You are on page 1of 25

Chapter 8: Conclusions and Future Research Work

CHAPTER 8

Conclusion and Future Research Work

The closing chapter aims at presenting a summary of the nine chapters of the study.
The research work presented in this thesis describes several techniques for
segmenting and recognising offline handwritten Arabic text. The main goal of this
thesis is the proposal, development and investigation of a segmentation and
recognition technique that could be integrated into a complete segmentation and
recognition system for offline handwritten Arabic documents. The hybrid feature
extraction approach, Arabic text segmenter and later recognition classifiers were
designed, implemented, tested and finally all techniques were integrated to perform
the goal of this research. Figure 8.1 shows the frame work of the system implemented
in this research work.

IHACDB

Handwritten Arabic document Image Digitization Pre-processing

Global segmentation approach (text


to words)

Feature extraction

Analytical segmentation approach


Recognition and classification (words to characters or segments)

Editable document

Figure 8.1: The frame work of the research work.

126
Chapter 8: Conclusions and Future Research Work

In pre-processing stage of the work, Otsu’s method for thresholding images is used to
binarize Arabic text images. We apply several filtering techniques such as Median
filter, morphological operations, and noise modeling techniques on off-line
handwritten Arabic text to reduce noise.

We also apply skew and size normalization techniques which remove some variations
in Arabic text images without affecting the identity of the text.

The concept of baseline, which plays a very important role in pre-processing stage of
handwritten Arabic text, is also implemented.

Segmentation is still the difficult and complex task not only in Arabic OCR but for
every cursive language and the correct recognition often depends on a correct
segmentation. The segmenter segments a text page into text lines, a text line into
words as well as a cursive word into its characters. Then, the isolated character is
divided into primary and secondary parts if it has more than one component.

This Ph.D thesis presents a newly hybrid feature extraction approach consists of
Moment Invariant, Zernike Moments and Eigen Value Eigen Vector techniques for
extracting appropriate features set which help to achieve improved performance rate,
decreasing the feature set dimensions and recognition time of handwritten Arabic text.

The total number of features extracted from this hybrid feature extraction approach is
27 features. These features represent the minimal features used in the field of Arabic
OCR.

The main advantage of the hybrid feature extraction approach is that, the features
extracted are invariants under translation, scale, rotation, and reflection. It overcomes
the problems in verities of handwritten Arabic style.

Support Vector Machine (SVM), which is the best classifier till now, is adopted as the
classifier of this work.

127
Chapter 8: Conclusions and Future Research Work

This Ph.D research has built a new Isolated Handwritten Arabic Characters IHACDB
database in order to provide normalization data that will be used for training systems
to recognize unconstrained handwritten Arabic text. This database may be used in
handwriting recognition in other languages such as Farsi, Jawi, Urdu, and Kurd.
IHACDB is freely available for researchers on internet.

This research developed a novel data set consists of five classes. The first four classes
represent the different four shapes of a character in a cursive word and the fifth class
is for secondary components of a character as follows:
 Class1 is for isolated form,
 Class 2 is for beginning form,
 Class 3 is for middle form,
 Class 4 is for end form,
 Class 5 is for secondary components of Arabic characters.
This data set reduces the number of characters and their four different shapes from
110 shapes to 59 shapes only.

8.1: Summary of the Previous Chapters

Chapter one of this study is an introductory chapter. It includes the basic concepts and
principles of Optical Character Recognition (OCR). The two types of OCR and their
differences are also discussed. This chapter also gives an introduction to Arabic
language, its history, importance and the originating of Arabic language. List of
Arabic alphabets and their different shapes are also included. It discusses the
uniqueness of Arabic characters and its cursive nature that differentiate from English
language. The objectives of this study are included in this chapter too.

Chapter two discusses the OCR stages in general. It surveys the state of the art in off-
line handwritten Arabic recognition. It also defines the existing of Arabic databases
for this field. The previous work has been done in the field of handwritten Arabic
OCR is included. The concept of baseline and its significance and roles in Arabic
writing are also discussed in this chapter.

128
Chapter 8: Conclusions and Future Research Work

Chapter three discusses the techniques that have been used in the segmentation and
recognition of off-line handwritten Arabic documents. This chapter discusses the
three feature extraction methods which are used throughout this research work i.e.
Moment Invariants, Zernike Moments and Eigenvalue Eigenvector. This chapter also
describes Support Vector Machine (SVM) classifier which is used to classify
handwritten Arabic text.
Chapter four introduces the new Isolated Handwritten Arabic Characters Database
(IHACDB). This chapter describes how this database was created, forms used in
collecting Arabic characters in their four shapes, segmentation, pre-processing etc.

Chapter five describes the handwritten Arabic numerals (Hindi digits). This chapter
surveys the previous work has been done in the field of recognition of handwritten
Arabic numerals. It also discusses the existing of numeral databases used in this field.
The experiments done by the researcher in recognizing offline handwritten Arabic
numerals (Hindi digit) are introduced too, in this chapter,.

Chapter six discusses the proposed and implemented pre-processing and segmentation
techniques used in this thesis. Pre-processing techniques are employed on Arabic text
to produce standard data for segmentation/recognition stages of the proposed system.
This chapter presents the new segmentation algorithm for off-line handwritten Arabic
text.

Chapter seven describes the feature extraction and recognition stages for off-line
handwritten Arabic text. The hybrid feature extraction approach for recognition of
off-line handwritten Arabic text based on Support Vector Machine (SVM) is
discussed. The novel five classes of Arabic dataset are developed. Features used
training and testing phases of SVM and recognition algorithm are also discussed in
this chapter.

Chapter eight outlines the experiments and results that have been obtained during this
research work. The results of the work are as follows:
 Segmenting a text page into text lines is 98%.
 Segmenting a text line into words is 99%
 Segmenting a cursive word into characters is 85%

129
Chapter 8: Conclusions and Future Research Work

 Recognition of cursive words is 82.89%.


 Fuzzy Based Recognition of Offline Handwritten Arabic Numerals is 95.14%
by using 161 features.
 Recognition of Offline Handwritten Arabic Numerals Using SVM is 97.31%
by using the hybrid feature extraction approach i.e. 27 features.

Comparing this work with other existing systems, the system is the first system uses a
hybrid feature extraction approach and has the ability to segment a word to its
characters; the comparison shows that a higher accuracy rate is obtained using this
system.

8.2: Contribution to Arabic OCR

The contributions work of this thesis is as follows:


 (IHACDB) A new database created in this thesis for the collection, storage
and retrieval of Arabic handwritten characters have been generated, which
considers a welcome addition to existing Arabic databases.

