Research On Human Balance Analysis Based On Multi-sensor Signal Fusion

Human balance is an important physiological indicator of human body and plays a vital role in people's daily life.Impaired human balance ability can easily lead to physical diseases such as falls,loss of control,dizziness,etc.In particular,falls have become the leading cause of accidental injuries among the elderly,greatly reducing people's quality of life,increasing family burden and social pressure.Therefore,the analysis and evaluation of human balance ability has important research significance for clinical pathological diagnosis and rehabilitation training program.However,there is still no uniform standard for the evaluation of human balance ability at this stage,and further research is needed for the analysis of eigenvalues and evaluation methods reflecting human body balance ability.According to the characteristics of human body balance,this paper designs a multi-sensor based human balance sensing system,that is the pressure sensor senses the position of the human body pressure center,the attitude module's acceleration sensor,gyroscope and angle sensor senses the hip joint attitude data.Then design test experiments from multiple multi-scale angles,provide data support for feature extraction,objective analysis and evaluation of human body balance ability.The human balance signal collected by the experiment is a highly nonlinear and complex time data sequence.In this paper,the maximum Lyapunov exponent and multivariate multi-scale entropy eigenvalue method are used to extract the feature of human balance data based on multi-sensor signals.It is found that the above two methods have the ability to accurately classify the balance of the human body when quantitatively analyzing the balance ability of the human body.Therefore,the method of extracting the human balance ability feature with mixed eigenvalues is proposed.The experimental results show that compared with the maximum Lyapunov exponent and multivariate multi-scale entropy,the mixed eigenvalues have better effect on local quantitative analysis of human body balance ability.In order to further improve the evaluation of balance ability,this paper uses principal component analysis method to fuse data features,and combines K-means cluster analysis and support vector machine to classify experimental signal data.The results show that,based on the eigenvectors composed of principal component analysis and mixed eigenvalues,the support vector machine method can be used to classify and evaluate the balance ability of human body.