Collect acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor;

[0012] Preprocess the collected acceleration and angular velocity data, and obtain the human body posture characteristics through Kalman filter and complementary filter fusion algorithm;

[0013] According to the acquired human posture characteristics, the acceleration data is converted to the reference coordinate system to obtain the acceleration characteristics in the reference system. After removing the gravity acceleration, the velocity characteristics are obtained by integration.

[0014] The acceleration features, velocity features and human posture features are input into the trained fall prediction model to perform fall prediction.

[0015] As a further technical solution, the main inertial sensor is arranged at the waist of the human body as a root node; the sub-inertial sensor is arranged at the leg of the human body as a sub-node; the root node and the sub-node are automatically networked via Bluetooth and perform data communication at a specific frequency.

[0016] As a further technical solution, the process of collecting acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor also includes: sampling synchronization between the main inertial sensor and the sub-inertial sensor.

[0017] As a further technical solution, the sampling synchronization process is as follows: the main inertial sensor will start broadcasting the clock externally as a slave; after receiving the broadcast of the main inertial sensor, the sub-inertial sensor will synchronize its own clock and feed back the synchronization result to the main inertial sensor; after the synchronization is completed, the main inertial sensor will stop broadcasting and enter normal mode.

[0018] As a further technical solution, the process of preprocessing the collected acceleration and angular velocity data includes:

[0019] The angular velocity data is de-biased by placing the main inertial sensor and the sub-inertial sensor on a horizontal plane, collecting angular velocity data for a certain period of time, averaging the data, and using the average value as the zero bias of each axis, which is then removed.

[0020] ω=ω′-bω

[0021] Where ω′ is the original reading of angular velocity, b ω is the desired zero bias, ω is the preprocessed data;

[0022] The acceleration measurement model can be simplified as:

[0023] a′=Sa+b a

[0024] Where S is the correction coefficient parameter matrix, b a is the zero bias, a is the real acceleration data, and a′ is the measured acceleration data. Therefore, the preprocessing process is to obtain the parameters S and b a , so as to correct the measured acceleration data;

[0025] The above formula can be simplified as:

[0026]

[0027] Using the six-position method, the main inertial sensor and the sub-inertial sensor are placed on a horizontal plane, and the positive and negative directions of the three axes are kept vertically upward. Data is collected and solved using the least squares method to finally obtain the parameter matrix M; specifically:

[0028]

[0029] In the formula, represents the average value of the acceleration measurement at the six positions; a1, a2, a3, a4, a5, a6 represent the true values ​​of the acceleration corresponding to the six positions; the above formula is abbreviated as:

[0030] L=MA

[0031] M=LA T (AA T ) -1

[0032] Where L is Symbolic abbreviation: A is Symbolic abbreviation: A T is the transpose of A.

[0033] The final corrected acceleration data is:

[0034] a=S -1 (a′-b a ).

[0035] As a further technical solution, the Kalman filter and complementary filter fusion algorithm is specifically:

[0036] The entire process of Kalman filtering is divided into a prediction phase and an update phase. The attitude quaternion is used as the state quantity. The prediction phase is expressed as follows:

[0037]

[0038]

[0039] in is the state quantity at time t, the superscript negative sign indicates the prior, the positive sign indicates the posterior, Φ is the state transfer matrix, P is the state error covariance matrix, and Q is the process noise covariance matrix;

[0040] The update phase first calculates the Kalman gain K t :

[0041]

[0042] Where H is the measurement matrix and R is the observation noise covariance matrix.

[0043]

[0044] P t + =(IK t H t ) t -

[0045] Where z is the measured value, which is only related to the acceleration data. I is the 4×4 identity matrix.

[0046] In the complementary filtering stage, only the heading angle is updated. The specific process is as follows:

[0047]

[0048] where h = [h x h y h z ] is the geomagnetic representation, m=[m x m y m z ] is the magnetometer three-axis reading, is the output of the Kalman filter at the current moment, for The conjugation of.

[0049] The component h of the Earth's magnetic field in the XY direction xy =[h x h y 0] T , corresponding to the reference north (magnetic north is different from geographic north)e y =

[010] T, through h xy With e y The heading angle can be corrected; there is a reference north around the rotation axis n r =

[001] T Rotate a certain angle to get h xy The unit vector of

[0050]

[0051] The correction angle of the magnetometer is defined as K m ∈[0,1] is the correction value, which is related to the credibility of the magnetic field. m =1 means the magnetic field is completely reliable, K m =0 means the magnetic field is completely unreliable;

[0052] Geographic North and Magnetic North do not completely coincide, corresponding to the angle

[0053] δ0=atan2(h x0 ,h y0 )

[0054] The corrected heading angle is corrected at this time

[0055]

[0056] The corresponding heading angle quaternion correction

[0057]

[0058] The corrected attitude quaternion is:

[0059]

[0060] As a further technical solution, the process of acquiring human body posture features also includes main sensor position calibration and sub-sensor position calibration.

