CN-121987189-A - Gait stability detection method based on multicellular body member-set estimation

CN121987189ACN 121987189 ACN121987189 ACN 121987189ACN-121987189-A

Abstract

The invention relates to the technical field of gait stability evaluation, in particular to a gait stability detection method based on multicellular body member collection estimation, which comprises the steps of constructing a stability area template based on multicellular bodies offline based on gait data of normal walking of a subject; collecting real-time gait data of a subject, adopting an ensemble member filtering algorithm, deducing a prediction state set at the current moment by using a posterior state set, carrying out measurement constraint on the prediction state set based on actual measurement data, and carrying out online recursion on a state feasible set, carrying out ensemble inclusion test on the state feasible set and a stability region template, judging whether the current gait is stable or not, calculating a risk score according to the distance between the state feasible set and the stability region template, and generating real-time early warning. Compared with the prior art, the invention can provide more accurate fall risk assessment, and obviously improve the safety and life quality of the old people group and the special group.

Inventors

ZHU YANFEI
LI HENG
LI CHUANJIANG
HE XUEFENG

Assignees

上海师范大学

Dates

Publication Date: 20260508
Application Date: 20260325

Claims (8)

1. The gait stability detection method based on the multicellular body member estimation is characterized by comprising the following steps: constructing a stability area template based on multicellular bodies offline based on gait data of normal walking of a subject; Collecting real-time gait data of a subject, adopting an ensemble member filtering algorithm, deducing a prediction state set at the current moment by using a posterior state set, and carrying out measurement constraint on the prediction state set based on actual measurement data to recursively construct a state feasible set on line; And carrying out set inclusion test on the state feasible set and the stability region template, judging whether the current gait is stable or not, calculating a risk score according to the distance between the state feasible set and the stability region template, and generating real-time early warning.
2. The method for detecting gait stability based on the multicellular carcinoma estimation according to claim 1, wherein the construction of the stability area template based on the multicellular carcinoma is specifically as follows: Detecting an initial landing event, dividing gait cycles, carrying out time normalization on each gait cycle, and resampling equal proportion of each gait cycle to a fixed gait phase point; For each gait phase point, a sliding window is adopted to aggregate the motion state samples of the gait cycle at the moment of the gait phase point, and the augmentation state vector is split into two-dimensional subspaces, namely a front-rear direction position-velocity space and a lateral position-velocity space; Defining a plurality of direction vectors in a two-dimensional subspace, projecting all motion state sample points in a window to each direction vector, and taking a set statistical score number of projection data as a boundary point in a corresponding direction; the boundary points in all directions are enveloped to obtain a convex polygon which covers all boundary points, a transfer function is called to fit a convex polygon top point set into a multicellular structure, and the center point and a generating matrix of the multicellular structure in a two-dimensional subspace are calculated; For each phase point in the gait cycle, a set of smoothed time-varying multicellular parameters is output, constituting a time-varying stability region template for the subject.
3. The gait stability detection method based on the multicellular body assembler estimation according to claim 1, wherein the predicting state set at the current moment is deduced by using the posterior state set, specifically as follows: By using the posterior state set at the previous moment and combining the state transition matrix and the boundary range of process noise, the prediction state set at the current moment is deduced : In the formula, , Respectively predicting the center of the state set multicellular body and generating a matrix at the current moment; , The center and the generation matrix of the multicellular body in the known state of the last time step respectively; is a state transition matrix; Is a control matrix; Is a control input; is process noise; The impact matrix of control inputs and process noise, respectively.
4. The gait stability detection method based on the multicellular aggregate estimation of claim 1, wherein the measurement constraint is performed on the prediction state set based on actual measurement data, specifically as follows: when the actual gait measurement data of the current moment is obtained, the prediction state set of the current moment Measurement consistency set of current measurement data Taking intersection sets to obtain a feasible state set The set of measurement consistency is a set of all states x that bring the measurement residuals within a bounded set of measurement noises.
5. The method of claim 4, wherein for the intersection-taking operation, a linear outer envelope method is used to calculate a conservative multiple in vitro approximation set containing intersections : Calculating gain matrix : Using actual measurements And prediction state multicellular center Correcting deviation of the updated state feasible set multicellular body center : Updating a generator matrix of a state feasible set: Wherein, the Generating a matrix for a prediction state set at the current moment; measuring the center of the noise multicellular body and generating a matrix respectively; is an observation matrix; The method is to prevent micro regularization parameters of matrix singular; Is an identity matrix.
6. The method for detecting gait stability based on multi-cell soma member estimation according to claim 4 or 5, wherein the maximum order is set for iterative concatenation of generator matrices Calculating L 2 norms of each generated matrix array vector in each iteration, and reserving the front with the maximum L 2 norms A matrix is generated.
7. The gait stability detection method based on the multiple cell population estimation of claim 1, wherein the determining whether the current gait is stable is specifically as follows: the feasible set of states obtained by online calculation Stability zone template corresponding to phase Performing a set inclusion test to determine whether the state feasible set is completely included in the stable region; If the current state feasible set is completely contained in the stability area template, the gait is considered to be in a stable state; Otherwise, calculating the minimum Euclidean distance between the multicellular body center of the current state feasible set and the stability area template by solving the quadratic programming problem, and taking the minimum Euclidean distance as a risk score.
8. The method for detecting gait stability based on the multi-cell collector estimation of claim 7, wherein the quadratic programming problem starts with the center of the current state viable set of multi-cells, the stability area template is the end point, and the risk score The calculation formula of (2) is as follows: The constraint conditions are as follows: in the formula, Template representing stability region Any point inside; the center of the multicellular body is a feasible set in the current state; , Respectively, stability zone templates The center of multicellular bodies and a generating matrix; Is a coefficient vector.

