A hand gesture recognition to control wheelchair to help People with special needs

Abstract

 Accelerometer sensor to take several different readings to obtain a quantity of information until the machine recognizes the various movements by defining the X, Y, and Z coordinates. Collects information about the location of X, Y, and Z coordinates during the movement. Accelerometer sensors convert motion into electrical signal so, signal be suitable for measure. The sensor makes several movements so, that Edge Impulses can recognize it. The motion must be repeated several times so, Edge Impulse can recognize it. Every two minutes we repeat the movement. The more attempts which do it more exact results we get. All the data which measure by Accelerometer sensor be store it in Edge impulse. In Edge Impulse there are a lot of steps I should go through it. Data acquisition use to collect data. Then create impulse by add processing block choose spectrogram. Next, adding Learning block then choose Neural Network (Keras). It can generate feature by select MFCC. It will show 3D shape with different colour for x, y, and z. NN classifier use to training output. It is showing percentage of accuracy. In Model testing show green colour if identify colour if it is unknown will be shown orang colour. Last one is  Deployment. It used to build on device model. After the machine recognizes several movements, so can build model. Then program it by using a microcontroller to do three different action.The importance of this project lies in the use of machine learning to identify hand movement.