[0061] A second aspect of the present invention provides an expandable inertial sensor fall risk instant warning system.

[0062] A scalable inertial sensor fall risk instant warning system, including:

[0063] The inertial sensor module is configured to: set a main inertial sensor and a sub-inertial sensor, and perform networking communication connection;

[0064] The data acquisition module is configured to: collect acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor;

[0065] The human body posture acquisition module is configured to: pre-process the collected acceleration and angular velocity data, and acquire the human body posture through the Kalman filter and complementary filter fusion algorithm;

[0066] The feature acquisition module is configured to: convert the acceleration data into a reference coordinate system according to the acquired human body posture, obtain the acceleration feature in the reference system, and integrate to obtain the velocity feature after removing the gravity acceleration;

[0067] The fall prediction module is configured to: input acceleration features, velocity features and human posture features into a trained fall prediction model to perform fall prediction.

[0068] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of an expandable inertial sensor fall risk instant warning method as described in the first aspect of the present invention.

[0069] A fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of an expandable inertial sensor fall risk instant warning method as described in the first aspect of the present invention are implemented.

[0070] One or more of the above technical solutions have the following beneficial effects:

[0071] The present invention provides an expandable device networking solution. According to different device states, a corresponding prediction scheme for impending falls is designed to automatically adapt to the access of multiple devices. In multi-device mode, a relationship model based on time, waist and other position postures is proposed to improve the recognition rate of fall prediction. The expansion module can adapt to more application fields by changing the installation position, such as motion capture, gait analysis, medical rehabilitation, etc.

[0072] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0074] Figure 1 This is a flow chart of the method of the first embodiment.

[0075] Figure 2 It is a system structure diagram of the second embodiment. DETAILED DESCRIPTION

[0076] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0077] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0078] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0079] Embodiment 1

[0080] This embodiment discloses an extensible inertial sensor fall risk instant warning method, which performs fall risk warning based on an extensible inertial sensor, wherein the extensible inertial sensor refers to a product that includes multiple forms, such as a product that only includes a root node, and a product that includes sub-nodes. When only the root node product is used, only a single node is used for processing. When sub-node products are expanded on this basis, the root node product will automatically identify the sub-node products and form a network to form a multi-node product.

[0081] like Figure 1 As shown, a scalable inertial sensor fall risk instant warning method includes:

[0082] Step S1, setting a main inertial sensor and a sub-inertial sensor, and performing networking communication connection;

[0083] The main inertial sensor is set at the waist of the human body as the root node; the sub-inertial sensor is set at the leg of the human body as the sub-node; the root node and the sub-node are automatically networked through Bluetooth and communicate data at a frequency of 60Hz. When at least one sub-node joins the system, the prediction algorithm in the root node will fuse the characteristic data of the root node and each sub-node to make an instant fall prediction. Due to the addition of sub-nodes, the characteristic information of different parts of the body is integrated, making the instant prediction more accurate.

[0084] In step S1, due to the addition of the sub-inertial sensor, the synchronization of the sampling data between nodes is an issue that cannot be ignored. Another feature of the present invention is to achieve millisecond-level sampling synchronization between nodes. In order to achieve sampling synchronization, when the sub-inertial sensor is added to the system, the main inertial sensor will enter the synchronization mode. At this time, the main inertial sensor will start to broadcast the clock externally as a slave. Specifically, a synchronization time is added to the broadcast of the main inertial sensor, and the time information carried in each broadcast is dynamically changed. After monitoring this time, the sub-inertial sensor will use this time value as its own operating clock and feed back the synchronization result to the main inertial sensor. After the synchronization is completed, the main inertial sensor will stop broadcasting and enter the normal mode to achieve data collection and solution.

[0085] Step S2, collecting acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor;

[0086] Step S3, preprocessing the collected acceleration and angular velocity data, and obtaining the human body posture through Kalman filtering and complementary filtering fusion algorithm;

[0087] The process of preprocessing the collected acceleration and angular velocity data includes:

[0088] The angular velocity data is debiased by placing the main inertial sensor and the sub-inertial sensor on a horizontal plane, collecting angular velocity data for a certain period of time, averaging the data, taking the average value as the zero bias of each axis, and then removing it:

[0089] ω=ω′-b ω

[0090] Where ω′ is the original reading of angular velocity, b ω is the desired zero bias, and ω is the preprocessed data.