 IHACDB is a new Arabic database containing, handwritten isolated Arabic


characters i.e. the five Arabic classes.

 In the pre-processing stage, traditional tools such as morphological operations


are used.

 In the feature extraction stage, a hybrid feature extraction approach for


recognition of off-line handwritten Arabic text is constructed and tested in the
classification stage.

 In the feature extraction stage, moment invariants features and the divisions of
numeral images into several parts for recognition of off-line handwritten
Arabic numerals is constructed and tested in the classification stage.

 In the segmentation stage, the segmentation algorithm is simple and efficient


and has given a high correct segmentation rate.

130
Chapter 8: Conclusions and Future Research Work

 A SVM technique is used to train and test Arabic handwritten text.

 A fuzzy conceptual approach is used to classify handwritten Arabic numerals.

8.3: Future Research Work:

The following are future research work:

1- Due to the cursive nature of Arabic language, serious researches should be


done in segmenting handwritten Arabic text to get the highest segmentation
rate.

2- Segmenting Arabic ligatures is still a challenging task in machine printed as


well as handwritten Arabic text. Therefore, it is recommended to do further
researches in segmenting Arabic ligatures.

3- A study of other pre-processing methods to provide clean and easy Arabic text
images for processing.

4- Further researches should be done to enhance the IHACDB database of this


study.

5- A study of other feature extraction methods that depend on topology of the


characters such as vectors, strokes, loops, dots ...etc.

6- Due to the similarity among Arabic words, post-processing methods should be


studied exhaustively. Post-processing will definitely improve the recognition
rate.

8.4: Conclusion:

This chapter gives a summary of the previous chapters, and work that has been
done during this research work. It presents the contribution of this work to Arabic
OCR. Future research work are offered in this chapter.

131
Appendix A: Arabic Forms

APPENDIX A

Appendix A consists of the six forms that have been used to collect Arabic characters
and diacritical marks as well. Every form is printed on one separate page.

‫ا‬
‫أ‬
‫إ‬
‫آ‬
‫ـا‬
‫ب‬
‫ت‬
‫ث‬
‫ج‬
‫ح‬
‫خ‬
‫د‬
‫ذ‬
‫ر‬
‫ز‬
‫س‬
‫ش‬
‫ص‬
‫ض‬
‫ط‬
132
‫‪Appendix A: Arabic Forms‬‬

‫ظ‬
‫ع‬
‫‪Form 1‬‬

‫غ‬
‫ف‬
‫ق‬
‫ك‬
‫ل‬
‫م‬
‫ن‬
‫هـ‬
‫و‬
‫ؤ‬
‫ي‬
‫ئ‬
‫ى‬
‫ال‬
‫أل‬
‫إل‬
‫بـ‬
‫ـبـ‬
‫ـب‬
‫تـ‬
‫‪133‬‬
‫‪Appendix A: Arabic Forms‬‬

‫ـت‬
‫ـتـ‬
‫‪Form 2‬‬

‫ثـ‬
‫ـثـ‬
‫ـث‬
‫جـ‬
‫ـجـ‬
‫ـج‬
‫حـ‬
‫ـحـ‬
‫ـح‬
‫خـ‬
‫ـخـ‬
‫ـخ‬
‫ـد‬
‫ـذ‬
‫ـر‬
‫ـز‬
‫سـ‬
‫ـسـ‬
‫ـس‬
‫شـ‬
‫‪134‬‬
‫‪Appendix A: Arabic Forms‬‬

‫ـشـ‬
‫ـش‬
‫‪Form 3‬‬

‫صـ‬
‫ـصـ‬
‫ـص‬
‫ضـ‬
‫ـضـ‬
‫طـ‬
‫ـطـ‬
‫ظ‬
‫ـظ‬
‫عـ‬
‫ـعـ‬
‫ـع‬
‫غـ‬
‫ـغـ‬
‫ـغ‬
‫فـ‬
‫ـفـ‬
‫ـف‬
‫قـ‬
‫ـقـ‬
‫‪135‬‬
‫‪Appendix A: Arabic Forms‬‬

‫ـق‬
‫كـ‬
‫‪Form 4‬‬

‫ـكـ‬
‫ـك‬
‫لـ‬
‫ـلـ‬
‫ـل‬
‫مـ‬
‫ـمـ‬
‫ـم‬
‫نـ‬
‫ـنـ‬
‫ـن‬
‫هـ‬
‫ـهـ‬
‫ـه‬
‫ـة‬
‫ـو‬
‫يـ‬
‫ـيـ‬
‫ـي‬
‫‪-‬ؤ‬
‫‪136‬‬
Appendix A: Arabic Forms

‫ـئ‬
‫ـى‬
Form5

~
~
~
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ
ْ

Form 6: Diacritical Marks

137
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

APPENDIX B

Appendix B consists of the Mean and Standard Deviation that represent the features
of handwritten Arabic numerals extracted using Moment Invariants method. Table
B.1 shows the trained data i.e. Mean and Standard deviation of Moment Invariants
features of Feature Group1as shown in chapter 4.

Table B.1 (A)


Mean and Standard Deviation of UPPER part of Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.9206 15.0628 23.0403 25.1375 49.7266 33.2938 50.5401
0
0.3686 1.5295 1.4925 2.2107 4.2755 3.1396 3.6706
6.4066 13.1327 23.7098 24.8195 49.4261 31.8808 50.7098
1
0.2699 0.6986 1.7949 1.9868 4.0205 2.4309 3.8771
6.1196 13.1028 19.6941 21.2585 42.6708 28.8277 42.3158
2
0.2855 1.0113 1.0169 1.1009 2.3019 1.6444 2.3651
6.1791 13.2538 20.3195 21.8553 43.7165 29.3793 43.6027
3
0.3296 1.1524 1.1577 1.2971 2.7502 1.9748 2.5236
6.2692 14.0764 20.3208 22.3528 44.4691 30.1704 44.4036
4
0.2573 1.2832 1.3637 1.5111 2.9963 2.0332 3.0008
6.0734 13.0258 19.2752 21.0043 41.9842 28.4559 41.7879
5
0.3665 1.1117 1.164 1.3036 2.7098 1.9208 2.6133
5.9434 12.2833 19.486 20.9888 41.9112 28.0309 41.8859
6
0.2959 0.7581 1.1083 1.2825 2.7218 1.9144 2.5108
5.4316 11.5663 18.1963 19.7086 39.2469 26.3026 39.6976
7
0.3737 1.035 1.2863 1.6362 3.403 2.3788 2.7522
6.5526 15.713 20.4044 22.9859 45.5179 31.5087 45.354
8
0.2667 1.5415 0.9624 1.6081 2.9325 2.1374 2.9875
6.291 14.5636 20.8082 22.8422 45.4021 30.9601 45.3711
9
0.3128 1.4589 1.4731 1.8826 3.5411 2.4954 3.6102

