A fall detection method based on multi-point posture sensor cooperation

Abstract

The present invention provides a fall detection method based on the cooperation of multi-point posture sensors, and relates to the field of data processing technology. The method includes acquiring multi-point detection data of different objects, performing feature analysis of normal posture, and forming normal posture feature data; performing posture-based fall feature analysis according to the multi-point detection data of different objects, and forming fall posture feature data; collecting real-time posture data of the target object, and combining the normal posture feature data and the fall posture feature data, performing real-time fall monitoring and analysis, and forming real-time fall monitoring and analysis result data. The method combines the big data of the falling limb information and the limb information during normal action of the same type of objects to establish a systematic method that can accurately, efficiently and in real time perform fall detection, which greatly improves the effect and efficiency of fall detection and provides an important basis for detection results for fall prevention and care.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a fall detection method based on the cooperation of multi-point posture sensors. Background Art

[0002] Safety accidents caused by elderly people falling occur frequently, which also poses a huge safety hazard to the lives of the elderly. And with the increase of age, falls will be aggravated due to poor control of limbs, and targeted training and adjustment are needed to improve physical functions. Therefore, fall detection is particularly important. At present, most of the fall detection is still empirically determined through regular inspections, which is highly dependent on the subjective initiative of the test object and cannot achieve accurate and timely fall detection.

[0003] With the development of science and technology, it is possible to establish an analytical evaluation method for fall detection through big data. At present, there is no systematic fall detection analysis based on big data. The STEADI fall prevention risk screening toolkit is mainly used for evaluation. The evaluation coverage is not comprehensive and cannot be personalized for different people. Therefore, it is important to improve the accuracy and timeliness of fall detection and improve the big data fall analysis method.

[0004] Therefore, designing a fall detection method based on multi-point posture sensors is an urgent problem to be solved. By combining the big data of falling limb information and limb information during normal actions of the same type of objects, a systematic method that can accurately, efficiently and real-time detect falls is established, which greatly improves the effect and efficiency of fall detection and provides an important basis for detection results for fall prevention and care. Summary of the invention

[0005] The purpose of the present invention is to provide a fall detection method based on the cooperation of multi-point posture sensors, by acquiring multi-point detection data of different objects, performing feature extraction and analysis of normal motion posture, obtaining feature data of normal posture, and also relatively extracting posture features in the fall state to form fall posture feature data. On the basis of fully combining normal posture feature data and fall posture feature data, the real-time posture data is compared and analyzed, and it can be quickly and accurately determined whether the current target object has the possibility of falling. The timeliness and accuracy of fall detection are greatly improved, making fall detection analysis more efficient and reasonable.

[0006] In a first aspect, the present invention provides a fall detection method based on the cooperation of a multi-point posture sensor, comprising acquiring multi-point detection data of different objects, performing feature analysis of normal posture, and forming normal posture feature data; performing posture-based fall feature analysis based on the multi-point detection data of different objects, and forming fall posture feature data; collecting real-time posture data of the target object, and combining the normal posture feature data and the fall posture feature data, performing real-time fall monitoring and analysis, and forming real-time fall monitoring and analysis result data.

[0007] In the present invention, the method obtains multi-point detection data of different objects, performs feature extraction and analysis of normal motion postures, obtains feature data of normal postures, and also relatively extracts posture features in a falling state to form falling posture feature data. On the basis of fully combining normal posture feature data and falling posture feature data, a comparative analysis is performed on real-time posture data, and it can be quickly and accurately determined whether the current target object has the possibility of falling. The timeliness and accuracy of fall detection are greatly improved, making fall detection analysis more efficient and reasonable.

[0008] As a possible implementation method, multi-point detection data of different objects are obtained, and feature analysis of normal posture is performed to form normal posture feature data, including: for different objects, normal posture data of each point in the multi-point detection data are extracted, and posture change analysis based on the reference point is performed to form normal posture change data of the object; for the normal posture change data of different objects, clustering is performed based on limb size to form normal posture change data sets of different body shapes; for the normal posture change data sets of different body shapes, angle change coordination analysis based on different points relative to the reference point is performed to form normal posture angle change coordination data corresponding to the normal posture change data sets of different body shapes; and all body shape normal posture change data sets and corresponding normal posture angle change coordination data are combined to form normal posture feature data.

[0009] In the present invention, the normal posture feature analysis of different objects under multi-point detection data is mainly considered in the following aspects: on the one hand, since the characteristics of human action behavior are mainly expressed based on limbs, it is necessary to fully extract the posture data of each limb used for analysis when performing fall detection analysis. On the other hand, considering that objects of different body shapes may have large changes in behavior posture due to large differences in body shape whether in normal action or when falling, in order to obtain more accurate feature data, it is necessary to reasonably divide the posture data of different objects extracted based on the limb conditions, so as to facilitate the subsequent formation of more targeted feature information based on this reasonable division, and ensure that the accuracy of fall detection is higher. The third aspect is that for a single individual, whether it is a normal posture action or a posture action when falling, the limb movements need to be coordinated. These limb movements are not independent of each other, but need to be synchronously and comprehensively considered when performing fall detection. The coordination between limbs is an important feature information.

[0010] As a possible implementation method, for different objects, the normal posture data of each point in the multi-point detection data is extracted, and the posture change analysis based on the reference point is performed to form the normal posture change data of the object, including: for each object, the trunk point is calibrated as the reference point, the posture space coordinate system is established with the reference point as the origin, and the feature analysis period and the size information extraction amount threshold K are set; the left upper limb point is determined based on the left upper limb coordinate position of the posture space coordinate system within the feature analysis period And determine the coordinate position of the left upper limb Left upper limb reference vector relative to the reference point in, (X0, Y0, Z0) are the coordinates of the reference point in the posture space coordinate system, and n represents the number of different objects; within the feature analysis cycle, K position points of the left upper limb coordinate position are randomly extracted, and the effective left upper limb detection size of the left upper limb point relative to the reference point is determined Determine the coordinate position of the right upper limb based on the posture space coordinate system during the feature analysis period And determine the coordinate position of the right upper limb Right upper limb reference vector relative to the reference point in, During the feature analysis cycle, randomly extract K position points of the right upper limb coordinate position, and determine the effective right upper limb detection size of the right upper limb point relative to the reference point Determine the left lower limb point position based on the left lower limb coordinate position in the posture space coordinate system within the feature analysis cycle And determine the coordinate position of the left lower limb Left lower limb reference vector relative to the reference point in, During the feature analysis cycle, randomly extract K position points of the left lower limb coordinate position, and determine the effective left lower limb detection size of the left lower limb point relative to the reference point Determine the coordinate position of the right lower limb based on the posture space coordinate system within the feature analysis period And determine the coordinate position of the right lower limb Right lower limb reference vector relative to the reference point in, During the feature analysis cycle, randomly extract K position points of the right lower limb coordinate position, and determine the effective right lower limb detection size of the right lower limb point relative to the reference point. Determine the head point based on the head coordinate position in the posture space coordinate system during the feature analysis cycle And determine the head coordinate position Head reference vector relative to the reference point in,

[0011] In the present invention, the data extraction of normal posture actions includes the movement characteristics of limbs including upper limbs, lower limbs and head, and these movement characteristics are not static, but change in coordination in the time dimension. Therefore, when extracting feature information, it is necessary to set a reasonable time period. However, it should be noted that for the set feature analysis period, it should be understood as the time when the object can complete a complete action at least once, that is, the time required for the object to walk and complete a swing of the arms and cross the feet. Of course, in order to ensure that the data are highly comparable, the time starting point of the feature analysis period also needs to be reasonably selected to ensure that the coordinated movement of the limbs is consistent. For the feature extraction of limb points, the main consideration is the position of the limb points in the time dimension, and the movement of the limbs relative to the object itself. The position of the limb points in the time dimension is expressed by parameterized vectors, which is more intuitive. Regarding the movement of the limbs relative to the object itself, on the one hand, the body shape of the object represented by the limbs is expressed here using vector modulus. For the upper and lower limbs, the detection points are fixed, and the distance between the detection points and the reference points is stable. It can be determined by extracting the distance size of any k limb points relative to the reference point and averaging it. On the other hand, the position change of the limb points relative to the reference point is expressed by parameterized coordinate vectors.

