Event-Driven Gait Recognition Method Based on Dynamic Temporal Segmentation

Recognizing human gait with Body Sensor Networks (BSNs) is a significant research in pervasive computing. A real-time gait recognition method driven by leg movements is proposed which uses gyroscopes as main sensors for collecting angular velocities of legs and waist. According to the fluctuation of legs´ angular velocities, sensor data can be segmented into gait cycles. And then from the segmented data in each cycle, a serial of features are extracted which will be given to a classification model for gait recognition. By experimenting four commonly used machine learning algorithms, the best classifier for gait recognition is determined as the final classification model. Experimental results show that our proposed method can recognize 12 kinds of gaits effectively. Compare to other methods, it has the characteristics of more recognizable actions, higher accuracy, better real-time performance and less calculation.