Table B.1(B)
Mean and Standard Deviation of LOWER part of Handwritten Arabic Numerals
image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.8141 14.6718 22.6501 24.4779 48.4683 32.3904 49.5396
0
0.4194 1.5593 1.7663 2.4866 4.8267 3.4447 4.0571
6.2996 12.8743 22.7978 23.6944 47.1689 30.4965 48.9642
1
0.2905 0.703 1.9894 2.2755 4.5645 2.7634 4.1217

138
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

5.8942 11.9446 21.3219 22.3415 44.5488 28.8976 45.8475


2
0.2908 0.6561 1.7779 1.9952 4.0401 2.4514 3.7991
5.986 12.4712 20.7353 22.1109 43.9989 28.9565 44.9651
3
0.2664 0.878 1.9108 2.0077 4.0493 2.3297 3.9367
5.9581 12.6184 19.1702 20.4152 40.7981 27.4817 40.9153
4
0.3248 1.1617 0.9779 1.1625 2.518 1.9738 2.1847
5.8937 12.3839 19.0413 20.7582 41.5253 27.8552 41.2836
5
0.3043 0.8493 0.9473 1.0311 2.1876 1.5695 2.193
6.1308 12.4824 22.1923 23.279 46.3895 30.1358 47.5951
6
0.261 0.5938 1.8669 2.0917 4.2048 2.5158 3.9464
6.5085 15.1744 20.2005 22.4863 44.6289 30.7514 44.5342
7
0.2844 1.5091 0.9454 1.5364 2.8699 2.1735 2.8496
5.4195 11.5634 18.1358 19.4776 38.8538 26.0864 39.3122
8
0.3967 1.1254 1.3907 1.7157 3.5704 2.5283 2.8303
5.7185 11.9574 18.3533 20.0185 39.8421 26.9274 40.1018
9
0.3197 0.7706 1.6792 1.7804 3.6428 2.214 3.5293

Table B.1(C)
Mean and Standard Deviation of LEFT part of Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.9881 15.5021 23.3192 25.6573 50.6412 34.0191 51.4184
0
0.3532 1.5807 1.5027 2.2383 4.2353 3.1155 3.8035
5.7669 11.6365 19.9537 20.2666 40.4232 26.1819 44.2481
1
0.3748 0.8238 2.0435 2.2576 4.5176 2.5615 3.7023
6.107 13.0906 19.3149 21.0947 42.2612 28.6606 41.8276
2
0.2565 0.913 0.7867 0.848 1.9381 1.5257 1.8333
6.161 12.8606 19.9497 21.4085 42.3006 28.1797 43.9844
3
0.2993 0.9058 1.3027 1.8256 3.5038 2.4806 2.9935
5.8202 12.0803 20.1647 21.5209 42.683 27.9741 44.079
4
0.32 0.8845 1.6672 1.982 3.9939 2.5129 3.3514
6.1457 13.6981 19.6242 21.2758 42.5246 28.9504 42.4151
5
0.3076 1.2003 1.1043 1.1621 2.4115 1.8182 2.405
6.0439 12.5179 21.1583 22.2654 44.4758 29.2258 45.2855
6
0.3678 0.8491 2.4838 2.6225 5.2686 2.9224 5.0084
5.897 11.9785 21.2237 22.3327 44.5142 28.9067 45.64
7
0.2714 0.6749 1.6377 1.7416 3.5558 2.1421 3.4682
5.9599 12.1161 21.135 22.1034 44.0182 28.6118 45.546
8
0.3327 0.7894 1.6037 1.7997 3.648 2.2518 3.4695
6.2792 14.857 21.2082 22.6894 45.4073 30.8728 45.3972
9
0.2968 1.4919 1.7607 1.9766 3.8601 2.6289 3.8037

139
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

Table B.1(D)
Mean and Standard Deviation of RIGHT part of Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.9865 15.4798 23.3228 25.6141 50.5964 33.9781 51.2555
0
0.3025 1.5659 1.5388 2.2158 4.2125 3.1055 3.6997
5.5839 11.2437 19.5712 19.9233 39.7494 25.7059 43.4116
1
0.4134 0.8718 2.3333 2.5682 5.1335 3.0138 4.2652
5.0676 10.5886 17.2582 18.1209 36.0483 23.8378 37.8421
2
0.3556 0.8772 1.7111 2.0098 4.042 2.5645 3.2805
5.5846 11.7843 17.8683 18.8145 37.451 25.1707 39.2272
3
0.4751 1.4057 1.8403 2.2362 4.5234 3.0816 3.5412
5.3041 11.2062 16.9826 18.1764 36.1466 24.4035 36.9934
4
0.4267 1.0569 1.8027 2.0873 4.2595 2.7567 3.7652
6.1514 13.7431 19.5873 21.248 42.4971 28.9915 42.3501
5
0.321 1.262 1.1108 1.2123 2.5561 1.8787 2.4916
5.7837 11.7053 20.7349 21.8704 43.5629 28.3397 44.6659
6
0.2465 0.5389 1.5704 1.7113 3.4969 2.147 3.3452
5.8913 11.9879 20.7968 21.8162 43.4535 28.279 44.8286
7
0.312 0.7736 1.4803 1.6171 3.3089 2.0408 3.1903
5.826 11.7843 21.1428 22.1269 44.0893 28.5467 45.6364
8
0.2951 0.669 1.7059 1.8934 3.8249 2.3633 3.5605
5.8022 11.7767 20.2003 21.0211 41.8322 27.2627 43.8151
9
0.2476 0.5794 1.3356 1.5702 3.1798 1.9974 2.7589

