MỤC LỤC
MỞ ĐẦU
6
CHƢƠNG I TỔNG QUAN VỀ MẠNG NƠ -RON NHÂN TẠO VÀ HỆ
THỐNG NHẬN DẠNG VĂN BẢN
8
1.1. Giới thiệu về mạng nơ ron
8
1.1.1. Mô hình mạng nơ-ron nhân tạo
8
1.1.2. Ƣu và nhƣợc điểm của mạng nơ-ron
12
1.2. Phân loại mạng nơ ron
14
1.3. Hệ thống nhận dạng văn bản
17
1.3.1. Các hoạt động tiền xử lý
20
1.3.2. Phân tích trang văn bản
22
1.3.3. Trích chọn đặc trƣng
27
1.3.4. Học và nhận dạng
33
1.3.5. Hậu xử lý
34
3.2.3. Hậu xử lý
85
3.3. Kết quả đạt đƣợc
85
3.3.1. Kết quả nhận dạng các ký tự riêng lẻ
86
3.3.2. Kết quả nhận dạng văn bản
86
KẾT LUẬN
89
TÀI LIỆU THAM KHẢO
92 - 6 -
MỞ ĐẦU
Mạng nơ ron nhân tạo ngày nay đang là một lĩnh vực nghiên cứu nóng
hổi, thu hút sự quan tâm đặc biệt của các nhà khoa học trong nhiều lĩnh vực. Đó
là nhờ những thành công rực rỡ cả về mặt lý thuyết và ứng dụng của nó. Phạm
vi áp dụng của mạng nơ ron rất rộng lớn: trong lĩnh vực xử lý, điều khiển nhƣ
xử lý tín hiệu, khử nhiễu, phân lớp, nhận dạng, dự báo. Ngoài ra mạng nơ ron
còn đƣợc ứng dụng trong các lĩnh vực toán học, y học, kinh doanh, tài chính,
nghệ thuật. Mạng nơ ron đang ngày càng trở thành một hƣớng nghiên cứu, một
giải pháp ngày càng hứa hẹn.
Một trong những ứng dụng quan trọng của mạng nơ ron là nhận dạng và
phân loại mẫu. Khả năng học và thích ứng của chúng đã làm cho chúng là lựa
chọn hàng đầu đối với nhiệm vụ so sánh các tập dữ liệu hoặc trích chọn các mẫu
thích hợp từ các dữ liệu phức hợp. Nhận dạng mẫu trong mạng nơ ron là một
lĩnh vực rất rộng, nhƣng phổ biến nhất vẫn là nhận dạng chữ in và chữ viết tay.
Chƣơng ba trình bày về hệ thống nhận dạng chữ Việt in sử dụng mạng
nơ ron.
- 8 -
CHƢƠNG I
TỔNG QUAN VỀ MẠNG NƠ RON NHÂN TẠO VÀ HỆ THỐNG
NHẬN DẠNG KÝ TỰ 1.1. Giới thiệu về mạng nơ-ron
Mạng nơ-ron nhân tạo đƣợc xây dựng từ những năm 1940, nhằm mô
phỏng một số chức năng bộ não của con ngƣời. Nếu nhìn não từ góc độ tính
toán, chúng ta dễ thấy rằng cách thức tính toán của não khác xa với tính toán
theo thuật toán và chƣơng trình truyền thống. Sự khác biệt thể hiện ở hai điểm:
+ Quá trình tính toán đƣợc tiến hành song song và phân tán trên nhiều nơ
ron gần nhƣ đồng thời
+ Tính toán thực chất là quá trình học chứ không phải theo một sơ đồ định
sẵn từ trƣớc.
Mạng nơ-ron nhân tạo đƣợc thiết kế tƣơng tự nhƣ nơ-ron sinh học, sẽ có
khả năng giải quyết hàng loạt các bài toán mà sự suy luận tƣơng đối “mờ”,
không đầy đủ dữ liệu, cần có quá trình “học” từ các ví dụ.
1.1.1. Mô hình mạng nơ-ron nhân tạo
1.1.1.1. Nơ-ron sinh học
Bộ não ngƣời có khoảng 10
11
tế bào thần kinh kết nối với mật độ cao gọi
là các nơ-ron (một nơ ron có khoảng 10
4
liên kết). Có nhiều loại nơ-ron khác
(i) Dạng tuyến tính:
N
1j
jiji
pwn
(1.1)
(ii) Dạng toàn phƣơng:
- 10 -
N
1j
2
jiji
pwn
(1.2)
(iii) Dạng mặt cầu:
N
Hình 1.3. Mô hình một nơ ron
class="bi x35 y41 wa h18" src="data:image/png;base64,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