[0012] As a possible implementation method, the normal posture change data of different objects are clustered based on limb size to form different body shape normal posture change data sets, including: for each object, extracting the corresponding effective left upper limb detection size Effective right upper limb detection size Effective left lower limb detection size And the effective right lower limb detection size Determine the corresponding effective limb size S n ,in, Among them, α is the influencing factor of the upper limb point size, and β is the influencing factor of the lower limb point size; set the cluster analysis number limit P, according to the effective limb size S corresponding to all objects n Perform the following cluster analysis of limb size: Substitute the effective limb size S n Arrange them in order from small to large to form an effective limb size set; determine the effective limb sizes at both ends of the effective limb size set as a small-direction clustering benchmark object and a large-direction clustering benchmark object respectively; extract the effective limb size closest to the small-direction clustering benchmark object in turn, determine the small-direction size difference of the effective limb size relative to the small-direction clustering benchmark object, and obtain the cumulative value of all small-direction size differences to form a small-direction size difference sum value; extract the effective limb size closest to the large-direction clustering benchmark object in turn, determine the large-direction size difference of the effective limb size relative to the large-direction clustering benchmark object, and obtain the cumulative value of all large-direction size differences to form a large-direction size difference sum value; extract the effective limb sizes to form the number of small-direction size difference sum values ​​to ensure that the gap between the small-direction size difference sum value and the large-direction size difference sum value is the smallest, and form a new effective limb size set with the small-direction clustering benchmark object and the corresponding extracted effective limb size set, and form a new effective limb size set with the large-direction clustering benchmark object and the corresponding extracted effective limb size set, and then perform P clustering analyses on the newly formed effective limb size sets to form different effective limb size sets; for each effective limb size set, collect the head benchmark vector of the corresponding object Left upper limb reference vector Right upper limb reference vector Left lower limb reference vector and the right lower limb reference vector Form a corresponding dataset of normal body shape and posture changes.

[0013] In the present invention, a large difference in body shape will result in the loss of more feature information when directly extracting features. Therefore, it is very important to perform reasonable clustering based on body shape. Here, considering that both upper limbs and lower limbs will affect the shape of the object, all limb data are sorted before cluster analysis to form an accurate data reference that is easy to perform clustering. The clustering is performed by arranging the effective limb sizes and then splitting two subsets each time in a way that the distance difference is close. This clustering analysis method will form a sub-data set that is twice as large after each analysis. The limit value P can be set according to the needs of the actual situation, or it can be determined through big data analysis.

[0014] As a possible implementation method, for different body shape normal posture change data sets, the angle change coordination analysis based on different points relative to the reference points is performed to form normal posture angle change coordination data corresponding to different body shape normal posture change data sets, including: for each body shape normal posture change data set, the left upper limb reference vector of each object is determined respectively. Relative head reference vector Angular acceleration change curve of the left upper limb during the characteristic analysis period Right upper limb reference vector Relative head reference vector Angular acceleration change curve of the right upper limb during the characteristic analysis period Left lower limb reference vector Relative head reference vector Angular acceleration change curve of the left lower limb during the characteristic analysis period Right lower limb reference vector Relative head reference vector Angular acceleration change curve of the right lower limb during the characteristic analysis period In the normal posture change data set, all left upper limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the left upper limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, all the right upper limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the right upper limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, all the left lower limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the left lower limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, all the right lower limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the right lower limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, for all head reference vectors Determine the head reference vector The similarity range of the head vector to the standard vertical vector (0, 0, 1) And the head point size variation range m represents the number of the normal posture change data set of different body shapes; the angular acceleration change boundary curve information of the left upper limb, the angular acceleration change boundary curve information of the right upper limb, the angular acceleration change boundary curve information of the left lower limb, and the angular acceleration change boundary curve information of the right lower limb are mapped one by one in the time dimension, and the head vector similarity range is collected And the head point size variation range The coordinated data of normal posture angle changes corresponding to the normal posture change data set of body shape are formed.

[0015] In the present invention, after completing the clustering division, feature extraction of each divided set can more accurately grasp the feature information contained in the set. When extracting feature information of objects in the set, the present application mainly considers three aspects of data. One is the change in the angular acceleration of each limb point relative to the object itself, that is, the change in the angular acceleration of the vector formed by each limb point relative to the head reference vector; the second is the change in the angle between the vector of the head point relative to the origin of the coordinate system and the reference axis of the coordinate system. After all, it is impossible for the object to keep the head reference vector unchanged all the time, but has a certain change and fluctuation, so this change and fluctuation should be reasonably extracted as a basis for detecting and judging whether the posture of the object is normal; the third is that the distance of the head point relative to the reference point is not constant like the limb point. Considering the bendability of the spine and the different postures and movements of different objects, it is necessary to consider the range of distance changes of the head point relative to the reference point, and use it as an important comparison data for whether a fall occurs. Of course, the angular acceleration data related to limb points obtained under big data as comparative expression data of normal posture should be to envelop possible angular acceleration data. Therefore, boundary fitting is to extract the maximum and minimum including boundaries of the limb angular acceleration change curves extracted from different objects, and form a reasonable limb angular acceleration change value range corresponding to normal posture.

[0016] As a possible implementation method, posture-based fall feature analysis is performed based on multi-point detection data of different objects to form fall posture feature data, including: for all objects in the normal posture change data set, the change range of the head point relative to the reference point during the fall cycle is determined. And the similarity range of the head fall reference vector formed by the head point relative to the reference point and the head fall vector of the standard vertical vector (0, 0, 1) Perform boundary fitting on the acceleration change curve of the angle between the left upper limb fall reference vector and the head fall reference vector formed by the left upper limb point position corresponding to each object in the normal body posture change data set relative to the reference point position to form the left upper limb fall angular acceleration change curve information, and compare the left upper limb fall angular acceleration change curve information with the corresponding left upper limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​in the left upper limb fall angular acceleration change curve information that do not belong to the left upper limb angular acceleration change boundary curve information in the time dimension, and form the corresponding left upper limb fall angular acceleration characteristic change curve; Perform boundary fitting on the acceleration change curve of the angle between the right upper limb fall reference vector and the head fall reference vector formed by the right upper limb point position corresponding to each object in the normal body posture change data set relative to the reference point position to form the right upper limb fall angular acceleration change curve information, and compare the right upper limb fall angular acceleration change curve information with the corresponding right upper limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​in the right upper limb fall angular acceleration change curve information that do not belong to the right upper limb angular acceleration change boundary curve information in the time dimension, and form the corresponding right upper limb fall angular acceleration characteristic change curve; Perform boundary fitting on the acceleration change curve of the angle between the left lower limb fall reference vector and the head fall reference vector formed by the left lower limb point position corresponding to each object in the normal body posture change data set relative to the reference point position to form the left lower limb fall angular acceleration change curve information, and compare the left lower limb fall angular acceleration change curve information with the corresponding left lower limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​in the left lower limb fall angular acceleration change curve information that do not belong to the left lower limb angular acceleration change boundary curve information in the time dimension, and form the corresponding left lower limb fall angular acceleration characteristic change curve; The acceleration change curve of the angle between the right lower limb fall reference vector and the head fall reference vector formed by the right lower limb point corresponding to each object in the normal body posture change data set relative to the reference point is subjected to boundary fitting to form the right lower limb fall angular acceleration change curve information, and the right lower limb fall angular acceleration change curve information is compared with the corresponding right lower limb angular acceleration change boundary curve information, and the angular acceleration change values ​​in the right lower limb fall angular acceleration change curve information that do not belong to the right lower limb angular acceleration change boundary curve information are arranged in the time dimension to form the corresponding right lower limb fall angular acceleration characteristic change curve.