Table B.1(E)
Mean and Standard Deviation of LOWER LEFT part of handwritten Arabic Numerals
image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0109 15.7535 23.6868 26.1064 51.5605 34.5961 52.0877
0
0.3906 1.6673 2.2322 3.0042 5.7224 3.8986 5.3213
5.9625 12.1688 19.7941 20.1419 40.1707 26.3048 42.8309
1
0.5017 1.178 2.8576 2.9918 5.9738 3.3972 5.1967
6.4901 13.3764 22.9693 24.4216 48.6663 31.863 49.1578
2
0.2507 0.6628 1.5453 1.6907 3.4452 2.1326 3.3193
6.2627 12.95 20.7982 22.0567 43.8428 29.0414 44.886
3
0.3426 0.939 1.4655 1.8331 3.6996 2.4543 3.1231
6.5319 14.2161 21.0916 23.0499 45.8592 30.9166 45.9724
4
0.3598 1.2685 1.3883 1.7367 3.4567 2.4075 3.2746
6.375 13.3638 21.0807 22.7856 45.6112 30.433 45.2909
5
0.2643 0.8016 1.3833 1.4009 2.8627 1.8428 2.8898
3.9804 8.2148 13.1717 13.3284 26.6909 17.6195 37.7638
6
1.5699 3.3429 4.6273 4.7463 9.5075 6.3175 20.3696
6.4403 13.3432 22.2004 23.662 47.1874 31.1014 47.5698
7
0.2631 0.7593 1.6414 1.7898 3.6509 2.1937 3.5007
8 6.4565 13.3559 23.2415 24.5454 48.8799 31.8165 49.7777

140
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

0.3224 0.9103 1.6643 1.9236 3.9002 2.4458 3.4665


6.114 12.7446 20.4276 21.8969 43.6167 28.9407 44.5295
9
0.6428 1.4765 2.3794 2.8934 5.7206 3.7589 5.1375

Table B.1(F)
Mean and Standard Deviation of LOWE RIGHT part of handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.9832 15.5335 23.2901 25.576 50.5512 33.9509 51.0287
0
0.3383 1.5293 2.0409 2.6512 5.0514 3.4521 4.7869
6.0964 12.3685 22.4078 23.1129 46.0795 29.6203 48.0055
1
0.3663 0.8478 2.1634 2.3856 4.8072 2.8521 3.9762
6.1845 12.6983 21.751 22.7643 45.39 29.6555 46.7658
2
0.4191 0.9735 2.1332 2.4627 4.9486 3.0898 4.2934
5.6157 11.6054 18.735 19.4173 38.7465 25.6678 40.8
3
0.8196 1.7464 3.4597 3.8341 7.6365 4.8125 6.7262
5.875 12.4843 19.205 20.5428 41.0037 27.5954 41.5455
4
0.3357 0.9532 1.9238 1.9401 3.9863 2.3085 3.9645
6.3538 13.2423 21.3294 22.9776 46.0075 30.5847 45.7835
5
0.2863 0.8087 1.5445 1.5886 3.2246 2.0133 3.2333
6.088 12.3667 22.0827 23.1863 46.227 30.0141 47.3248
6
0.258 0.5797 1.7881 1.9476 3.9643 2.3982 3.7172
6.532 13.858 22.2582 23.7883 47.323 31.2999 47.9007
7
0.2806 1.2462 1.4017 1.5743 3.1889 2.1204 3.0643
6.2864 12.9012 22.632 23.7385 47.2952 30.7491 48.6626
8
0.3459 0.8933 1.8236 2.1351 4.2912 2.7377 3.7344
5.9946 12.2397 20.9861 22.1413 44.0728 28.781 45.2699
9
0.2457 0.6107 1.6125 1.7561 3.5841 2.1932 3.3943

Table B.1(G)
Mean and Standard Deviation of UPPER LEFT part of Handwritten Arabic Numerals
image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0112 15.6469 23.2792 25.7325 50.7847 34.1635 51.4131
0
0.39 1.5247 1.9095 2.61 4.9501 3.4244 4.542
6.3103 12.8673 23.2588 24.1514 48.08 30.9237 49.7594
1
0.3038 0.7358 2.1004 2.329 4.6789 2.762 3.8948
6.4675 13.7922 21.1208 22.9765 45.8609 30.7164 45.7132
2
0.2391 0.8282 1.264 1.2993 2.6646 1.691 2.7527
6.6412 14.6766 21.4475 23.4755 46.7146 31.6197 46.6621
3
0.2493 1.1817 1.1915 1.5229 2.8969 2.0736 2.8802
6.3395 14.158 20.2426 22.1123 43.9471 29.9123 44.2155
4
0.2846 1.4019 1.2031 1.5048 3.0201 2.1947 2.7293
5 6.4267 13.4625 21.7603 23.3191 46.528 30.8355 46.7372

141
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

0.3194 0.9818 1.6072 1.739 3.4966 2.2681 3.38


6.1211 12.5695 21.8374 23.0806 46.0771 30.0426 46.7802
6
0.2973 0.7373 1.896 1.9127 3.8617 2.2645 3.8417
6.4315 13.3575 22.8946 24.2088 48.2172 31.5069 49.0813
7
0.3147 0.9579 1.754 1.9463 3.9272 2.4201 3.6597
6.545 13.7845 22.4545 23.9298 47.561 31.3693 48.2629
8
0.2838 1.0611 1.4515 1.7009 3.395 2.2561 3.1973
6.2869 14.5067 21.3893 22.5137 45.2293 30.537 45.1923
9
0.2848 1.544 1.5223 1.7506 3.2716 2.3988 3.3086

Table B.1(H)
Mean and Standard Deviation of UPPER RIGHT part of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0514 15.8733 23.7455 26.3023 51.8821 34.8535 52.399
0
0.3289 1.5332 2.0207 2.76 5.2466 3.6069 4.9568
5.8854 11.9873 20.0364 20.3686 40.6481 26.4463 43.0808
1
0.4913 1.1603 2.6755 2.7994 5.614 3.1956 5.1458
6.5325 13.9162 22.2546 23.5962 47.1363 31.32 47.5176
2
0.3754 1.2553 1.9125 2.1852 4.3529 2.8124 4.0394
6.4581 14.1996 21.4168 23.1026 46.0971 31.0454 46.0929
3
0.2906 1.2957 1.4787 1.5866 3.2042 2.1783 3.0813
5.7196 12.0559 18.589 19.3286 38.5219 25.6494 40.9875
4
0.8107 2.1472 3.0562 3.7151 7.3195 4.9125 5.8701
6.4299 13.5275 21.6914 23.1939 46.2577 30.7112 46.5877
5
0.3497 1.06 1.7107 1.9216 3.8669 2.4811 3.6204
6.681 14.2456 22.8153 24.6867 49.0606 32.5117 49.3168
6
0.2366 0.9533 1.6544 1.8487 3.6912 2.3763 3.5077
6.4094 13.2324 23.2774 24.5633 48.9257 31.7795 49.8416
7
0.3083 0.873 1.6985 1.9084 3.8561 2.3895 3.5381
6.4439 13.3247 22.236 23.6282 47.0474 30.9548 47.6654
8
0.2613 0.7656 1.4096 1.6184 3.3185 2.1175 3.0393
6.5995 14.149 21.873 23.4228 46.5423 31.0642 47.2278
9
0.3048 1.127 1.5497 1.9672 3.9014 2.6237 3.414

