The Study Of Gait Recognition Using Kinematic Acceleration Signal

Gait recognition refers to automatic identification of an individual based on his/her walking style. As one of biometrics, gait has the unique advantages of being distant recognition, non-invasive, difficult to conceal and obtained without a walker's attention. So the research of human gait, as well as its development as a biometric for identification purpose is currently an active research area. A lot of research on the gait recognition has been done in the domestic and foreign, but the research is just at beginning, and there is no prefect product.This thesis presents an improved method for gait recognition based on 3D kinematic acceleration signal. The kinematic acceleration data is collected using a portable device which can be attached to a person's waist, where it detects the waist's movement in horizontal, vertical and sideway direction as the person walks. The data is used in an attempt to authenticate the walking person.At first, the state of the art of gait recognition is studied. Then this thesis presents an effective method for wireless gathering kinematic acceleration signal, and designs the wireless gathering system for kinematic acceleration signal, which is composed of the 3D accelerometer MMA7260 and CC1010, a RF transceiver chip with 8051 compatible microcontroller. The wireless gathering system is mainly made up of the transmitting module and the receiving module. In the transmitting module, after capturing the 3D kinematic acceleration signal through MMA7260, the signal is sampled and converted to the characteristic data via the built-in ADC of CC1010. Then the gait characteristic data is wirelessly transmitted. At the same time the receiving module wirelessly receives the data and sends them to the computer through serial port. Then the kinematic acceleration data is processed in real time.The thesis creates a database of kinematic acceleration using the wireless gathering system. The database contains 36 test subjects and 180 groups of data.The thesis discusses a method of gait recognition based on the nearest neighbor algorithm. After the preprocessing of the kinematic acceleration data, such as denoising and normalizing, the thesis extracts gait feature of time domain and frequency domain and applies two kinds of algorithms, time domain algorithm and frequency domain algorithm, to verify his/her identity. Then the equal error rates of 12.6% and 22.2% are achieved, respectively.The method presented differs significantly from the mainstream gait recognition research relying on computer vision or sensors installed in the floor. The gait recognition based on the kinematic acceleration signal is proved to have certain recognition ability, which is a new attempt in the biometric recognition domain, is of great practical value and research significance.