[0017] In the present invention, after completing the extraction of normal posture data feature information, it is also necessary to extract the posture feature information during the fall. After all, the body movements of the human body are still diverse, and it is not necessarily judged as a state feature of falling if it is not outside the normal range collected and extracted by big data. The feature information extraction of the falling posture here mainly includes the size change range of the head point relative to the reference point and the angle change range of the vector formed by the head point and the reference point on the same coordinate axis, as well as the angular acceleration change curve data of the limb point. It should be noted that, especially the angular acceleration of the limb point, considering that falling is an action process, the feature data of the angular acceleration of the limb point in the early stage of the fall may have an intersection with the angular acceleration data under the normal posture, so the accurate angular acceleration feature information generated by the falling posture is obtained through comparison processing.

[0018] As a possible implementation method, real-time posture data of the target object is collected, and combined with normal posture feature data and fall posture feature data, real-time fall monitoring and analysis is performed to form real-time fall monitoring and analysis result data, including: performing limb point-based size analysis according to the real-time posture data to determine the corresponding normal posture change data of the body shape, normal posture angle change coordination data and fall posture feature data; performing fall monitoring and analysis on the real-time posture data according to the normal posture angle change coordination data and the fall posture feature data to form real-time fall monitoring and analysis result data.

[0019] In the present invention, after obtaining the limb feature data of the normal posture and the falling posture, the fall can be accurately detected in combination with the real-time detected object posture data. Of course, the real-time monitoring data still needs to be reasonably processed to be compared with the corresponding feature data to ensure the rationality and accuracy of the comparative analysis.

[0020] As a possible implementation method, according to the real-time posture data, a limb point-based size analysis is performed to determine the corresponding body normal posture change data, normal posture angle change coordination data and fall posture feature data, including: determining the real-time left upper limb size of the target object's left upper limb point relative to the reference point within the feature analysis period Real-time right upper limb size relative to the reference point Real-time left lower limb size relative to the reference point And the real-time right lower limb size of the right lower limb point relative to the reference point According to the real-time left upper limb size Real-time right upper limb size Real-time left lower limb size And real-time right lower limb size Determine real-time effective limb size According to the real-time effective limb size T, the effective limb size set corresponding to the real-time effective limb size T is determined; according to the effective limb size set, the corresponding normal posture angle change coordination data and falling posture characteristic data are determined.

[0021] In the present invention, the first point of reasonable analysis of real-time data is to determine the body shape of the object to be analyzed in real time, so as to determine the corresponding feature data according to the body shape for accurate and reasonable comparison. The body shape data of the target object is also determined by the size information of the limb points, and after obtaining the effective limb size, it is determined to which corresponding effective limb size set it belongs.

[0022] As a possible implementation method, according to the normal posture angle change coordination data and the fall posture feature data, the real-time posture data is subjected to fall monitoring analysis to form real-time fall monitoring analysis result data, including: determining the real-time head point size change range Q of the target object's head point relative to the reference point within the feature analysis period brain ; Determine the real-time head point similarity range Q of the target object's head point relative to the standard vertical vector (0, 0, 1) within the feature analysis cycle θ ; Determine the real-time head vector of the target object's head point relative to the basic point within the feature analysis period; determine the real-time angular acceleration change curve of the left upper limb of the target object's left upper limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the right upper limb of the target object's right upper limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the left lower limb of the target object's left lower limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the right lower limb of the target object's right upper limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the right lower limb of the target object's right upper limb point relative to the reference point and the real-time head vector within the feature analysis period; according to the real-time head point size change range Q brain , Real-time head point similarity range Q θ , the real-time angular acceleration change curve of the left upper limb, the real-time angular acceleration change curve of the right upper limb, the real-time angular acceleration change curve of the left lower limb, and the real-time angular acceleration change curve of the right lower limb, and combined with the normal posture angle change coordination data and the falling posture feature data, fall monitoring analysis is performed to form the fall monitoring analysis result data.

[0023] In the present invention, fall detection of the target object requires extraction of real-time feature information that is the same as the feature comparison information. This information includes the real-time distance range of the head point relative to the reference point, the angle range of the vector formed by the head point relative to the reference point and the coordinate axis, and the real-time angular acceleration change curve of each limb.

[0024] As a possible implementation method, according to the real-time head point size change range Q brain , Real-time head point similarity range Q θ , the real-time angular acceleration change curve of the left upper limb, the real-time angular acceleration change curve of the right upper limb, the real-time angular acceleration change curve of the left lower limb and the real-time angular acceleration change curve of the right lower limb, and combined with the normal posture angle change coordination data and the fall posture feature data, the fall monitoring analysis is performed to form the fall monitoring analysis result data, including: when the real-time head point size change range Q brain Belongs to the corresponding head point size variation range Real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range The real-time angular acceleration change curve of the left upper limb does not exceed the range specified by the corresponding left upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the right upper limb does not exceed the range specified by the corresponding right upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the left lower limb does not exceed the range specified by the corresponding left lower limb angular acceleration change boundary curve information, and the real-time angular acceleration change curve of the right lower limb does not exceed the range specified by the corresponding right lower limb angular acceleration change boundary curve information, then a normal action posture prompt information is formed; when the real-time head point size change range Q brain Belongs to the corresponding head point size variation range Real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range If the real-time angular acceleration change curve of the left upper limb exceeds the range specified by the corresponding left upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the right upper limb exceeds the range specified by the corresponding right upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the left lower limb exceeds the range specified by the corresponding left lower limb angular acceleration change boundary curve information, and the real-time angular acceleration change curve of the right lower limb exceeds the range specified by the corresponding right lower limb angular acceleration change boundary curve information, a fall warning information for the action posture is generated; when the real-time head point size change range Q brain The corresponding head point size range For a non-inclusion relationship, the real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range Then the action posture fall warning information is formed; when the real-time head point size change range Q brain Belongs to the corresponding head point size variation range Real-time head point similarity range Q θ Similarity range with the corresponding head vector If it is a non-inclusion relationship, a fall warning information of the action posture is formed.

[0025] In the present invention, here, when performing comparative analysis on fall detection of the target object, the comparison of feature data is sequential, and the data of the head point is the most important data. After all, a large change in the head point data is likely to be a precursor to a fall. Therefore, when making a judgment, the size range and angle range of the head point are first compared and judged. Only when these two data are within the normal range, the limb feature data is further compared. It can be understood that when the size range and angle range of the head point are within the normal range, as long as not all limb feature data exceed the normal range, the action posture can be considered normal. If any of the size range and angle range of the head point is not within the normal range, it is judged that there is a possibility of falling.

[0026] The beneficial effects of the fall detection method based on the cooperation of multi-point posture sensors provided by the present invention are as follows:

[0027] This method obtains multi-point detection data of different objects, performs feature extraction and analysis of normal motion postures, obtains feature data of normal postures, and also relatively extracts posture features in a falling state to form feature data of falling postures. On the basis of fully combining the feature data of normal postures and the feature data of falling postures, a comparative analysis of real-time posture data is performed to quickly and accurately determine whether the current target object has the possibility of falling. This greatly improves the timeliness and accuracy of fall detection, making fall detection analysis more efficient and reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention 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.

[0029] Figure 1 A step diagram of a fall detection method based on cooperation with multi-point posture sensors provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0031] Safety accidents caused by elderly people falling occur frequently, which also poses a huge safety hazard to the lives of the elderly. And with the increase of age, falls will be aggravated due to poor control of limbs, and targeted training and adjustment are needed to improve physical functions. Therefore, fall detection is particularly important. At present, most of the fall detection is still empirically determined through regular inspections, which is highly dependent on the subjective initiative of the test object and cannot achieve accurate and timely fall detection.

