Human Gait Signal Classification Based On Multivariate Multiscale Fuzzy Entropy

Gait signal contains a large number of useful information which can reflect our own health, mental state, on the research of gait signal has always attracts a lot of scholars’ attention. Acceleration reflects an important aspect of gait information, it contains a wealth of information which can reflects the mental state and motion state of the human body. In this paper, the human gait acceleration data is researched, the analysis of human gait acceleration signal can discover the intrinsic differences between normal,abnormal behavior and different gait, the result may be provide an important basis to detection of abnormal gait and evaluation of recovery treatment.After studying domestic and foreign commonly used method of human gait acceleration data, this paper embedded system of STM32f103 MCU as the core, sensor of MPU9150 and embedded the quaternion rotation matrix algorithm into the main chip, completed the human gait acceleration data correction. Through this system, each subjects’ gait acceleration data were collected under different kinds of conditions. It provides a reliable source of data for the human gait analysis.It focuses on the development of multivariate multiscale entropy algorithm, for the shortcomings of traditional multivariate multiscale entropy algorithm. This paper presents multivariate multi-scale fuzzy entropy algorithm. The method improved coarse-grained way of traditional algorithm and make coarse-grained time series equal to the length of original time series on each scale by sliding mean filter, reduce the discrete of multivariate multi-scale entropy. In addition, algorithm maintains the advangtage of tradition method hard threshold, counting the distance of two composite delay vector slightly greater than threshold value by defining fuzzy membership function, not only reducing the dependence of threshold of the traditional algorithm, but also solving the instability caused by traditional threshold. Finally, the algorithm was validated in the simulation data, demonstrate the effectiveness of multi-scale multivariate fuzzy entropy.This article applied multivariate multiscale entropy fuzzy and multivariate multiscale sample entropy to extracting and classifying the feature of collected human gait acceleration and extracting the feature of different behaviors of gait acceleration data in foreign public databases and categorize. The results show that our algorithm is better than traditional methods. Through classification under normal walking gait, upstairs, downstairs gait data set which in the foreign public database, obtained the highest recognition rate reached 96.5%.Compared with the international obtained recognition rate for this data set, further reflecting the superiority of our algorithm. At last, We combine the EMD algorithm with multivariate multiscale fuzzy entropy to extract and analysis the feature of Parkinson’s gait, the young, the elderly gait in the MIT database, the final results fully proved this algorithm not only can extract characteristics of human gait effectively but also has good statistical properties and classification accuracy.