Table B.2 shows Mean and Standard Deviation of Moment Invariants features of
Feature Group 2.
Table B.2(A)
Mean and Standard Deviation of SWAPPING UPPER part of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.364 14.0478 25.1198 27.217 53.8838 34.9987 54.7022
0
0.5601 1.4794 1.4925 2.2107 4.2745 2.691 3.7074

142
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

5.2163 10.5005 25.7892 26.899 53.5839 32.6888 54.8633


1
0.2644 0.5559 1.7949 1.9868 4.0199 2.4041 3.8649
5.5838 12.3638 21.7736 23.3379 46.824 30.2494 46.4665
2
0.2509 0.9212 1.0169 1.1009 2.3009 1.5886 2.3672
5.6706 12.9382 22.3989 23.9347 47.8712 31.2806 47.7588
3
0.2316 1.0646 1.1577 1.2971 2.7491 1.635 2.5234
5.5193 11.4388 22.4002 24.4322 48.6232 30.9984 48.5579
4
0.2737 0.6638 1.3637 1.5111 2.9966 1.8452 3.001
5.8945 14.3544 21.3547 23.0837 46.1369 31.086 45.9415
5
0.2997 0.8067 1.164 1.3036 2.7096 1.5874 2.6143
5.4746 12.5731 21.5654 23.0682 46.0676 29.9556 46.0436
6
0.2526 0.8376 1.1083 1.2825 2.7204 1.7119 2.5097
5.3806 14.1545 20.2757 21.788 43.4023 29.6302 43.8496
7
0.2871 1.5044 1.2863 1.6362 3.4009 2.0355 2.7537
5.7793 12.1844 22.4839 25.0654 49.6713 31.9569 49.5068
8
0.2408 0.5917 0.9624 1.6081 2.9332 1.8409 2.9886
5.6786 12.3875 22.8876 24.9216 49.5572 32.0521 49.5269
9
0.2271 0.6728 1.4731 1.8826 3.5412 2.0452 3.6102

Table B.2(B)
Mean and Standard Deviation OF SWAPPING LOWER part of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.2157 13.7757 24.7296 26.5574 52.6257 34.1314 53.7306
0
0.5482 1.4562 1.7663 2.4866 4.8257 3.0016 4.2103
5.0747 10.2136 24.8772 25.7738 51.327 31.2717 53.0562
1
0.2735 0.5697 1.9894 2.2755 4.5638 2.6623 3.8258
5.385 11.5902 23.4013 24.4209 48.7061 31.1547 50.0039
2
0.26 0.8054 1.7779 1.9952 4.0392 2.1342 3.7984
5.2019 10.6727 22.8148 24.1903 48.1555 30.2157 49.1189
3
0.2541 0.6294 1.9108 2.0077 4.0487 2.281 3.938
5.3813 11.9166 21.2496 22.4946 44.955 28.8669 45.0737
4
0.2748 0.9041 0.9779 1.1625 2.5166 1.373 2.1841
5.7458 14.5034 21.1207 22.8376 45.6768 30.779 45.4381
5
0.2635 0.8169 0.9473 1.0311 2.1875 1.3238 2.1931
5.1933 10.5189 24.2718 25.3584 50.5469 31.4633 51.7513
6
0.2599 0.5158 1.8669 2.0917 4.204 2.3608 3.9459
5.6769 12.0191 22.28 24.5658 48.7825 31.4046 48.6865
7
0.2485 0.5944 0.9454 1.5364 2.8705 1.8292 2.8505
5.2933 13.702 20.2152 21.5571 43.0093 29.1967 43.4635
8
0.2878 1.5986 1.3907 1.7157 3.5684 2.0146 2.8314
5.2708 11.4611 20.4328 22.0979 43.9963 28.5006 44.2577
9
0.2646 0.9699 1.6792 1.7804 3.6418 2.0266 3.5282

143
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

Table B.2(C)
Mean and Standard Deviation of SWAPPING LEFT part of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.3689 13.8145 25.3986 27.7367 54.7982 35.387 55.563
0
0.5985 1.5438 1.5027 2.2383 4.2345 2.6919 3.7752
5.2151 12.279 22.0332 22.346 44.582 28.7843 48.3993
1
0.2569 0.9351 2.0435 2.2576 4.5173 2.4694 3.7045
5.6131 12.6802 21.3944 23.1741 46.4129 30.1951 45.9854
2
0.2505 0.6675 0.7867 0.848 1.9385 1.2792 1.8329
5.5866 12.4279 22.0292 23.4879 46.4588 30.229 48.1383
3
0.237 0.8239 1.3027 1.8256 3.5029 1.9249 2.994
5.5061 13.2531 22.2441 23.6003 46.8406 30.9662 48.2341
4
0.2778 0.9563 1.6672 1.982 3.9929 2.2725 3.3512
5.9132 13.8404 21.7036 23.3553 46.6758 30.8977 46.5686
5
0.2509 0.916 1.1043 1.1621 2.4125 1.5694 2.4065
5.2037 10.5902 23.2377 24.3448 48.6322 30.278 49.4389
6
0.2741 0.612 2.4838 2.6225 5.2683 2.8085 5.0108
5.5242 12.1737 23.3032 24.4121 48.6712 31.4129 49.7929
7
0.2441 0.8054 1.6377 1.7416 3.5552 1.964 3.4695
5.6247 12.9255 23.2144 24.1828 48.176 31.498 49.702
8
0.2782 0.8508 1.6037 1.7997 3.6473 2.0028 3.4688
5.5499 11.8606 23.2877 24.7689 49.5603 31.3956 49.5514
9
0.2423 0.5812 1.7607 1.9766 3.8612 2.2304 3.8046

Table B.2(D)
Mean and Standard Deviation of SWAPPING RIGHT part of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.2635 13.5763 25.2271 27.4631 54.3209 34.9526 55.0939
0
0.5888 1.4984 1.6304 2.3731 4.5333 2.8278 3.841
4.5548 10.5426 20.0555 20.282 40.4851 25.8036 45.4687
1
0.6743 1.4151 3.1999 3.4325 6.7988 3.768 6.9451
4.9836 13.9294 18.9622 19.6617 39.144 27.4753 41.4356
2
0.327 1.7462 1.8654 2.1577 4.3303 2.6708 3.4464
5.1709 12.4991 19.8827 20.6927 41.2485 27.6163 43.3736
3
0.3286 1.3091 2.1863 2.668 5.3556 2.7428 4.0954
4.9545 11.9931 18.8028 19.8044 39.3942 26.4698 40.6903
4
0.342 1.5543 2.1301 2.4526 4.9474 2.5396 4.2389
5.8601 13.5806 21.6792 23.3755 46.7216 30.8797 46.5768
5
0.2609 0.9018 1.0982 1.2034 2.5255 1.5931 2.4716
5.4007 12.718 22.7974 23.9624 47.7736 31.0371 48.7942
6
0.2579 1.0463 1.6316 1.7897 3.66 2.1433 3.4168
5.5561 12.9228 23.1399 24.1503 48.1405 31.4072 49.5299
7
0.2684 1.0044 1.52 1.6642 3.4111 1.9427 3.2553