[0091] The acceleration measurement model can be simplified as:

[0092] a′=Sa+b a

[0093] Where S is the correction coefficient parameter matrix, b a is the zero bias, a is the real acceleration data, and a′ is the measured acceleration data. Therefore, the preprocessing process is mainly to obtain the parameters S and b. a , thereby correcting the measured acceleration data.

[0094] The above formula can be simplified as:

[0095]

[0096] In this embodiment, the six-position method is used, the main inertial sensor and the sub-inertial sensor are placed on a horizontal plane, the positive and negative directions of the three axes are kept vertically upward, data is collected, and the least square method is used to solve, and finally the parameter matrix M is obtained. Specifically:

[0097]

[0098] In the formula, represents the average value of the acceleration measurement at the six positions; a1, a2, a3, a4, a5, a6 represent the true values ​​of the acceleration corresponding to the six positions; the above formula is abbreviated as:

[0099] L=MA

[0100] M=LA T (AA T ) -1

[0101] Where L is Symbolic abbreviation: A is Symbolic abbreviation: A T is the transpose of A.

[0102] The final corrected acceleration data is:

[0103] a=S -1 (a′-b a )

[0104] In the process of acquiring the human body posture, the Kalman filter and complementary filter fusion algorithm are used to obtain the human body posture. Among them, the traditional Kalman integrated update method cannot distinguish the inclination angle from the heading angle, which makes it have a high degree of coupling. In fact, the correction of the inclination angle depends on the acceleration data, and the correction of the heading angle depends on the magnetometer data. Therefore, when there is a problem with the data of one party, it will affect the result of the other party. Based on this, the Kalman filter and complementary filter fusion algorithm are used in this embodiment. The Kalman filter only updates the inclination posture, and the complementary filter only updates the heading angle. All posture representation methods are in quaternion form. The following is a detailed description of the two-stage update:

[0105] In the Kalman filtering stage, the entire Kalman process can be divided into a prediction stage and an update stage. Taking the quaternion as the state quantity, the prediction stage can be expressed as follows:

[0106]

[0107] in is the state quantity at time t, the negative superscript indicates the prior, the positive superscript indicates the posterior, Φ is the state transfer matrix, P is the state error covariance matrix, and Q is the process noise covariance matrix.

[0108] The update phase first calculates the Kalman gain K t :

[0109]

[0110] Where H is the observation matrix and R is the observation noise covariance matrix

[0111]

[0112] P t + =(IK t H t ) t -

[0113] Where z is the measured value, which is only related to the acceleration data; I is the 4×4 identity matrix.

[0114] In this embodiment, the heading angle update is separated from the inclination angle update, that is, the magnetometer data does not participate in the Kalman calculation, and only the heading angle is updated in the complementary filtering stage. The specific method is:

[0115]

[0116] where h = [h x h y h z ] is the geomagnetic representation, m=[m x m y m z ] is the magnetometer three-axis reading, q t + is the output of the Kalman filter at the current moment, for The conjugation of.

[0117] The component h of the Earth's magnetic field in the XY direction xy =[h x h y 0] T , corresponding to the reference north (magnetic north is different from geographic north)e y =

[010] T , through h xy With e y The heading angle can be corrected. There is a reference north around the rotation axis n r =

[001] T Rotating a certain angle can get h xy The unit vector of

[0118]

[0119] The correction angle of the magnetometer is defined as K m ∈[0,1] is the correction value, which is related to the credibility of the magnetic field. m =1 means the magnetic field is completely reliable, K m =0 means the magnetic field is completely unreliable.

[0120] Geographic North and Geomagnetic North do not completely coincide, the corresponding angle is:

[0121] δ0=atan2(h x0 ,h y0 )

[0122] At this time, the corrected heading angle is corrected to:

[0123]

[0124] The corresponding heading angle quaternion correction is:

[0125]

[0126] The corrected attitude quaternion is:

[0127]

[0128] Step S4, according to the acquired human body posture, the acceleration data is converted to a reference coordinate system to obtain the acceleration characteristics in the reference system, and after removing the gravity acceleration, the velocity characteristics are obtained by integration;

[0129]

[0130] in is the rotation matrix from the sensor coordinate system to the reference coordinate system represented by the quaternion, a b is the acceleration data in the sensor coordinate system, g n It is the representation of the gravitational acceleration in the reference frame.

