Gesture Recognition Based On Acceleration Sensor

With the continuous improvement of electronic technology, mobile smart devices are becoming an important channel to get information. And with the advance in Micro Electro Mechanical Systems(MEMS) technology, intelligent wearable devices are also presented the outbreak of the trend. Because the intelligent wearable device has advantages of small, portable, low cost, low power consumption, it is becoming an important platform for human-computer interaction. In the conventional mobile intelligent device platform, gesture recognition based on touchscreen, and voice recognition are mature human-computer interaction, but in the wearable device platform, since the size, cost, ease of operation and other restrictions, the human-computer interaction of recognition based on the touch screen and voice exists a great constraints. Gesture is a natural and intuitive way to express, it is a very import means of communication in people’s daily lives, and it is a very suitable way as a human-computer interaction. In recent years, MEMS acceleration sensors are becoming the standard for intelligent wearable device, it provides the material basis for gesture recognition using an acceleration sensor. With the outbreak of wearable devices, gesture recognition based on acceleration sensors will play a huge role in human-computer interaction, sports and health monitoring.This paper designs a gesture recognition system based on accelerometer for wearable device platform, by extracting the key points of information, and using of Error Back propagation neural network technology, and matching the key point feature, achieve higher precision gesture recognition purposes with a smaller computational time and space complexity.First, select the ADXL345accelerometer as the algorithm acceleration data acquisition sources, and using ARM Cortex-MO microprocessor as the platform to achieve the acceleration acquisition system. Then transfer data to computer through the serial port and save documents in txt for MATALB to process.Secondly, by removing the jump point, using sliding mean filter, smooth the acceleration data. By determining the steady state, get the start and end of the gesture recognition data, it can be semi-automatically obtain the true acceleration data of a user’s gesture, without the need for other auxiliary operation. This will make it possible for gesture recognition based on acceleration on a wearable device platform. Finally, using BP neural network algorithm, designed gestures, achieve gesture recognition based on acceleration, and by designing experiments, indicating that the gesture recognition based on the acceleration using BP neural network has a better recognition rate.All in all, smart mobile devices and smart wearable devices is becoming more and more fashion, gesture recognition based on acceleration sensor will be important part of gesture recoginition, it is worthy of further study because it will play an important role in human-computer interaction and user experience.