System and method for analyzing gait and postural balance of a person
A method and system is provided for finding and analyzing gait parameters and postural balance of a person using a Kinect system. The system is easy to use and can be installed at home as well as in clinic. The system includes a Kinect sensor, a software development kit (SDK) and a processor. The temporal skeleton information obtained from the Kinect sensor to evaluate gait parameters includes stride length, stride time, stance time and swing time. Eigenvector based curvature detection is used to analyze the gait pattern with different speeds. In another embodiment, Eigenvector based curvature detection is employed to detect static single limb stance (SLS) duration along with gait variables for evaluating body balance.
1. A method for determining postural balance of a person, the method comprising processor implemented steps of:
capturing a skeleton data of the person, wherein the skeleton data comprises spatiotemporal variations of an ankle joint data of the person in a x plane, a y plane and a z plane;
removing a plurality of noises from the captured skeleton data using a noise cleaning module, wherein the plurality of noises comprises one or more of errors, and wherein the plurality of noises is removed by:
converting the captured skeleton data into a set of features related to a physical structure of the person;
computing static features of the physical structure in each frame of the captured skeleton data; and
tracking changes in the static features based on a previous and a following frames of each frame of the captured skeleton data;
tracking ankle coordinates of the person from the captured skeleton data only in the x-plane and the y-plane;
determining a plurality of gait parameters of the person using the tracked ankle coordinates of the person by:
obtaining a velocity profile of the tracked ankle coordinates;
computing curvature points based on the velocity profile, wherein the curvature points are two points which indicate a beginning of two consecutive stance phases for a given leg;
detecting peak and trough regions from the velocity profile and obtaining a set of data points in the peak and trough regions;
computing a covariance matrix of standardized data, wherein the standardized data is obtained by calculating mean and standard deviation of the set of data points; and
performing a first eigenvector based curvature analysis by computing an eigenvalue decomposition of the covariance matrix to obtain a curvature point indicative of a beginning of a swing phase for the leg, wherein an eigenvector corresponding to a least eigenvalue provides a direction of minimum variance of the set of data points and reveals a direction towards the curvature point indicative of the beginning of said swing phase; wherein the plurality of gait parameters are determined by using the obtained curvature point indicative of the beginning of said swing phase, the curvature points of the beginning of two consecutive stance phases, and the obtained set of data points in the peak and trough regions of the velocity profile; and
measuring a static single limb stance (SLS) duration of the person by:
capturing variations of a left ankle and a right ankle of the person in the tracked ankle coordinates of the person from the captured skeleton data in the x-plane and the y-plane; and
performing a second eigenvector based curvature analysis on the captured the variations in the tracked ankle coordinates to determine two negative curvature points, wherein a duration between the determined negative curvature points is the static SLS duration.
2. The method of claim 1 , wherein the peak and trough regions are detected by applying a peak and trough detection algorithm to the velocity profile of the person.
3. The method of claim 1 , wherein a k-means clustering algorithm is employed to capture the variations of the left ankle and the right ankle of the person in the tracked ankle coordinates of the person over each frame for measuring the static SLS duration of the person.
4. The method of claim 1 , wherein the plurality of gait parameters include stride length, stride time, swing time and stance time.
5. A system for determining postural balance of a person, the system comprises:
a 3D motion sensor for capturing a skeleton data of the person, wherein the skeleton data comprises spatiotemporal variations of an ankle joint data of the person in a x plane, a y plane and a z plane;
a noise cleaning module for removing a plurality of noises from the captured skeleton data, wherein the plurality of noises comprises one or more of errors, and wherein the plurality of noises is removed by:
converting the captured skeleton data into a set of features related to a physical structure of the person;
computing static features of the physical structure in each frame; and
tracking changes in the static features based on previous and following frames;
a memory; and
a processor in communication with the memory, wherein the processor is further configured to perform the steps of:
tracking ankle coordinates of the person from the captured skeleton data only in the x-plane and the y-plane;
determining a plurality of gait parameters of the person using the tracked ankle coordinates of the person by:
obtaining a velocity profile of the tracked ankle coordinates;
computing curvature points based on the velocity profile, wherein the curvature points are two points which indicate a beginning of two consecutive stance phases for a given leg;
detecting peak and trough regions from the velocity profile and obtaining a set of data points in the peak and trough regions;
computing a covariance matrix of standardized data, wherein the standardized data is obtained by calculating mean and standard deviation of the set of data points; and
performing a first eigenvector based curvature analysis by computing an eigenvalue decomposition of the covariance matrix to obtain a curvature point indicative of a beginning of a swing phase for the leg, wherein an eigenvector corresponding to a least eigenvalue provides a direction of minimum variance of the set of data points and reveals a direction towards the curvature point indicative of the beginning of said swing phase; wherein the plurality of gait parameters are determined by using the obtained curvature point indicative of the beginning of said swing phase, the curvature points of the beginning of two consecutive stance phases, and the obtained set of data points in the peak and trough regions of the velocity profile; and
measuring a static single limb stance (SLS) duration of the person by:
capturing variations of a left ankle and a right ankle of the person in the tracked ankle coordinates of the person from the captured skeleton data in the x-plane and the y-plane; and
performing a second eigen vector based curvature analysis on the captured the variations in the tracked ankle coordinates to determine two negative curvature points, wherein a duration between the determined negative curvature points is the static SLS duration.
6. A non-transitory computer-readable medium having embodied thereon a computer program for determining postural balance of a person, by executing the steps comprising:
capturing a skeleton data of the person, wherein the skeleton data comprises spatiotemporal variations of an ankle joint data of the person in a x plane, a y plane and a z plane;
removing a plurality of noises from the captured skeleton data using a noise cleaning module, wherein the plurality of noises comprises one or more of errors, and wherein the plurality of noises is removed by:
converting the captured skeleton data into a set of features related to a physical structure of the person;
computing static features of the physical structure in each frame; and
tracking changes in the static features based on previous and following frames;
tracking ankle coordinates of the person from the captured skeleton data only in the x-plane and the y-plane;
determining a plurality of gait parameters of the person using the tracked ankle coordinates of the person by:
obtaining a velocity profile of the tracked ankle coordinates;
computing curvature points based on the velocity profile, wherein the curvature points are two points which indicate a beginning of two consecutive stance phases for a given leg;
detecting peak and trough regions from the velocity profile and obtaining a set of data points in the peak and trough regions;
computing a covariance matrix of standardized data, wherein the standardized data is obtained by calculating mean and standard deviation of the set of data points; and
performing a first eigenvector based curvature analysis by computing an eigenvalue decomposition of the covariance matrix to obtain a curvature point indicative of a beginning of a swing phase for the leg, wherein an eigenvector corresponding to a least eigenvalue provides a direction of minimum variance of the set of data points and reveals a direction towards the curvature point indicative of the beginning of said swing phase; wherein the plurality of gait parameters are determined by using the obtained curvature point indicative of the beginning of said swing phase, the curvature points of the beginning of two consecutive stance phases, and the obtained set of data points in the peak and trough regions of the velocity profile; and
measuring a static single limb stance (SLS) duration of the person by:
capturing variations of a left ankle and a right ankle of the person in the tracked ankle coordinates of the person from the captured skeleton data in the x-plane and the y-plane; and
performing a second eigenvector based curvature analysis on the captured the variations in the tracked ankle coordinates to determine two negative curvature points, wherein a duration between the determined negative curvature points is the static SLS duration.