144
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

5.4389 12.1832 23.4074 24.4308 48.7148 31.3396 50.1406


8
0.2564 0.8205 1.6979 1.8594 3.7837 2.1404 3.5775
5.4364 12.677 22.4108 23.2989 46.4058 30.575 48.125
9
0.2503 0.8301 1.4274 1.6255 3.2972 1.9511 2.9845

Table B.2(E)
Mean and Standard Deviation of SWAPPING UPPER LEFT DIAGONAL of
Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0924 17.1001 24.4709 27.9466 54.2118 36.5718 82.9177
0
0.2953 2.5564 2.5739 3.2531 5.8168 4.0661 4.8212
6.0136 12.7863 18.835 21.0996 41.1873 27.672 71.3196
1
0.2917 0.8782 1.0236 1.066 2.1013 1.3398 2.0554
6.1698 15.8811 22.9437 25.5112 49.7945 33.5759 78.7404
2
0.2716 2.2833 2.3139 2.7055 4.4055 3.15 3.4519
6.415 15.5458 21.2004 25.106 48.297 33.0105 77.351
3
0.2418 2.0841 1.9532 2.5428 4.1723 2.694 3.1888
5.9674 13.5188 20.3026 22.434 43.9089 29.3549 74.146
4
0.3173 1.9573 2.1145 1.9728 3.7041 2.3816 3.1268
6.2591 16.9004 23.1474 25.2936 49.5627 33.8263 79.1845
5
0.3011 2.3036 2.384 2.5215 4.1771 2.9013 3.3519
5.6159 11.9255 17.6482 19.7196 38.5322 25.8743 69.9537
6
0.2887 0.7355 1.1423 1.1308 2.2535 1.3532 2.0335
5.6619 13.9953 18.6182 22.061 42.4439 29.1916 72.9347
7
0.32 2.4381 1.9549 2.7432 4.5075 3.2148 3.5967
6.4057 16.6094 20.6292 25.2147 48.1867 33.5853 77.6355
8
0.2702 2.3268 1.3955 2.404 3.7725 3.0393 3.0489
6.2451 14.438 22.878 25.0738 49.136 32.4434 78.2886
9
0.2776 1.5468 2.3976 2.4011 4.371 2.6514 3.3887

Table B.2(F)
Mean and Standard Deviation of SWAPPING LOWER RIGHT DIAGONAL of
Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0546 16.7931 24.3589 27.5586 53.5632 36.0224 82.1638
0
0.3262 2.5687 2.6418 3.3801 6.0492 4.2131 5.0466
6.0246 12.887 18.7443 20.9867 40.9731 27.6095 70.8392
1
0.3126 1.0449 1.0551 1.1316 2.2089 1.4669 2.149
5.4967 12.3792 18.6805 20.9038 40.7558 27.2547 71.2981
2
0.2777 1.9172 2.3336 2.2956 4.2338 2.3693 3.314
5.732 13.6922 18.2256 22.1531 42.4204 29.0711 72.3812
3
0.3181 2.1673 1.8351 2.3265 3.7587 2.7916 3.1444
5.7273 12.3619 18.9149 20.9666 41.021 27.3293 71.5764
4
0.3848 1.4334 2.3938 2.4117 4.6259 2.8538 3.6924

145
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

6.2401 16.8191 22.7521 25.6201 49.8412 34.0968 79.0653


5
0.2909 2.3136 2.2324 2.5427 4.1872 2.9274 3.3731
5.6427 11.8522 18.1389 20.4245 39.8308 26.5371 70.6177
6
0.2675 0.9762 1.3423 1.3575 2.6947 1.4401 2.2043
6.4219 15.9695 20.7724 25.373 48.4872 33.4167 77.5567
7
0.2679 2.2944 1.3492 2.7124 4.2931 3.4205 3.3229
5.6514 13.3886 19.0759 21.1679 41.338 28.0082 71.8567
8
0.3153 2.3566 2.1312 2.3637 4.1114 2.7102 3.2864
5.7028 11.9341 18.6029 20.8636 40.7179 27.0143 71.3416
9
0.3366 1.092 1.766 1.7486 3.4555 1.9324 2.8812

Table B.2(G)
Mean and Standard Deviation of SWAPPING UPPER RIGHT DIAGONAL of
Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0058 15.5225 24.4264 26.6703 52.7493 35.0842 53.2675
0
0.2796 1.5643 1.7822 2.4527 4.6407 3.2162 4.3459
5.7443 11.6058 21.2149 22.0134 43.8642 28.2229 46.0892
1
0.273 0.627 1.8774 2.0644 4.1453 2.4964 3.6452
5.8027 13.0866 18.3822 19.6926 39.6106 27.1156 39.1675
2
0.2534 1.1228 0.8911 0.9348 2.1741 1.765 1.8291
6.0401 14.0664 18.3672 20.876 41.1654 28.51 41.2394
3
0.2404 1.3407 0.8273 1.4475 2.7338 2.2471 2.5597
5.7682 12.3525 19.0305 20.6665 41.2526 27.783 41.2042
4
0.274 0.9253 1.288 1.424 2.9011 1.8762 2.8544
6.2245 15.2445 22.7581 24.2856 48.5813 32.6207 48.5304
5
0.2776 1.3567 1.5613 1.6415 3.0689 2.1839 3.0127
5.6047 11.8078 17.5678 19.5255 38.9359 26.38 38.7438
6
0.2447 0.6454 0.8626 0.8965 1.9614 1.4171 1.9672
5.9659 14.3542 18.5403 20.5878 41.0061 28.3215 40.7887
7
0.2697 1.3972 0.9696 1.1969 2.3794 1.7856 2.3866
6.0573 14.496 18.8597 20.6817 41.2883 28.3971 41.0788
8
0.2556 1.3035 1.0113 1.1992 2.4154 1.726 2.3865
6.001 13.0043 18.9185 20.3437 40.2619 27.2076 41.2052
9
0.2329 0.8562 0.8512 1.2438 2.5043 1.87 2.1229

