Mobile phone gyroscope and dynamic time warping-based action gesture identification method

Abstract

The invention provides a mobile phone gyroscope and dynamic time warping (DTW)-based action gesture identification method. The method comprises a gesture template library establishment stage and a gesture identification stage. In the gesture template library establishment stage, for each action gesture, a triaxial angular velocity data sequence in a mobile phone internal gyroscope is collected, starting and ending of a gesture are automatically identified according to an angular velocity mean square error, and gesture data is captured. After the gesture data is subjected to smoothing, normalization, cubic spline interpolation and fake peak valley preprocessing, gesture data samples are formed. A sample with the maximum similarity with other data samples is selected from the data samples ofthe same gesture by adopting a DTW algorithm to serve as a final template of the gesture. In the gesture identification stage, the gesture data is collected and is subjected to DTW matching with templates in a gesture template library, the most similar template is selected, and the gesture corresponding to the most similar template is a current user gesture. Numbers, letters, up, down, left and right, and user-defined gestures can be accurately identified.

Description

technical field [0001] The invention relates to the field of gesture recognition, in particular to a gesture recognition method based on a mobile phone gyroscope and dynamic time warping. Background technique [0002] In recent years, with the development and maturity of Micro-Electro-Mechanical System (MEMS) technology, MEMS sensors with low price, small size and high sensitivity are widely used in various smart phones. The three-axis angular velocity of the mobile phone can be measured through the built-in gyroscope in the smart phone, and human gestures can be recognized through machine learning algorithms. Gesture recognition using sensors is an emerging potential research direction in the field of human-computer interaction and pattern recognition. Compared with traditional mouse, keyboard and other interactive methods, motion gesture recognition is more flexible and free, and the user experience is more in line with human habits. It has the following advantages: (1) m...

AI Technical Summary

Problems solved by technology

There are some problems in the current gesture recognition: (1) The recognition of the start and end of the gesture is inaccurate, resulting in low accuracy of gesture recognition; (2) The shaking of the hand before the start of the gesture is mistakenly regarded as gesture data, which reduces the accuracy of gesture recognition; (3) The user is required to keep the posture of the mobile phone fixed when making gestures, and changes in the way of holding will reduce the recognition accuracy; (4) There is confusion in gesture recognition

Images