[0032] With the development of science and technology, it is possible to establish an analytical evaluation method for fall detection through big data. At present, there is no systematic fall detection analysis based on big data. The STEADI fall prevention risk screening toolkit is mainly used for evaluation. The evaluation coverage is not comprehensive and cannot be personalized for different people. In order to greatly improve the accuracy and timeliness of fall detection, improving the big data fall analysis method has become an important research direction.

[0033] refer to Figure 1 , an embodiment of the present invention provides a fall detection method based on the cooperation of multi-point posture sensors. The method obtains multi-point detection data of different objects, performs feature extraction and analysis of normal movement posture, obtains feature data of normal posture, and also relatively extracts posture features under the fall state to form fall posture feature data. On the basis of fully combining normal posture feature data and fall posture feature data, a comparative analysis of real-time posture data is performed to quickly and accurately determine whether the current target object has the possibility of falling. The timeliness and accuracy of fall detection are greatly improved, making fall detection analysis more efficient and reasonable. Compared with the STEADI anti-fall assessment method, fall monitoring analysis based on big data is more accurate and reasonable.

[0034] The fall detection method based on multi-point posture sensors specifically includes the following steps:

[0035] S1: Acquire multi-point detection data of different objects, perform feature analysis of normal postures, and form normal posture feature data.

[0036] Acquire multi-point detection data of different objects, perform feature analysis of normal posture, and form normal posture feature data, including: for different objects, extract normal posture data of each point in the multi-point detection data, and perform posture change analysis based on the reference point to form normal posture change data of the object; perform clustering based on limb size for normal posture change data of different objects to form normal posture change data sets of different body shapes; perform synergy analysis of angle changes based on different points relative to the reference point for normal posture change data sets of different body shapes to form normal posture angle change synergy data corresponding to normal posture change data sets of different body shapes; and gather all body shape normal posture change data sets and corresponding normal posture angle change synergy data to form normal posture feature data.

[0037] The following aspects are mainly considered in the analysis of normal posture features under multi-point detection data of different objects: On the one hand, since the characteristics of human action behavior are mainly expressed based on limbs, it is necessary to fully extract the posture data of each limb used for analysis when performing fall detection analysis. On the other hand, considering that objects of different body shapes may have large changes in behavior posture due to large differences in body shape whether they are acting normally or falling, in order to obtain more accurate feature data, it is necessary to reasonably divide the posture data of different objects extracted based on the limb conditions, so as to facilitate the subsequent formation of more targeted feature information based on this reasonable division, and ensure a higher accuracy of fall detection. The third aspect is that for a single individual, whether it is a normal posture movement or a posture movement during a fall, the movements of the limbs need to be coordinated. These limb movements are not independent of each other, but need to be considered synchronously and comprehensively when performing fall detection. The coordination between limbs is an important feature information.

[0038] For different objects, the normal posture data of each point in the multi-point detection data is extracted, and the posture change analysis based on the reference point is performed to form the normal posture change data of the object, including: for each object, the trunk point is calibrated as the reference point, the posture space coordinate system is established with the reference point as the origin, and the feature analysis cycle and the size information extraction amount threshold K are set; the left upper limb point is determined based on the left upper limb coordinate position of the posture space coordinate system within the feature analysis cycle And determine the coordinate position of the left upper limb Left upper limb reference vector relative to the reference point in, (X0, Y0, Z0) are the coordinates of the reference point in the posture space coordinate system, and n represents the number of different objects; within the feature analysis cycle, K position points of the left upper limb coordinate position are randomly extracted, and the effective left upper limb detection size of the left upper limb point relative to the reference point is determined Determine the coordinate position of the right upper limb based on the posture space coordinate system during the feature analysis period And determine the coordinate position of the right upper limb Right upper limb reference vector relative to the reference point in, During the feature analysis cycle, randomly extract K position points of the right upper limb coordinate position, and determine the effective right upper limb detection size of the right upper limb point relative to the reference point Determine the left lower limb point position based on the left lower limb coordinate position in the posture space coordinate system within the feature analysis cycle And determine the coordinate position of the left lower limb Left lower limb reference vector relative to the reference point in, During the feature analysis cycle, randomly extract K position points of the left lower limb coordinate position, and determine the effective left lower limb detection size of the left lower limb point relative to the reference point Determine the coordinate position of the right lower limb based on the posture space coordinate system within the feature analysis period And determine the coordinate position of the right lower limb Right lower limb reference vector relative to the reference point in, During the feature analysis cycle, randomly extract K position points of the right lower limb coordinate position, and determine the effective right lower limb detection size of the right lower limb point relative to the reference point. Determine the head point based on the head coordinate position in the posture space coordinate system during the feature analysis cycle And determine the head coordinate position Head reference vector relative to the reference point in,

[0039] The data extraction of normal posture actions includes the movement characteristics of the limbs, including the upper limbs, lower limbs and head. These movement characteristics are not static, but change in a coordinated manner in the time dimension. Therefore, when extracting feature information, it is necessary to set a reasonable time period. However, it should be noted that the set feature analysis period should be understood as the time for the object to complete at least one complete action, that is, the time required for the object to swing its arms and cross its feet. Of course, in order to ensure that the data are highly comparable, the time starting point of the feature analysis period also needs to be reasonably selected to ensure that the coordinated movement of the limbs is consistent. For the feature extraction of limb points, the main consideration is the position of the limb points in the time dimension and the movement of the limbs relative to the object itself. The position of the limb points in the time dimension is expressed by parameterized vectors, which is more intuitive. Regarding the movement of the limbs relative to the object itself, on the one hand, the body shape of the object represented by the limbs is expressed here using vector modulus. For the upper and lower limbs, the detection points are fixed, and the distance between the detection points and the reference points is stable. It can be determined by extracting the distance size of any k limb points relative to the reference point and averaging it. On the other hand, the position change of the limb points relative to the reference point is expressed by parameterized coordinate vectors.

[0040] For the normal posture change data of different objects, clustering based on limb size is performed to form different normal posture change data sets of different body shapes, including: for each object, extracting the corresponding effective left upper limb detection size Effective right upper limb detection size Effective left lower limb detection size And the effective right lower limb detection size Determine the corresponding effective limb size S n ,in, Among them, α is the influencing factor of the upper limb point size, and β is the influencing factor of the lower limb point size; set the cluster analysis number limit P, according to the effective limb size S corresponding to all objects nThe following cluster analysis of limb dimensions is performed: the effective limb dimensions Sn are arranged in order from small to large to form an effective limb dimension set; the effective limb dimensions at both ends of the effective limb dimension set are respectively determined as the small-direction clustering benchmark object and the large-direction clustering benchmark object; the effective limb dimensions closest to the small-direction clustering benchmark object are sequentially extracted to determine the small-direction dimension difference of the effective limb dimension relative to the small-direction clustering benchmark object, and the cumulative value of all small-direction dimension differences is obtained to form the small-direction dimension difference sum value; the effective limb dimensions closest to the large-direction clustering benchmark object are sequentially extracted to determine the large-direction dimension difference of the effective limb dimension relative to the large-direction clustering benchmark object size difference, and obtain the cumulative value of all large-dimension size differences to form a large-dimension size difference sum value; extract the effective limb size to form the number of small-dimension size difference sum values ​​to ensure that the difference between the small-dimension size difference sum value and the large-dimension size difference sum value is minimal, and form a new effective limb size set with the small-dimension clustering benchmark object and the corresponding extracted effective limb size set, and form a new effective limb size set with the large-dimension clustering benchmark object and the corresponding extracted effective limb size set, and then perform P times of clustering analysis on the newly formed effective limb size set to form different effective limb size sets; for each effective limb size set, collect the head benchmark vector of the corresponding object Left upper limb reference vector Right upper limb reference vector Left lower limb reference vector and the right lower limb reference vector Form a corresponding dataset of normal body shape and posture changes.

