Handwriting Analysis Using IMU For Small Handwriting Size

Amran, Fatin Afiera (2017) Handwriting Analysis Using IMU For Small Handwriting Size. Project Report. UTeM, Melaka. (Submitted)

Text Handwriting Analysis Using IMU For Small Handwriting Size - Fatin Afiera Amran - 24 pages.pdf - Submitted Version Download (642kB)

Abstract

Handwritten character recognition is one of the example of constant development of computer tools leads to a requirement of easier interfaces between man and the computer. Handwritten character recognition is one of the computer’s ability where it can receive and interpret the handwritten input data from source as the document and also can transform it to machine readable and editable format. In this project the system was implemented using MATLAB software tools. There are several steps involved in this system including feature generation, feature selection and extraction and data classification. In signal preprocessing, there are some procedures involved which are calibration, filtering and normalization. During the preprocessing phase, the errors are filtered. During the feature generation phase, some formulas involved to process the acceleration include mean and root mean square (RMS). The purpose of feature selection and extraction is to increase the accuracy of the classification. Artificial Neural Network is recommended as the classifier for handwritten digit and hand gesture recognition in this project. Neural Network can make classification decision accurately and has advantage in high speed of learning. After completing the feature generation phase, the data classification process will take place and the extracted handwriting character will be classified. By applying the equation of mean, alphabet ‘i’ shows the highest accuracy which is 96.71% followed by alphabet ‘u’ that has accuracy of 90.71%. Alphabets ‘a’, ‘e’ and ‘o’ have almost same accuracy which are 82.81%, 83.72% and 85.13%. The accuracy depends on the acceleration and the hand gesture of the writing. After implementing the neural network, the overall performance of the digital pen achieved is 57%.

Actions (login required)

View Item