Nonspecific-user hand gesture recognition by using MEMS accelerometer

Abstract

Hand gestures are a form of nonverbal communication, which allow a person to communicate a range of thoughts and feelings with or without speech. Here MEMS 3 axis accelerometer to detect the input gestures as X, Y, Z direction. The axis is to detect the four types of gesture, which includes up, down, left, right. The hand motion of data collected will directly send to the microcontroller to run on a PC with the help of wireless module. The data will compressed by different users, gesture to extract from sign sequence and template matching. A single gesture has contain an 8 numbers code. This code reduces the hundreds of data values in single gesture and also to compare with the stored templates. In this paper three models has introduced and discussed about its accuracy. The sequence of gesture contain 85 experiments, finally the results achieves an overall accuracy of 96% based on the sign sequence generation and template matching, this each recognition contains the ranging from 94% to 100%.