Based On Sensor Gesture Recognition System Design And Realization In The Wearable Device

Since wearable devices have a lot of advantages,such as portable,low energy consumption and low production cost,they have been used in various fields and have obtained more and more people's attraction in recent years.What's more,wearable devices are a significant component of the platform for human-computer interaction(HCI).Dynamic gesture is a natural and efficient way for communication not only between human and human,but also between human and computer.Therefore,dynamic gesture is very suitable for smart wearable devices.Based on the property of dynamic gesture,we propose real-time continuous dynamic gesture recognition system,and realize this system on the wearable device.In this paper,we first focus on the feature extraction of continuous dynamic gesture signal.The feature extraction solutions which fit with the characteristic of dynamic gesture signal are proposed based on three aspects:the continuity of dynamic gesture signal,user variability and duration variability.Then,we study the pre-processing of dynamic gesture signal,which including:amplitude fluctuation based threshold segmentation method,dynamic gesture based low-pass filters and feature extraction method.Second,we study the traditional dynamic time warping(DTW),support vector machine(SVM)and dimension reduction + classification methods,respectively.Based on the limitations of the existing methods,the improved dynamic gesture recognition methods are proposed under different scenario.What's more,the improved DTW method which updates the gallery sets regularly is proposed to improve the recognition rate of dynamic gesture recognition task.Extensive experimental results show the performance of different methods under different scenario.Moreover,these experimental results demonstrate the effective of our proposed methods when it is applied in practical scenario.Finally,different dynamic gesture recognition methods are casted into a unified dynamic gesture recognition framework.We design the software scheme of the whole system and transform this software framework into wearable devices which is also used in the smart-home scenario.Therefore,our framework can achieve the control and management of smart-home by using wearable devices.