Accelerometer Based Digital Pen for Handwritten Digit Recognition

Author(s):

Ms. Veena Ravikiran Sarje , M.G.M's collage of Engineering

Keywords:

Accelerometer, Microcontroller, handwritten Recognition Algorithm, kernel based class separability (KBCS), linear discriminate analysis (LDA), probabilistic neural network (PNN)

Abstract:

In this research we have planned Accelerometer based digital pen for handwritten digit recognition. In future it can be used for security purpose, Vehicle model analysis and Robotic application. This project work includes the status of hand written digits. On-line as well as offline detection is included in this work. It consists of a 3D measuring instrument tool, Arduino Uno and zigbee module to collect accelerations of written digits. The popularity law includes acquisition of acceleration information, signal process, feature generation, feature selection and extraction. The rule is in use to convert time-series acceleration signals to desirable feature vectors. A pen mounted with measuring instrument signals are transferred to computer wirelessly for recognition. The kernel based class separability (KBCS) selects imperative options or features. The reduced options are sending to Linear Discriminative Analysis (LDA) to once more reduce dimension of feature house. Then for final recognition all the information is move to probabilistic neural network (PNN) for final recognition. Experimental outcome have with success valid recognition of numerical digits by digital pen. The system is predicated on AVR ATmega328P-PU microcontroller is integral on Arduino board, Zigbee and recognition algorithm. ADXL335 is employed to capture the motion data. The popularity rule familiar to settle for information from measuring system method and show the information. In future it can be used to detect hand gestures also.

Other Details:

Manuscript Id	:	IJSTEV5I1033
Published in	:	Volume : 5, Issue : 1
Publication Date	:	01/08/2018
Page(s)	:	78-88

Download Article