[0041] Too much difference in body size will cause direct feature extraction to lose a lot of feature information. Therefore, it is very important to perform reasonable clustering based on body size. Here, considering that both upper and lower limbs will affect the shape of the object, all limb data are sorted out before cluster analysis to form an accurate data reference that is easy to cluster. Clustering is done by arranging the effective limb sizes and then splitting two subsets each time in a way that the distance difference is close. This clustering analysis method will form a sub-data set that is twice as large after each analysis. The limit value P can be set according to the actual needs, or it can be determined through big data analysis.

[0042] For different body shape normal posture change data sets, the angle change coordination analysis based on different points relative to the reference points is performed to form normal posture angle change coordination data corresponding to different body shape normal posture change data sets, including: for each body shape normal posture change data set, the left upper limb reference vector of each object is determined respectively. Relative head reference vector Angular acceleration change curve of the left upper limb during the characteristic analysis period Right upper limb reference vector Relative head reference vector Angular acceleration change curve of the right upper limb during the characteristic analysis period Left lower limb reference vector Relative head reference vector Angular acceleration change curve of the left lower limb during the characteristic analysis period Right lower limb reference vector Relative head reference vector Angular acceleration change curve of the right lower limb during the characteristic analysis period In the normal posture change data set, all left upper limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the left upper limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, all the right upper limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the right upper limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, all the left lower limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the left lower limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, all the right lower limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the right lower limb angular acceleration change corresponding to the normal body posture change data set; in the normal body posture change data set, for all head reference vectors Determine the head reference vector The similarity range of the head vector to the standard vertical vector (0, 0, 1) And the head point size variation range m represents the number of the normal posture change data set of different body shapes; the angular acceleration change boundary curve information of the left upper limb, the angular acceleration change boundary curve information of the right upper limb, the angular acceleration change boundary curve information of the left lower limb, and the angular acceleration change boundary curve information of the right lower limb are mapped one by one in the time dimension, and the head vector similarity range is collected And the head point size variation range The coordinated data of normal posture angle changes corresponding to the normal posture change data set of body shape are formed.

[0043] After completing the clustering division, feature extraction for each divided set can more accurately grasp the feature information contained in the set. When extracting feature information for objects in the set, this application mainly considers three aspects of data. One is the change in the angular acceleration of each limb point relative to the object itself, that is, the change in the angular acceleration of the vector formed by each limb point relative to the head reference vector; the second is the change in the angle between the vector of the head point relative to the origin of the coordinate system and the reference axis of the coordinate system. After all, it is impossible for the object to keep the head reference vector unchanged all the time, but has a certain change and fluctuation, so this change and fluctuation should be reasonably extracted as a basis for detecting and judging whether the posture of the object is normal; the third is that the distance of the head point relative to the reference point is not constant like the limb point. Considering the bendability of the spine and the different postures and movements of different objects, it is necessary to consider the range of distance changes of the head point relative to the reference point, and use it as an important comparison data for whether a fall has occurred. Of course, the angular acceleration data related to limb points obtained under big data as comparative expression data of normal posture should be to envelop possible angular acceleration data. Therefore, boundary fitting is to extract the maximum and minimum including boundaries of the limb angular acceleration change curves extracted from different objects, and form a reasonable limb angular acceleration change value range corresponding to normal posture.

[0044] S2: Perform posture-based fall feature analysis based on multi-point detection data of different objects to form fall posture feature data.

[0045] According to the multi-point detection data of different objects, posture-based fall feature analysis is performed to form fall posture feature data, including: for all objects in the normal posture change data set, the head point relative to the reference point in the fall cycle is determined. And the similarity range of the head fall reference vector formed by the head point relative to the reference point and the head fall vector of the standard vertical vector (0, 0, 1) Perform boundary fitting on the acceleration change curve of the angle between the left upper limb fall reference vector and the head fall reference vector formed by the left upper limb point position corresponding to each object in the normal body posture change data set relative to the reference point position to form the left upper limb fall angular acceleration change curve information, and compare the left upper limb fall angular acceleration change curve information with the corresponding left upper limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​in the left upper limb fall angular acceleration change curve information that do not belong to the left upper limb angular acceleration change boundary curve information in the time dimension, and form the corresponding left upper limb fall angular acceleration characteristic change curve; Perform boundary fitting on the acceleration change curve of the angle between the right upper limb fall reference vector and the head fall reference vector formed by the right upper limb point position corresponding to each object in the normal body posture change data set relative to the reference point position to form the right upper limb fall angular acceleration change curve information, and compare the right upper limb fall angular acceleration change curve information with the corresponding right upper limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​in the right upper limb fall angular acceleration change curve information that do not belong to the right upper limb angular acceleration change boundary curve information in the time dimension, and form the corresponding right upper limb fall angular acceleration characteristic change curve; Perform boundary fitting on the acceleration change curve of the angle between the left lower limb fall reference vector and the head fall reference vector formed by the left lower limb point position corresponding to each object in the normal body posture change data set relative to the reference point position to form the left lower limb fall angular acceleration change curve information, and compare the left lower limb fall angular acceleration change curve information with the corresponding left lower limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​in the left lower limb fall angular acceleration change curve information that do not belong to the left lower limb angular acceleration change boundary curve information in the time dimension, and form the corresponding left lower limb fall angular acceleration characteristic change curve; The acceleration change curve of the angle between the right lower limb fall reference vector and the head fall reference vector formed by the right lower limb point corresponding to each object in the normal body posture change data set relative to the reference point is subjected to boundary fitting to form the right lower limb fall angular acceleration change curve information, and the right lower limb fall angular acceleration change curve information is compared with the corresponding right lower limb angular acceleration change boundary curve information, and the angular acceleration change values ​​in the right lower limb fall angular acceleration change curve information that do not belong to the right lower limb angular acceleration change boundary curve information are arranged in the time dimension to form the corresponding right lower limb fall angular acceleration characteristic change curve.

[0046] After completing the extraction of normal posture data feature information, it is also necessary to extract the posture feature information during the fall. After all, the body's limb movements are still diverse, and it is not necessarily judged as a state feature of falling if it is not outside the normal range collected and extracted by big data. The feature information extraction of the falling posture here mainly includes the size change range of the head point relative to the reference point and the angle change range of the vector formed by the head point and the reference point on the same coordinate axis, as well as the angular acceleration change curve data of the limb point. It should be noted that, especially the angular acceleration of the limb point, considering that falling is an action process, the feature data of the angular acceleration of the limb point in the early stage of the fall may intersect with the angular acceleration data under the normal posture, so the accurate angular acceleration feature information generated by the falling posture is obtained through comparison processing.

[0047] S3: Collect the real-time posture data of the target object, and combine the normal posture feature data and the fall posture feature data to perform real-time fall monitoring and analysis to form real-time fall monitoring and analysis result data.

[0048] The real-time posture data of the target object is collected, and combined with the normal posture feature data and the fall posture feature data, real-time fall monitoring and analysis is performed to form real-time fall monitoring and analysis result data, including: performing limb point-based size analysis according to the real-time posture data to determine the corresponding normal posture change data of the body shape, normal posture angle change coordination data and fall posture feature data; performing fall monitoring and analysis on the real-time posture data according to the normal posture angle change coordination data and the fall posture feature data to form real-time fall monitoring and analysis result data.

[0049] After obtaining the limb feature data of normal posture and falling posture, combined with the real-time detected object posture data, the fall detection can be accurately performed. Of course, the real-time monitoring data still needs to be reasonably processed to compare with the corresponding feature data to ensure the rationality and accuracy of the comparative analysis.