[0131] The above process also includes the main sensor position calibration and the sub-sensor position calibration. In the calibration process, the wearer enters the calibration posture, calculates the calibration relationship of the root position, and then obtains the calibration posture relationship of other positions based on the calibration posture of the root position and the sensor posture. The root position calibration process is: in the calibration posture (the person is in a standing state, the arms are naturally spread out, the palms are facing down, and the human body is similar to a T shape), the posture rotation matrix of the root position is:

[0132]

[0133] Among them, ψ is the heading angle of the human body root position relative to the reference coordinate system, which can be obtained by measuring the inertial sensor at the root position.

[0134] The root position calibration relationship is as follows:

[0135]

[0136] in, is the rotation matrix of the sensor relative to the reference frame, is the calibration relationship matrix of the sensor relative to the root position.

[0137] The specific method of other position calibration is: the calibration relationship of other parts of the body depends on the root sensor, the corresponding part sensor, the root position calibration relationship and other information. The specific form is as follows

[0138]

[0139] in, are the rotation matrices of the root sensor and other position sensors, It is the rotation matrix of other parts relative to the root position under the calibration posture, which can be determined according to the specific form of the human body coordinate system and the calibration posture of the human body.

[0140] Subsequently, the attitude information of each position is calculated according to the calibration relationship of each position, and then the acceleration data is converted to the reference coordinate system to obtain the acceleration representation in the reference system. After removing the gravity acceleration, the velocity characteristics are obtained by integration.

[0141] Step S5, inputting the acceleration features, velocity features and human body posture features into the trained fall prediction model to perform fall prediction.

[0142] In the process of fall prediction using the trained fall prediction model, it is determined whether the acceleration and velocity characteristics of the waist node reach the set threshold. If so, it is further determined whether the waist posture information and other position posture information match the established model to a set threshold. If the threshold is reached, an alarm message is issued.

[0143] Embodiment 2

[0144] This embodiment discloses an expandable inertial sensor fall risk instant warning system;

[0145] like Figure 2 As shown, an expandable inertial sensor fall risk instant warning system includes:

[0146] The inertial sensor module is configured to: set a main inertial sensor and a sub-inertial sensor, and perform networking communication connection;

[0147] The data acquisition module is configured to: collect acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor;

[0148] The human body posture acquisition module is configured to: pre-process the collected acceleration and angular velocity data, and acquire the human body posture through the Kalman filter and complementary filter fusion algorithm;

[0149] The feature acquisition module is configured to: convert the acceleration data into a reference coordinate system according to the acquired human body posture, obtain the acceleration feature in the reference system, and integrate to obtain the velocity feature after removing the gravity acceleration;

[0150] The fall prediction module is configured to: input acceleration features, velocity features and human posture features into a trained fall prediction model to perform fall prediction.

[0151] Embodiment 3

[0152] The purpose of this embodiment is to provide a computer-readable storage medium.

[0153] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an expandable inertial sensor fall risk instant warning method as described in Example 1.

[0154] Embodiment 4

[0155] The purpose of this embodiment is to provide an electronic device.

[0156] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the scalable inertial sensor fall risk instant warning method as described in Example 1 are implemented.

[0157] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0158] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0159] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A scalable inertial sensor fall risk instant warning method, characterized in that: include: Set up the main inertial sensor and sub-inertial sensor, and connect them for network communication; Collect acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor; Preprocess the collected acceleration and angular velocity data, and obtain the human body posture characteristics through Kalman filter and complementary filter fusion algorithm; According to the acquired human posture characteristics, the acceleration data is converted to the reference coordinate system to obtain the acceleration characteristics in the reference system. After removing the gravity acceleration, the velocity characteristics are obtained by integration. The acceleration features, velocity features and human posture features are input into the trained fall prediction model to perform fall prediction.

2. The scalable inertial sensor fall risk instant warning method according to claim 1, characterized in that: The main inertial sensor is arranged at the waist of the human body as a root node; the sub-inertial sensor is arranged at the leg of the human body as a sub-node; the root node and the sub-node are automatically networked through Bluetooth and perform data communication at a specific frequency.

3. The scalable inertial sensor fall risk instant warning method according to claim 1, characterized in that: The process of collecting acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor also includes: sampling synchronization between the main inertial sensor and the sub-inertial sensor.

