Human Behavior Prediction through Handwriting Analysis Abhishek Biradar 1, Priyanka Humbre 2, Kartik Jagdale 3, Savita Phale 4, Akash Shelke 5 (Computer,JSPM S Rajarshi Shahu College of Engineering/ Pune University, India) Abstract :-Handwriting Analysis or Graphology is a scientific method of identifying, evaluating and understanding personality through the strokes and patterns revealed by handwriting. Handwriting reveals the true personality including emotional outlay, fears, honesty, defenses and many others. Professional handwriting examiners called graphologist often identify the writer with a piece of handwriting. Accuracy of handwriting analysis depends on how skilled the analyst is. Although human intervention in handwriting analysis has been effective, it is costly and prone to fatigue. Hence the proposed methodology focuses on developing a tool for behavioral analysis which can predict the personality traits automatically with the aid of a computer without the human intervention. In this paper a method has been proposed to predict the personality of a person from the baseline, the pen pressure and the letter t. as found in an individual s handwriting. These parameters are the inputs to the Artificial Neural Network which outputs the personality trait of the writer. The performance is measured by examining multiple samples. Keywords :- Artificial Neural Network, Behavior Analysis, Behavior Prediction, Graphology, Handwriting Analysis. I. INTRODUCTION Graphology is the study of handwriting. It is a scientific method of identifying, evaluating, and understanding a person s personality via the strokes and patterns revealed by his handwriting. The handwriting is done by the brain and not the hand. Hence handwriting is also known as brain writing. Research scientists in the fields of neuro-science have categorized neuro-muscular movement tendencies as they are correlated with specific observable personality traits. Each personality trait is represented by a neurological brain pattern. Each neurological brain pattern produces a unique neuro-muscular movement that is the same for every person who has that personality trait. When writing, these tiny movements occur unconsciously. Each written movement or stroke reveals a specific personality trait. Graphology is the science of identifying these strokes as they appear in handwriting and describe the corresponding personality trait. In this paper, a method has been proposed to predict the behavior of a person from the features extracted from his handwriting. The personality traits revealed by baseline, letter slant, pen pressure, size, zone and word spacing as found in individual s handwriting are explored in this paper. Six parameters 7 are input to the ANN which outputs the personality trait of the writer. The evaluation of the baseline and letter slant is using the polygonization method, the evaluation of pen pressure utilizes grey-level threshold value, and evaluations of zone use Pythagora s theorem. Eclipse is the tool used for this purpose. The performance is measured by examining multiple samples. II. RELATED WORK Handwriting Analysis or Graphology is a scientific method of identifying, evaluating and understanding personality through the strokes and patterns revealed by handwriting. Among the many aspects of handwriting that can serve as scheme to predict personality traits are baseline, letter slant, pen pressure, size, zone and word spacing. Writer individuality rests on the hypothesis that each individual has consistent handwriting, which is distinct from the handwriting of another individual. However, this hypothesis has not been subjected to rigorous scrutiny with the accompanying experimentation, testing and peer review. 795 Page
III. METHODS AND DATA DESCRIPTION Pattern recognition performed on a sheet of A4 handwritten scanned using a scanner in jpeg format as Fig 1. The image is divided into two areas including handwriting and signature area. Fig. 1. Design of handwriting area of image of testing data Pattern recognition and handwritten signatures are used to predict personalities. The system is developed consists of two phases: pre-processing and pattern recognition of each feature stage. A. Data Aquisition and Image Pre Processing Handwriting image samples of different individuals are used in this research which is digitally collected by scanning the handwriting of 50 different writers (training data) and 100 different writers (testing data). Each of them was asked to write a text document of simple. Most of the handwritings are printed but few of them are cursive handwriting. The samples we rewritten on A4 size paper without any lines. In pre-processing stage, image processing was done with gray scale and threshold so the handwriting image in a bit then handwriting area. B. Recognition of Handwriting System Handwriting are identify certain pixels that can inform structure pattern of handwriting. This method is used when the features have to be reviewed have simple structure pattern. Therefore this method is called multistructure algorithm. When the feature has a complex pattern, recognition using second methods that is ANN with multi-layer perceptron(mlp) architecture. The imags of handwriting area is done pre-processing stage. After that, the image is segmented correspond to the features that were reviewed. For handwriting has six features, that is baseline, letter slant, pen pressure, size, zone and word spacing identified using multi-structure algorithm. While the baseline features were identified using ANN. Segmentation and identification process as Fig 2. 796 Page