[0050] According to the real-time posture data, the size analysis based on the limb points is carried out to determine the corresponding normal posture change data of the body shape, the coordinated data of the normal posture angle change and the characteristic data of the falling posture, including: determining the real-time left upper limb size of the left upper limb point relative to the reference point during the characteristic analysis period of the target object Real-time right upper limb size relative to the reference point Real-time left lower limb size relative to the reference point And the real-time right lower limb size of the right lower limb point relative to the reference point According to the real-time left upper limb size Real-time right upper limb size Real-time left lower limb size And real-time right lower limb size Determine real-time effective limb size According to the real-time effective limb size T, the effective limb size set corresponding to the real-time effective limb size T is determined; according to the effective limb size set, the corresponding normal posture angle change coordination data and falling posture characteristic data are determined.

[0051] The first point for reasonable analysis of real-time data is to determine the body shape of the object being analyzed in real time, so as to determine the corresponding feature data based on the body shape for accurate and reasonable comparison. The body shape data of the target object is also determined by the size information of the limb points, and after obtaining the effective limb size, it is determined to which corresponding effective limb size set it belongs.

[0052] According to the normal posture angle change coordination data and the fall posture feature data, the real-time posture data is subjected to fall monitoring analysis to form the real-time fall monitoring analysis result data, including: determining the real-time head point size change range Q of the target object's head point relative to the reference point within the feature analysis period brain ; Determine the real-time head point similarity range Q of the target object's head point relative to the standard vertical vector (0, 0, 1) within the feature analysis cycle θ ; Determine the real-time head vector of the target object's head point relative to the basic point within the feature analysis period; determine the real-time angular acceleration change curve of the left upper limb of the target object's left upper limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the right upper limb of the target object's right upper limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the left lower limb of the target object's left lower limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the right lower limb of the target object's right upper limb point relative to the reference point and the real-time head vector within the feature analysis period; determine the real-time angular acceleration change curve of the right lower limb of the target object's right upper limb point relative to the reference point and the real-time head vector within the feature analysis period; according to the real-time head point size change range Q brain , Real-time head point similarity range Q θ , the real-time angular acceleration change curve of the left upper limb, the real-time angular acceleration change curve of the right upper limb, the real-time angular acceleration change curve of the left lower limb, and the real-time angular acceleration change curve of the right lower limb, and combined with the normal posture angle change coordination data and the falling posture feature data, fall monitoring analysis is performed to form the fall monitoring analysis result data.

[0053] To detect the fall of the target object, it is necessary to extract real-time feature information that is the same as the feature comparison information. This information includes the real-time distance range of the head point relative to the reference point, the angle range of the vector formed by the head point relative to the reference point and the coordinate axis, and the real-time angular acceleration change curve of each limb.

[0054] According to the real-time head point size change range Q brain , Real-time head point similarity range Q θ , the real-time angular acceleration change curve of the left upper limb, the real-time angular acceleration change curve of the right upper limb, the real-time angular acceleration change curve of the left lower limb and the real-time angular acceleration change curve of the right lower limb, and combined with the normal posture angle change coordination data and the fall posture feature data, the fall monitoring analysis is performed to form the fall monitoring analysis result data, including: when the real-time head point size change range Q brain Belongs to the corresponding head point size variation range Real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range The real-time angular acceleration change curve of the left upper limb does not exceed the range specified by the corresponding left upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the right upper limb does not exceed the range specified by the corresponding right upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the left lower limb does not exceed the range specified by the corresponding left lower limb angular acceleration change boundary curve information, and the real-time angular acceleration change curve of the right lower limb does not exceed the range specified by the corresponding right lower limb angular acceleration change boundary curve information, then a normal action posture prompt information is formed; when the real-time head point size change range Q brain Belongs to the corresponding head point size variation range Real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range If the real-time angular acceleration change curve of the left upper limb exceeds the range specified by the corresponding left upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the right upper limb exceeds the range specified by the corresponding right upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the left lower limb exceeds the range specified by the corresponding left lower limb angular acceleration change boundary curve information, and the real-time angular acceleration change curve of the right lower limb exceeds the range specified by the corresponding right lower limb angular acceleration change boundary curve information, a fall warning information for the action posture is generated; when the real-time head point size change range Q brain The corresponding head point size range For a non-inclusion relationship, the real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range Then the action posture fall warning information is formed; when the real-time head point size change range Q brainBelongs to the corresponding head point size variation range Real-time head point similarity range Q θ Similarity range with the corresponding head vector If it is a non-inclusion relationship, a fall warning information of the action posture is formed.

[0055] Here, when performing comparative analysis on fall detection of the target object, the comparison of feature data is sequential, and the data on the head point is the most important data. After all, a large change in the head point data is likely to be a precursor to a fall. Therefore, when making a judgment, the size range and angle range of the head point are first compared and judged. Only when these two data are within the normal range, the limb feature data is further compared. It can be understood that when the size range and angle range of the head point are within the normal range, as long as not all limb feature data exceed the normal range, the action posture can be considered normal. If any of the size range and angle range of the head point is not within the normal range, it is judged that a fall may occur.

[0056] In summary, the fall detection method based on the cooperation of multi-point posture sensors provided in the embodiment of the present invention has the following beneficial effects:

[0057] This method obtains multi-point detection data of different objects, performs feature extraction and analysis of normal motion postures, obtains feature data of normal postures, and also relatively extracts posture features in a falling state to form feature data of falling postures. On the basis of fully combining the feature data of normal postures and the feature data of falling postures, a comparative analysis of real-time posture data is performed to quickly and accurately determine whether the current target object has the possibility of falling. This greatly improves the timeliness and accuracy of fall detection, making fall detection analysis more efficient and reasonable.

[0058] In the embodiments of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, wherein there is an association between the other information and the information to be indicated. It is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can also be achieved by means of the arrangement order of each piece of information agreed in advance (for example, specified by the protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common parts of each piece of information can also be identified and indicated uniformly to reduce the indication overhead caused by indicating the same information separately.

[0059] In addition, the specific indication method may also be various existing indication methods, such as but not limited to the above-mentioned indication methods and various combinations thereof. The specific details of the various indication methods can refer to the prior art and will not be repeated herein. As can be seen from the above, for example, when it is necessary to indicate multiple information of the same type, different indication methods may be used for different information. In the specific implementation process, the desired indication method can be selected according to specific needs. The embodiment of the present application does not limit the selected indication method. In this way, the indication method involved in the embodiment of the present application should be understood to cover various methods that can enable the party to be indicated to obtain the information to be indicated.

[0060] It should be understood that the information to be indicated can be sent as a whole, or divided into multiple sub-information and sent separately, and the sending period and / or sending time of these sub-information can be the same or different. The specific sending method is not limited in the embodiment of the present application. Among them, the sending period and / or sending time of these sub-information can be pre-defined, for example, pre-defined according to a protocol, or can be configured by the sending end device by sending configuration information to the receiving end device.

[0061] "Pre-definition" or "pre-configuration" can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in the device, and the embodiments of the present application do not limit the specific implementation method. Among them, "saving" can mean saving in one or more memories. The one or more memories can be set separately or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories can also be partially set separately and partially integrated in a decoder, a processor, or a communication device. The type of memory can be any form of storage medium, which is not limited by the embodiments of the present application.

[0062] The "protocol" involved in the embodiments of the present application may refer to a protocol family in the communication field, a standard protocol with a similar protocol family frame structure, or a related protocol used in future communication systems, and the embodiments of the present application do not make specific limitations on this.

[0063] In the embodiments of the present application, descriptions such as "when...", "in the case of...", "if" and "if" all mean that the device will make corresponding processing under certain objective circumstances. It does not limit the time, nor does it require the device to have a judgment action when implementing it, nor does it mean that there are other limitations.

