Research On Gait Recognition Algorithm And Implementation Based On MEMS Sensors

Biological characteristics, such as gait recognition, identification and gait certification, are gaining more and more attention in recent years. With the continuous development and maturity of Micro-Electro-Mechanical System(MEMS), human gait recognition based on the acceleration sensor has become an emerging field of research. It is a process which can realize the qualitative judgment of human gait through the analysis of gait acceleration data. Human gait recognition based on the acceleration sensor has been widely used, which has very broad application prospects in intelligent human-computer interaction, intelligent monitoring, biomedical research, athletic, exercise energy consumption evaluation and many other fields. Therefore, the research on gait recognition based on MEMS sensor has great significance.In order to carry out a more detailed gait classification and improve the gait recognition rate, in this thesis, the algorithm of human gait recognition based on MEMS sensor is improved in following contents:1.This thesis constructed data of accelerometer and gyroscope to support human gait recognition research based on MEMS sensors. We also designed gait data acquisition device with nine axis sensor LSM9DS0, and micro controller STM32F103 used for gait data collection. This device was placed on man's waist which can collect 10 kinds of gait data such as normal walking, fast walking, walking backwards, walking downstairs,walking upstairs, jumping, running, step walking, relaxing and bicycling.2. This thesis has carried out a series of preprocess work on the collected gait data: smooth denoising, normalization, windowing, changing the coordinate system and filtering the gravity component. Based on these data after preprocess work, we use methods of features in time domain(TF), FFT coefficients and DCT coefficients for feature extraction and complete the design of NaiveBayes,C4.5 decision tree and support vector machine(SVM).3. In this thesis, three kinds of features from extraction and three kinds of classifiers we designed are used for comparison of gait recognition research. It is found that the gait recognition rate is highest when FFT coefficients and SVM classification algorithm are used. The recognition rate of running and jumping is 97.14%, and the static recognition rate is 94.29%. Compared with these three movements of high recognition rate, some other activities are difficult to recognize, such as the aliasing between walking upstairs and downstairs, and the aliasing among normal walking, fast walking, and step walking.4.In this thesis, the recognition algorithm is improved in order to reach a more detailed distinction between confused gait. The wrapper feature subset selection algorithm is introduced into the gait recognition algorithm based on MEMS sensor. We select the optimal feature subset by wrapper feature subset selection algorithm and then use SVM classifier for recognition. In this way, the gait recognition rate is increased by 13.15% in average. The aliasing between walking upstairs and downstairs, and the aliasing among step walking,normal walking and fast walking can be effectively distinguished.5.At last, we combine the improved gait recognition algorithm with hardware platform, in order to make the improved algorithm in practical application.This thesis implements a more detail classification of human gait and improved the recognition rate. The improved algorithm is superior to the traditional methods. The research of this subject has important theoretical value and application value, and it is worth to be continued both in detail and depth.