Description

Gait stability detection method based on multicellular body member-set estimation Technical Field The invention relates to the technical field of gait stability evaluation, in particular to a method for evaluating gait stability based on Zonotope set member filtering. Background With the advent of the aging society, the problem of elderly falls has become a significant problem affecting health and quality of life. How to realize accurate and real-time monitoring of gait stability in daily life of the elderly and early warning in time before danger occurs becomes a key technical problem to be solved urgently in the fields of intelligent care, rehabilitation engineering and man-machine interaction. Conventional mechanical evaluation methods based on fixed thresholds generally perform the determination by extracting gait parameters (such as step size, pace, trajectory, etc.), and setting fixed empirical thresholds. However, for the elderly with different heights, weights and health conditions, the normal gait parameters of the elderly have great differences, and the simple fixed threshold value often leads to high misjudgment rate. In addition, the method usually only focuses on static or quasi-static indexes, can not effectively capture stability boundaries under dynamic disturbance in the walking process, and is difficult to adapt to complex and changeable real walking scenes. A statistical estimation method based on a probability model. As machine learning techniques develop, some studies have attempted to model gait data using Hidden Markov Models (HMMs), gaussian Mixture Models (GMMs), or Deep Neural Networks (DNNs). The method depends on a large amount of labeling data, and achieves a certain effect in a controlled environment, but has the bottleneck that in practical application, robustness is insufficient, sensor noise, uneven ground, burst interference and the like in a real environment are often non-Gaussian and limited uncertainties, a probability model is difficult to accurately describe the extreme conditions, and an evaluation model is easy to destabilize. The calculation complexity is high, while the accuracy of complex models such as a deep neural network is high, the reasoning time is long, and the millisecond real-time requirement for fall early warning is difficult to meet. The method is lack of interpretability, namely the output of the black box model only gives a stable/unstable binary conclusion, the specific degree of deviation from a stable region cannot be quantitatively evaluated, and the physical basis and the guiding significance for providing risk early warning are lacked. Therefore, a gait stability detection method is needed to solve the problems of low accuracy, poor real-time performance, inability to deal with dynamic uncertainty and the like of the existing gait stability assessment method. Disclosure of Invention The invention aims to solve the problems that the existing gait stability assessment method is low in precision, poor in real-time performance, incapable of processing dynamic uncertainty and the like, and provides a gait stability detection method based on multicellular body member-gathering estimation. The aim of the invention can be achieved by the following technical scheme: As a first aspect of the present invention, there is provided a gait stability detection method based on a multicellular panelist estimation, the steps including: constructing a stability area template based on multicellular bodies offline based on gait data of normal walking of a subject; Collecting real-time gait data of a subject, adopting an ensemble member filtering algorithm, deducing a prediction state set at the current moment by using a posterior state set, and carrying out measurement constraint on the prediction state set based on actual measurement data to recursively construct a state feasible set on line; And carrying out set inclusion test on the state feasible set and the stability region template, judging whether the current gait is stable or not, calculating a risk score according to the distance between the state feasible set and the stability region template, and generating real-time early warning. As a preferred technical scheme, the construction of the stability region template based on multicellular bodies is specifically as follows: Detecting an initial landing event, dividing gait cycles, carrying out time normalization on each gait cycle, and resampling equal proportion of each gait cycle to a fixed gait phase point; For each gait phase point, a sliding window is adopted to aggregate the motion state samples of the gait cycle at the moment of the gait phase point, and the augmentation state vector is split into two-dimensional subspaces, namely a front-rear direction position-velocity space and a lateral position-velocity space; Defining a plurality of direction vectors in a two-dimensional subspace, projecting all motion state sample points in a window to each dir