[0064] In the description of the embodiments of the present application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship, for example, A / B can represent A or B; "and / or" in the embodiments of the present application is only a description of the association relationship of the associated objects, indicating that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In addition, in the description of the embodiments of the present application, unless otherwise specified, "multiple" refers to two or more than two. "At least one of the following" or its similar expression refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple. In addition, in order to facilitate the clear description of the technical solution of the embodiments of the present application, in the embodiments of the present application, the words "first" and "second" are used to distinguish the same or similar items with basically the same functions and effects. Those skilled in the art will appreciate that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit the difference. At the same time, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0065] It should be understood that the processor in the embodiment of the present application may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0066] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DRRAM).

[0067] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0068] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0069] In this application, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0070] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0071] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0072] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0073] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0074] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0075] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0076] If the functions are implemented in the form of software functional units 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 and includes several instructions for 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 the various embodiments of the present application. The aforementioned storage media include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0077] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A fall detection method based on multi-point posture sensors, characterized in that: include: Acquire multi-point detection data of different objects, perform feature analysis of normal posture, and form normal posture feature data; Performing posture-based fall feature analysis according to the multi-point detection data of different objects to form fall posture feature data; Collecting real-time posture data of the target object, and combining the normal posture feature data and the fall posture feature data to perform real-time fall monitoring and analysis to form real-time fall monitoring and analysis result data; Among them, obtaining multi-point detection data of different objects, performing feature analysis of normal posture, and forming normal posture feature data include: For different objects, extract the normal posture data of each point in the multi-point detection data, and perform posture change analysis based on the reference point to form normal posture change data of the object; Clustering the normal posture change data of different objects based on limb size to form different body shape normal posture change data sets; For different data sets of normal posture changes of said body shapes, a coordinated analysis of angle changes of different points relative to the reference points is performed to form coordinated data of normal posture angle changes corresponding to different data sets of normal posture changes of said body shapes; Collecting all the normal posture change data sets and the corresponding normal posture angle change collaborative data to form the normal posture feature data; For different objects, the normal posture data of each point in the multi-point detection data is extracted, and the posture change analysis based on the reference point is performed to form the normal posture change data of the object, including: For each object, the torso point is calibrated as the reference point, the posture space coordinate system is established with the reference point as the origin, and the feature analysis period and the size information extraction threshold K are set; Determine the left upper limb point position based on the left upper limb coordinate position of the posture space coordinate system within the feature analysis period And determine the coordinate position of the left upper limb The left upper limb reference vector relative to the reference point in, (X0, Y0, Z0) are the coordinates of the reference point in the posture space coordinate system, and n represents the number of different objects; During the feature analysis period, K position points of the left upper limb coordinate position are randomly extracted, and the effective left upper limb detection size of the left upper limb point relative to the reference point is determined. Determine the right upper limb point position based on the right upper limb coordinate position of the posture space coordinate system within the feature analysis period And determine the coordinate position of the right upper limb The right upper limb reference vector relative to the reference point in, During the feature analysis period, K position points of the right upper limb coordinate position are randomly extracted, and the effective right upper limb detection size of the right upper limb point relative to the reference point is determined. Determine the left lower limb point position based on the left lower limb coordinate position of the posture space coordinate system within the feature analysis period And determine the coordinate position of the left lower limb The left lower limb reference vector relative to the reference point in, During the feature analysis period, K position points of the left lower limb coordinate position are randomly extracted, and the effective left lower limb detection size of the left lower limb point relative to the reference point is determined. Determine the right lower limb point position based on the right lower limb coordinate position of the posture space coordinate system within the feature analysis period And determine the coordinate position of the right lower limb The right lower limb reference vector relative to the reference point in, In the feature analysis cycle, K position points of the right lower limb coordinate position are randomly extracted, and the effective right lower limb detection size of the right lower limb point relative to the reference point is determined. Determine the head point based on the head coordinate position of the posture space coordinate system within the feature analysis period And determine the head coordinate position The head reference vector relative to the reference point in, The normal posture change data of different objects are clustered based on limb size to form different body shape normal posture change data sets, including: For each object, extract the corresponding effective left upper limb detection size The effective right upper limb detection size The effective left lower limb detection size And the effective right lower limb detection size Determine the corresponding effective limb size S n ,in, Among them, α is the influencing factor of the upper limb point size, and β is the influencing factor of the lower limb point size; Set the cluster analysis limit P according to the effective limb size S corresponding to all objects n Cluster analysis of the following limb dimensions was performed: The effective limb size S n Arrange them in order from smallest to largest to form a set of effective limb sizes; Determine the effective limb sizes at both ends of the effective limb size set as a small-direction clustering reference object and a large-direction clustering reference object respectively; Sequentially extract the effective limb size closest to the small-direction clustering reference object, determine the small-direction size difference of the effective limb size relative to the small-direction clustering reference object, and obtain the cumulative value of all the small-direction size differences to form a small-direction size difference sum value; Sequentially extract the effective limb sizes closest to the large-direction clustering reference object, determine the large-direction size differences of the effective limb sizes relative to the large-direction clustering reference object, and obtain the cumulative values ​​of all the large-direction size differences to form a large-direction size difference sum value; Extracting the effective limb dimensions to form the number of small dimension difference sum values ​​ensures that the difference between the small dimension difference sum value and the large dimension difference sum value is minimal, and forming a new effective limb dimension set with the small dimension clustering reference object and the corresponding extracted effective limb dimension set, and forming a new effective limb dimension set with the large dimension clustering reference object and the corresponding extracted effective limb dimension set, and then performing P times of clustering analysis on the newly formed effective limb dimension sets to form different effective limb dimension sets; For each of the effective limb size sets, the head reference vector of the corresponding object is collected The left upper limb reference vector The right upper limb reference vector The left lower limb reference vector And the right lower limb reference vector Forming a corresponding data set of normal body posture changes; For different data sets of normal posture changes in body shapes, a synergy analysis of angle changes of different points relative to reference points is performed to form normal posture angle change synergy data corresponding to different data sets of normal posture changes in body shapes, including: For each of the normal posture change data sets, the left upper limb reference vector of each object is determined respectively. Relative to the head reference vector The left upper limb angular acceleration change curve during the characteristic analysis period The right upper limb reference vector Relative to the head reference vector The right upper limb angular acceleration change curve during the characteristic analysis period The left lower limb reference vector Relative to the head reference vector The angular acceleration change curve of the left lower limb during the characteristic analysis period The right lower limb reference vector Relative to the head reference vector The angular acceleration change curve of the right lower limb during the characteristic analysis period In the normal posture change data set, for all the left upper limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the left upper limb angular acceleration change corresponding to the normal posture change data set; In the normal posture change data set, for all the right upper limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the angular acceleration change of the right upper limb corresponding to the normal posture change data set of the body shape; In the normal posture change data set, for all the left lower limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the left lower limb angular acceleration change corresponding to the normal posture change data set; In the normal posture change data set, for all the right lower limb angular acceleration change curves Perform boundary fitting to determine the boundary curve information of the angular acceleration change of the right lower limb corresponding to the normal posture change data set of the body shape; In the body shape normal posture change data set, for all the head reference vectors Determine the head reference vector The similarity range of the head vector to the standard vertical vector (0, 0, 1) And the size range of head point m represents the number of the normal posture change data set of different body shapes; The left upper limb angular acceleration change boundary curve information, the right upper limb angular acceleration change boundary curve information, the left lower limb angular acceleration change boundary curve information, and the right lower limb angular acceleration change boundary curve information are mapped one by one in the time dimension, and the head vector similarity range is collected And the head point size variation range The normal posture angle change coordination data corresponding to the normal posture change data set of the body shape is formed.