Fig. 2. Recognition of handwriting system consist 4 features using multistructure algorithm and baseline feature using ANN. The process of feature extraction is done by saving the memory segmentation for classification. Segmentation performed on three stages, i.e. vertical segmentation, horizontal segementation and lines segmentation as Fig 3. Fig3. Three Step Segmentation 797 Page
Vertical segmentation divide the image into three sections that right and left side of the image used to analyze page margin. Whereas spacing between lines analysis using left side. Horizontal segmentation to divide into three parts which middle segment are processed into line segments. The process begins by taking the coordinates of x in the upper left corner and y coordinates of the lower left corner. Having found the black pixel value of x will be stored as the value CropX1. Then look for the value of the value of x Crop X2 started last form to the right y-axis until no black pixels are found. Then lookcropy1 and CropY2 the same process starting from the bottom point. It used to classify dominance zone, baseline patterns and spaces between words features. To identify baseline pattern, result of line segmentation process was extracted by Hill Valley feature as shown in Fig 4.The output becomes input for identification system using artificial neural network with Multilayer Perceptron (MLP) architecture. Classification of human personality based on page margin, space between word or line and the dominance of zone feature used multi structures. Analysis of the structure of the page is done by identifying the edge of black pixels in the upper left and upper right. The second classification is based on the spacing between lines is done by identifying the writing and spaces of color pixels after noise cancelling. Line spacing is calculated of the vertical segmentation as Fig 5. Meanwhile, classification based on space between word and dominance of vertical zones after line segmentation. For classification based on spacing between words, compute average of writing and space, as shown in Fig5. Classification based on dominance of vertical zones was proceed starting from the left edge. There will be a row of the array, where is searched for the highest series. Highest Array will be divided into three based on the number of existing zones, with the goal to find out where the time of writing is at its decline. From these comparisons it will produce the dominant type of zone as Fig 5.The last classification is based on baseline structure performed after feature extraction process in Fig. 4, using artificial neural network with as show in Fig.6 798 Page
The system consists of 2000 neurons of input layer, ten neurons of hidden layer and five neurons of output layer. Each input neuron xi is connected to hidden neurons by each weight vij (1). Then each hidden neuron zj, is connected to output neurons yk by wjk (2). Fungsi f(.) is sigmoid biner. 799 Page
The error between the output of the feed-forward network and the target (ek) correct weight by back propagation algorithm. This research used five output neurons or five Classes of the total pattern of the baseline. They are pattern of Straight lines, ascending base line, declining baseline, baseline convex and concave bottom line. IV. ACKNOWLEDGEMENTS It is our privilege to acknowledge with deep sense of gratitude to our project guide, Prof.K. P. Moholkar for their valuable suggestions and guidance throughout our course of study and timely help given to us in the completion of our project titled Human Behavior Prediction through Handwriting Analysis REFERENCES [1] D. John Antony, O. F. M. Cap. - Personality profile through handwriting analysis [2] Champa H N, K R Ananda Kumar- Artificial neural network for human behavior prediction through handwriting analysis. International Journal of Computer Applications, Vol. 2, May 2010, pp. 36-41 [3] G. Sheikholeslami, S. N. Srihari, V. Govindaraju Computer aided graphology Center of Excellence for Document Analysis and Recognition (New York) [4] Abdul Rahiman M, Diana Varghese, Manoj Kumar G HABIT- Handwriting Analysis Based Individualistic Traits Prediction [5] Ricard Coll, Alicia Forn es, JosepLlad os - Graph logical Analysis of Handwritten Text Documents for Human Resources Recruitment. 2009 10th International Conference on Document Analysis and Recognition [6] Parmeet Kaur Garewal, Deepak Prashar Behavior Prediction through Handwriting Analysis. [7] Huber RA, Headrick AM (1999). Handwriting Identification: Facts and Fundamentals. Boca Raton: CRC Press. ISBN:084931285X 800 Page