Table B.2(H)
Mean and Standard Deviation of SWAPPING LOWER LEFT DIAGONAL of
Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0066 15.5199 24.4376 26.689 52.7891 35.0991 53.2895
0
0.2734 1.5534 1.7687 2.4336 4.6015 3.1893 4.3371
5.7453 11.607 21.2174 22.0072 43.8551 28.2208 46.0934
1
0.274 0.6377 1.8643 2.066 4.1536 2.4985 3.7067

146
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

5.8084 13.0575 18.4306 19.7434 39.7048 27.1486 39.2712


2
0.2537 1.1028 0.9089 0.955 2.1825 1.7648 1.9017
6.0386 14.0381 18.3814 20.8418 41.1086 28.452 41.2045
3
0.2401 1.3384 0.8459 1.409 2.6582 2.2051 2.505
5.7753 12.3643 19.1013 20.7431 41.4057 27.8753 41.3352
4
0.2735 0.9232 1.3162 1.4529 2.9282 1.8808 2.9376
6.2248 15.2448 22.7666 24.284 48.5805 32.6166 48.5384
5
0.2777 1.3473 1.5437 1.6432 3.0715 2.1839 2.9984
5.5905 11.7726 17.5431 19.4852 38.8549 26.3209 38.6826
6
0.246 0.6536 0.8131 0.8423 1.8751 1.3899 1.8539
5.9726 14.3372 18.554 20.6096 41.0334 28.3168 40.8483
7
0.2663 1.3974 0.9722 1.2048 2.3928 1.7994 2.4114
6.0499 14.5229 18.858 20.6592 41.263 28.4075 41.0354
8
0.255 1.3081 1.0225 1.1667 2.377 1.6994 2.3331
5.9952 12.9857 18.8985 20.3057 40.185 27.1509 41.1554
9
0.2344 0.8526 0.8505 1.2318 2.4807 1.8575 2.1076

Table B.3 shows Mean and Standard Deviation of Moment Invariants features of
Feature Group 3.

Table B.3(A)
Mean and Standard Deviation of UPPER TRIANGLE of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0256 15.7029 23.2949 25.9338 51.0965 34.4227 51.6552
0
0.3214 1.4781 1.8302 2.6204 4.9521 3.4555 4.7553
6.4265 13.1775 23.3656 24.6178 49.0373 31.8149 49.9265
1
0.2713 0.7009 1.6553 1.9135 3.8994 2.4014 3.2864
6.279 13.0489 21.3717 22.8436 45.7117 30.2458 45.7183
2
0.2746 0.8453 1.4908 1.5439 3.142 1.9358 3.1532
6.0743 13.2498 19.1228 20.531 41.0264 27.9836 41.2952
3
0.4385 1.29 1.9107 2.1464 4.3289 2.8514 3.9187
6.2064 13.8641 19.9522 21.96 43.6441 29.6974 43.7051
4
0.3234 1.3019 1.3845 1.6237 3.2513 2.2444 3.0175
6.49 13.758 21.8938 23.483 46.8034 31.1138 47.1511
5
0.3687 1.1567 1.8221 2.052 4.106 2.6908 3.7891
6.4089 13.529 20.8568 22.5272 44.9115 30.0952 45.0067
6
0.2694 0.8464 1.2916 1.3914 2.8039 1.8989 2.9403
5.5564 11.8259 17.7743 18.1709 36.3427 24.4779 38.6746
7
1.1804 2.716 4.4276 4.6891 9.4105 6.1164 8.0816
6.7463 15.4007 21.8327 24.0957 47.7485 32.4349 47.8766
8
0.2374 1.4342 1.162 1.6271 3.0515 2.3004 3.0453
6.3478 13.9877 20.6934 22.2069 44.3757 30.0258 44.4149
9
0.3274 1.303 1.2631 1.5178 3.0579 2.2211 2.8235

147
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

Table B.3(B)
Mean and Standard Deviation of LOWER TRIANGLE of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.8849 15.2403 22.5431 24.8146 49.0094 33.012 49.6797
0
0.4596 1.658 2.1374 2.9206 5.5707 3.8421 5.3055
6.3259 12.9361 22.8836 23.8979 47.6253 30.8911 49.1667
1
0.3094 0.7476 1.842 2.204 4.4223 2.8238 3.7241
6.1534 12.5206 21.9294 22.8336 45.4851 29.5076 47.2167
2
0.2751 0.6498 1.659 1.9782 3.9892 2.4801 3.3868
6.1486 12.5671 22.0031 23.0297 45.901 29.8718 47.3803
3
0.3034 0.7129 1.9024 2.1466 4.3117 2.6465 3.9136
6.2211 13.171 20.4571 21.9686 43.9872 29.4318 43.8814
4
0.275 0.9295 1.3761 1.4317 2.8446 1.8356 3.1073
6.3675 13.1015 22.6099 24.1894 48.2683 31.5288 48.5034
5
0.2855 0.7315 1.8266 1.8641 3.7617 2.2663 3.7295
6.101 12.4472 21.654 22.6169 45.0438 29.3201 46.6539
6
0.5887 1.2746 2.4334 2.7434 5.4781 3.4959 4.4443
6.735 14.9798 22.1617 24.2037 48.0519 32.3152 48.2639
7
0.2592 1.438 1.3171 1.6522 3.1841 2.2508 3.1352
5.1587 10.9401 16.2905 16.6536 33.284 22.4529 36.0685
8
1.3341 3.0259 4.6678 4.9199 9.8546 6.5092 8.6109
5.6087 11.4841 18.7941 19.4558 38.8185 25.6135 41.1358
9
0.828 1.7572 3.4154 3.772 7.5208 4.7771 6.189

Table B.3(C)
Mean and Standard Deviation of RIGHT TRIANGLE of Handwritten Arabic
Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.0355 15.832 23.5261 26.2233 51.6223 34.7539 52.2651
0
0.3161 1.583 1.8263 2.6732 5.0008 3.5338 4.7587
6.661 14.8733 20.2976 22.0153 43.6102 29.9348 45.7413
1
0.2845 1.1129 1.087 1.7693 3.2671 2.4173 7.7932
5.3522 11.5661 17.0307 17.4577 34.9717 23.6278 37.4855
2
1.5775 3.7766 5.719 6.1414 12.2854 8.1626 10.5816
6.1569 13.1048 19.8655 21.3685 42.5566 28.6045 43.4112
3
0.825 1.9946 2.859 3.4714 6.8585 4.6665 5.6634
5.3325 11.1879 17.0555 17.5984 35.1087 23.5378 37.7182
4
0.9548 2.2828 3.6708 4.1619 8.2816 5.4108 6.8318
6.4245 13.7089 21.247 22.8687 45.7148 30.5476 45.6573
5
0.2884 1.0307 1.4871 1.4859 3.0579 2.0014 3.034
6.3562 13.181 20.881 22.2129 44.2601 29.5305 44.8127
6
0.3505 0.9309 1.5291 1.9073 3.8917 2.6087 3.3481
6.2356 12.7713 21.8787 23.2145 46.2477 30.3078 46.9685
7
0.2942 0.7504 1.5912 1.811 3.6697 2.3021 3.3906