2. The fall detection method based on multi-point posture sensors according to claim 1 is characterized in that: The step of performing posture-based fall feature analysis based on the multi-point detection data of different objects to form fall posture feature data includes: For all objects in the body shape normal posture change data set, determine the head point fall size change range relative to the reference point during the fall cycle And the similarity range of the head fall reference vector formed by the head point relative to the reference point and the head fall vector of the standard vertical vector (0, 0, 1) Perform boundary fitting on the acceleration change curve of the angle between the left upper limb fall reference vector and the head fall reference vector formed by the left upper limb point corresponding to each object in the normal body shape posture change data set relative to the reference point to form left upper limb fall angular acceleration change curve information, and compare the left upper limb fall angular acceleration change curve information with the corresponding left upper limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​of the left upper limb fall angular acceleration change curve information that do not belong to the left upper limb angular acceleration change boundary curve information in the time dimension, and form the corresponding left upper limb fall angular acceleration characteristic change curve; Perform boundary fitting on the acceleration change curve of the angle between the right upper limb fall reference vector and the head fall reference vector formed by the right upper limb point corresponding to each object in the normal body shape posture change data set relative to the reference point to form right upper limb fall angular acceleration change curve information, and compare the right upper limb fall angular acceleration change curve information with the corresponding right upper limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​of the right upper limb fall angular acceleration change curve information that do not belong to the right upper limb angular acceleration change boundary curve information in the time dimension, and form the corresponding right upper limb fall angular acceleration characteristic change curve; Perform boundary fitting on the acceleration change curve of the angle between the left lower limb fall reference vector and the head fall reference vector formed by the left lower limb point corresponding to each object in the normal body shape posture change data set relative to the reference point to form left lower limb fall angular acceleration change curve information, and compare the left lower limb fall angular acceleration change curve information with the corresponding left lower limb angular acceleration change boundary curve information, arrange the angular acceleration change values ​​of the left lower limb fall angular acceleration change curve information that do not belong to the left lower limb angular acceleration change boundary curve information in the time dimension, and form the corresponding left lower limb fall angular acceleration characteristic change curve; Boundary fitting is performed on the acceleration change curve of the angle between the right lower limb fall reference vector and the head fall reference vector formed by the right lower limb point corresponding to each object in the normal body posture change data set relative to the reference point to form the right lower limb fall angular acceleration change curve information, and the right lower limb fall angular acceleration change curve information is compared with the corresponding right lower limb angular acceleration change boundary curve information, and the angular acceleration change values ​​of the right lower limb fall angular acceleration change curve information that do not belong to the right lower limb angular acceleration change boundary curve information are arranged in the time dimension to form the corresponding right lower limb fall angular acceleration characteristic change curve.

3. The fall detection method based on multi-point posture sensors according to claim 2 is characterized in that: The real-time posture data of the target object is collected, and real-time fall monitoring and analysis is performed in combination with the normal posture feature data and the fall posture feature data to form real-time fall monitoring and analysis result data, including: According to the real-time posture data, a dimensional analysis based on limb points is performed to determine the corresponding normal posture change data of the body shape, the coordinated data of the normal posture angle change, and the characteristic data of the falling posture; According to the normal posture angle change coordination data and the fall posture characteristic data, the real-time posture data is subjected to fall monitoring analysis to form the real-time fall monitoring analysis result data.

4. The fall detection method based on multi-point posture sensors according to claim 3 is characterized in that: The method of performing a limb point-based size analysis based on the real-time posture data to determine the corresponding body normal posture change data, the normal posture angle change coordination data, and the falling posture feature data includes: Determine the real-time left upper limb size of the target object's left upper limb point relative to the reference point during the feature analysis period Real-time right upper limb size relative to the reference point Real-time left lower limb size relative to the reference point And the real-time right lower limb size of the right lower limb point relative to the reference point According to the real-time left upper limb size The real-time right upper limb size The real-time left lower limb size And the real-time right lower limb size Determine real-time effective limb size According to the real-time effective limb size T, determining an effective limb size set corresponding to the real-time effective limb size T; According to the effective limb size set, the corresponding normal posture angle change coordination data and the falling posture characteristic data are determined.

5. The fall detection method based on multi-point posture sensors according to claim 4 is characterized in that: The performing fall monitoring analysis on the real-time posture data according to the normal posture angle change coordination data and the fall posture feature data to form the real-time fall monitoring analysis result data includes: Determine the real-time head point size change range Q of the target object's head point relative to the reference point within the feature analysis cycle brain ; Determine the real-time head point similarity range Q of the target object's head point relative to the standard vertical vector (0, 0, 1) within the feature analysis cycle θ ; Determine a real-time head vector of the target object's head point relative to the basic point within a feature analysis period; Determine a left upper limb real-time angular acceleration change curve of a real-time left upper limb vector formed by a left upper limb point position of the target object relative to a reference point position and the real-time head vector within a feature analysis period; Determine a right upper limb real-time angular acceleration change curve of a real-time right upper limb vector formed by a right upper limb point position of the target object relative to a reference point position and the real-time head vector within a feature analysis period; Determine a left lower limb real-time angular acceleration change curve of a real-time left lower limb vector formed by a left lower limb point position of the target object relative to a reference point position and the real-time head vector within a feature analysis period; Determine a real-time angular acceleration change curve of the right lower limb formed by the real-time right upper limb vector of the right upper limb point of the target object relative to the reference point and the real-time head vector within the feature analysis period; According to the real-time head point size change range Q brain , the real-time head point similarity range Q θ , the real-time angular acceleration change curve of the left upper limb, the real-time angular acceleration change curve of the right upper limb, the real-time angular acceleration change curve of the left lower limb and the real-time angular acceleration change curve of the right lower limb, and combined with the normal posture angle change coordination data and the fall posture feature data, perform fall monitoring analysis to form fall monitoring analysis result data.

6. The fall detection method based on multi-point posture sensors according to claim 5 is characterized in that: According to the real-time head point size change range Q brain , the real-time head point similarity range Q θ , the real-time angular acceleration change curve of the left upper limb, the real-time angular acceleration change curve of the right upper limb, the real-time angular acceleration change curve of the left lower limb, and the real-time angular acceleration change curve of the right lower limb, and in combination with the normal posture angle change coordination data and the fall posture feature data, perform fall monitoring analysis to form fall monitoring analysis result data, including: When the real-time head point size changes within the range Q brain The corresponding head point size variation range The real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range The real-time angular acceleration change curve of the left upper limb does not exceed the range defined by the corresponding left upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the right upper limb does not exceed the range defined by the corresponding right upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the left lower limb does not exceed the range defined by the corresponding left lower limb angular acceleration change boundary curve information, and the real-time angular acceleration change curve of the right lower limb does not exceed the range defined by the corresponding right lower limb angular acceleration change boundary curve information, then a normal action posture prompt information is formed; When the real-time head point size changes within the range Q brain The corresponding head point size variation range The real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range If the real-time angular acceleration change curve of the left upper limb exceeds the range specified by the corresponding left upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the right upper limb exceeds the range specified by the corresponding right upper limb angular acceleration change boundary curve information, the real-time angular acceleration change curve of the left lower limb exceeds the range specified by the corresponding left lower limb angular acceleration change boundary curve information, and the real-time angular acceleration change curve of the right lower limb exceeds the range specified by the corresponding right lower limb angular acceleration change boundary curve information, then an action posture fall warning information is generated; When the real-time head point size changes within the range Q brain The corresponding head point size variation range is a non-inclusion relationship, the real-time head point similarity range Q θ Belongs to the corresponding head vector similarity range Then the action posture fall warning information is formed; When the real-time head point size changes within the range Q brain The corresponding head point size variation range The real-time head point similarity range Q θ The similarity range with the corresponding head vector If it is a non-inclusion relationship, a fall warning information of the action posture is formed.

Images