4. The scalable inertial sensor fall risk instant warning method as claimed in claim 3, characterized in that: The sampling synchronization process is as follows: the main inertial sensor will start broadcasting the clock as a slave; after receiving the broadcast of the main inertial sensor, the sub-inertial sensor will synchronize its own clock and feed back the synchronization result to the main inertial sensor; after the synchronization is completed, the main inertial sensor will stop broadcasting and enter normal mode.

5. The scalable inertial sensor fall risk instant warning method according to claim 1, characterized in that: The process of preprocessing the collected acceleration and angular velocity data includes: The angular velocity data is de-biased by placing the main inertial sensor and the sub-inertial sensor on a horizontal plane, collecting angular velocity data for a certain period of time, averaging the data, and using the average value as the zero bias of each axis, which is then removed. ω=ω′-b ω Where ω′ is the original reading of angular velocity, b ω is the desired zero bias, ω is the preprocessed data; The acceleration measurement model is simplified as follows: a′=Sa+b a Where S is the correction coefficient parameter matrix, b a is the zero bias, a is the real acceleration data, and a′ is the measured acceleration data. Therefore, the preprocessing process is to obtain the parameters S and b a , so as to correct the measured acceleration data; The above formula is simplified to: Using the six-position method, the main inertial sensor and the sub-inertial sensor are placed on a horizontal plane, and the positive and negative directions of the three axes are kept vertically upward. Data is collected and solved using the least squares method to finally obtain the parameter matrix M; specifically: In the formula, represents the average value of the acceleration measurement at the six positions; a1, a2, a3, a4, a5, a6 represent the true values ​​of the acceleration corresponding to the six positions; the above formula is abbreviated as: L=MA M=DAY T (CHALLENGE ACCEPTED T ) -1 Where L is Symbolic abbreviation: A is Symbolic abbreviation: A T is the transpose of A; The final corrected acceleration data is: a=S -1 (a′-b a )。 6. The scalable inertial sensor fall risk instant warning method according to claim 1, characterized in that: The Kalman filter and complementary filter fusion algorithm is specifically as follows: The entire process of Kalman filtering is divided into a prediction phase and an update phase. The attitude quaternion is used as the state quantity. The prediction phase is expressed as follows: in is the state quantity at time t, the superscript negative sign indicates the prior, the positive sign indicates the posterior, Φ is the state transfer matrix, P is the state error covariance matrix, and Q is the process noise covariance matrix; The update phase first calculates the Kalman gain K t : Where H is the observation matrix and R is the observation noise covariance matrix; Where z is the measured value, which is only related to the acceleration data; I is the 4×4 identity matrix; In the complementary filtering stage, only the heading angle is updated. The specific process is as follows: where h = [h x h y h z ] is the geomagnetic representation, m=[m x m y m z ] is the magnetometer three-axis reading, is the output of the Kalman filter at the current moment, for conjugation of; The component h of the Earth's magnetic field in the XY direction xy =[h x h y 0] T , corresponding to the reference north y =[010] T , through h xy With e y The heading angle can be corrected; there is a reference north around the rotation axis n r =[001] T Rotate a certain angle to get h xy The unit vector of The correction angle of the magnetometer is defined as K m ∈[0,1] is the correction value, which is related to the credibility of the magnetic field. m =1 means the magnetic field is completely reliable, K m =0 means the magnetic field is completely unreliable; Geographic North and Geomagnetic North do not completely coincide, the corresponding angle is: δ0=atan2(h x0 ,h y0 ) At this time, the corrected heading angle is: The corresponding heading angle quaternion correction is: The corrected attitude quaternion is:

7. The scalable inertial sensor fall risk instant warning method according to claim 1, characterized in that: The process of acquiring human body posture features also includes main sensor position calibration and sub-sensor position calibration.

8. An expandable inertial sensor fall risk immediate warning system, characterized in that: include: The inertial sensor module is configured to: set a main inertial sensor and a sub-inertial sensor, and perform networking communication connection; The data acquisition module is configured to: collect acceleration and angular velocity data of human behavior based on the main inertial sensor and the sub-inertial sensor; The human body posture acquisition module is configured to: pre-process the collected acceleration and angular velocity data, and acquire the human body posture through the Kalman filter and complementary filter fusion algorithm; The feature acquisition module is configured to: convert the acceleration data into a reference coordinate system according to the acquired human body posture, obtain the acceleration feature in the reference system, and integrate to obtain the velocity feature after removing the gravity acceleration; The fall prediction module is configured to: input acceleration features, velocity features and human posture features into a trained fall prediction model to perform fall prediction.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the expandable inertial sensor fall risk instant warning method as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the expandable inertial sensor fall risk instant warning method as described in any one of claims 1 to 7 are implemented.

Images