Gesture Rocognition Technology Research And Application On Android Platform

In recent years, with the rapid development of mobile Internet industry, the number of mobile devices which are equipped with mobile operating systems is also rapidly increasing. In a variety of today’s popular mobile operating system, Android system, by virtue of its openness, abundant applications and other advantages, is growing fast.However, human-computer interaction on the Android platform is still based on the style of touch-screen operation, and there is no much innovative interaction.This thesis mainly studies how to use the gesture recognition technology to the Android platform, which can provide a new way for human-computer interaction on Android platform. Through the study of gesture detection, gesture recognition algorithm,the thesis finally develops a gesture recognition APP on the Android platform, which can recognize six specific gestures. The APP can easily extend to other application on the Android platform. The mainly research work is as follows:1) On the gesture detection step, this thesis uses the Viola-Jones object detection algorithms to detect gestures, and chooses LBP feature to train the gesture classifier, for the LBP feature can describe local texture features effectively. Viola-Jones detection framework has been widely applied in the field of face recognition, but rarely in the gesture detection applications. Compared to the Haar feature, which is used in the Voila-Jones framework, LBP feature has almost the same assuracy, while using less time,so LBP feature is more suitable for Android platform.2) On the gesture recognition step, this thesis uses the detected gesture gesture region to find its contour information, using the way of skin-area segmentation, edge detection and contour extraction successively. Then, uses the template matching algorithm to recognize the gesture, while chooses the Hu moments as the feature, which is distance invariance, rotational invariance and scale invariance.3) By using the algorithm previously described, the writer programs to realise them to recognize six specific gestures. Then use NDK tool to encapsulate them to get the JNI interface, which can be used on the Android platform. And by using the gesture recognition interface, the thesis develops an APP on the Android platform, which shows as a floating window. The APP can send the gesture recognition information to other applications, using broadcast. Finally, the APP shows an example of using the APP tocontrol an image explorer. On the ZTE U950 mobile phone, which has a single-core1 GHz processer, and the Android system version is 4.0.4, the average time of processing(from the input to identifying the gesture) one frame, which size is 200 * 160, is 93 ms.