148
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

6.1795 12.6036 21.912 22.989 45.7623 29.8138 47.0392


8
0.2989 0.7418 1.6373 1.9132 3.8638 2.4359 3.4641
6.348 13.3426 20.2957 21.769 43.1685 28.9464 44.2369
9
0.3564 1.0819 1.1685 1.6587 3.3208 2.4513 2.6888

Table B.3(D)
Mean and Standard Deviation of RIGHT AND LEFT TRIANGLES of Handwritten
Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.5548 13.8805 23.0131 24.3619 48.4112 31.7722 49.5418
0
0.266 1.0033 1.7104 2.0269 3.9984 2.5698 3.5086
6.5675 14.8099 21.0126 23.5244 46.2976 31.4873 47.8236
1
0.3112 1.4725 1.4016 2.0175 3.8037 2.7648 5.7718
5.5576 12.1641 17.3058 17.5482 35.1484 23.8598 37.3945
2
0.5474 1.6704 2.5219 2.7441 5.4878 3.5303 4.8507
5.8796 12.6863 18.558 20.0095 39.7907 27.0586 40.341
3
0.3522 1.0253 1.4859 1.7308 3.5919 2.3503 3.0856
5.1349 10.915 16.495 17.2344 34.3182 23.0288 36.4958
4
0.5594 1.632 2.0773 2.4024 4.786 3.2352 3.9672
5.5783 11.9871 20.8989 21.938 43.7365 28.38 44.6358
5
0.2832 0.609 1.5471 1.7461 3.4443 2.2089 3.0834
4.9089 9.9956 17.2352 18.284 36.4099 23.8603 37.9705
6
0.3637 0.8 1.4856 1.9547 3.9123 2.6468 2.7558
5.6533 12.2418 18.4672 20.3582 40.5928 27.3718 40.4334
7
0.2827 0.7736 1.0494 1.1388 2.396 1.6198 2.3656
5.7279 12.5462 18.5614 20.2097 40.3391 27.3774 40.2971
8
0.2898 0.8294 1.046 1.169 2.5003 1.7968 2.2745
5.7759 12.2098 19.9225 21.0996 42.1115 27.815 42.8155
9
0.2985 0.8386 1.4847 1.635 3.4122 2.2392 2.7342

Table B.3(E)
Mean and Standard Deviation of UPPER AND LOWER TRIANGLES of
Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
6.5551 13.9944 22.5814 23.9224 47.5391 31.3899 48.717
0
0.27 1.0345 1.759 2.0477 4.04 2.6033 3.5286
5.6197 11.3116 21.0777 21.7291 43.3276 27.7463 45.857
1
0.2502 0.5236 1.745 2.0384 4.0998 2.4861 3.389
5.3755 11.427 18.2104 19.7639 39.3611 26.2213 39.5321
2
0.2543 0.6166 0.9959 1.249 2.5987 1.7975 2.3224
5.4696 11.5504 17.5652 19.6529 38.7745 26.0549 39.3177
3
0.2802 0.6511 0.9656 1.4677 2.786 1.9343 2.529
5.5193 11.5205 19.474 20.7037 41.3748 27.2794 41.8203
4
0.2843 0.673 1.3964 1.5058 3.0336 2.008 2.7864

149
Appendix B: Features of Handwritten Arabic Numerals (Trained Data)

5.6735 12.3574 20.7591 20.9868 42.1272 27.4162 43.4211


5
0.3107 0.6967 1.5193 1.5986 3.0939 2.01 2.8349
5.3395 10.9324 17.7463 19.0836 37.8739 25.0423 38.9568
6
0.2988 0.6392 1.2663 1.6358 3.318 2.1975 2.5789
5.4139 11.5851 17.3224 17.6635 35.5481 23.5668 36.9178
7
0.4064 1.0965 1.6952 2.0027 3.9369 2.4569 3.7177
5.6149 12.1653 17.8868 17.661 35.8954 23.8685 36.8871
8
0.4432 1.3117 1.8755 1.9478 3.9742 2.6272 3.6056
5.4209 11.1885 17.7113 18.5043 36.7302 24.2623 38.8304
9
0.336 0.8712 1.283 1.5851 3.1251 2.0336 2.5658

Table B.3(F)
Mean and Standard Deviation of HEART of Handwritten Arabic Numerals image
No. Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7
7.1277 16.6261 25.5021 28.0502 55.4687 37.0613 55.6652
0
0.2118 2.0757 1.8949 2.5699 4.792 3.4515 4.6403
6.0073 12.1514 22.5507 23.4195 46.6976 29.9428 48.7211
1
0.2608 0.5745 2.0694 2.2572 4.5876 2.6796 6.7288
5.9538 13.7348 19.0357 20.5263 41.1033 28.1246 41.0142
2
0.2536 1.1822 0.8848 0.9482 1.9869 1.6217 2.0647
6.282 15.3023 19.2815 22.4474 44.081 30.7902 44.0244
3
0.2256 1.4728 0.8586 1.5006 2.7457 2.2082 2.7078
5.9909 12.9612 20.2044 22.0717 43.9001 29.3797 44.0011
4
0.2542 0.9801 1.509 1.5259 2.9957 1.9655 2.9932
6.0131 16.199 22.6254 22.1825 45.3142 31.0108 45.3397
5
0.3928 1.6729 2.1047 2.095 4.0157 2.7182 3.9954
5.9114 12.5018 18.6007 20.6789 41.1391 27.8084 41.0501
6
0.2473 0.7376 0.8199 0.8484 1.8881 1.3751 1.9015
6.1253 15.2233 19.155 21.1226 41.9749 29.4734 42.1185
7
0.2702 1.5812 0.895 1.0606 2.1278 1.7345 2.2937
6.2081 15.4868 19.5131 21.371 42.5743 29.7974 42.6006
8
0.2553 1.5549 0.9524 0.9684 2.0843 1.6409 2.0851
6.2215 14.1657 20.3093 22.3194 44.3125 30.1288 44.3881
9
0.2331 1.1477 1.0281 1.4274 2.7364 2.0708 2.5647

150

You might also like