查詢結果分析
來源資料
相關文獻
- 快速學習類神經網路手寫英數與中文字辨認系統
- 以類神經網路為基礎之線上手寫中文字辨識系統
- 類神經網路手寫英數字辨認系統
- 多重專家應用於線上手寫中文文字識別
- 線上手寫中文文字辨認系統的筆段抽取
- On Speeding Radical-Based On-Line Chinese Character Recognition Through the Directional Matching of the First and Last Segments
- A Fast and Efficient Competitive Learning Design Algorithm Based on Weight Vector Training in Transform Domain
- Consonant Classification for Mandarin Syllables
- 傳真手寫數字自動辨識系統
- Reinforcement Learning for Ga-Based Neural Networks
頁籤選單縮合
題 名 | 快速學習類神經網路手寫英數與中文字辨認系統 |
---|---|
作 者 | 周義昌; | 書刊名 | 電信研究 |
卷 期 | 22:3 1992.06[民81.06] |
頁 次 | 頁317-334 |
專 輯 | 類神經網路專集 |
分類號 | 312.74 |
關鍵詞 | 中文字; 手寫; 英數; 學習; 辨認系統; 類神經網路; |
語 文 | 中文(Chinese) |
中文摘要 | 本文提出一個基於分離再合併特徵抽取法與類似Hebbian學習法則分類器之快速學習類神經網路,簡稱為F.L.N.N.,應用於手寫英數與中文字辨認,對任意圖形,經過特徵抽取網路,取出特徵後,比對網路僅須一次即可完成學習或辦認。特徵抽取過程中,分離部份在於抽取出字形之輪廓,而合併部份則在於計算字形輪廓之點密度做為字形之特徵。實驗分成兩部份,第一部份是F.L.N.N.與錯誤回傳模型(Back Propagation)應用於手寫英數字辦認之模擬結果與效益評估比較,結果顯示,因手寫字本身品質而異,當同類字形間偏移角度在20°之內,辦認率平均約95%,在學習速率與辦認率之效益評估上皆優於錯誤回傳類神經網路模型,且因類神經網路具有的大量平行性特點,在辦認速率上則遠優於傳統辦認方法;第二部份是擴充F.L.N.N.以處理筆畫較繁複的手寫中文字,實驗結果顯示,此法可解決一般線上手寫中文字辦認系統所面臨之連筆、筆順變動問題。 |
英文摘要 | In this paper, we propose a Hebbian like classifier based on split-and-merge feature extracting method to set up an Alpha-Numeric and Chinese Character Recongition System. In the feature extracting procedure, the split process is to form the binary maps by capturing the contour of a character with one or more depths, and in the merge process the feature information is extracted by merging the pixel density of the binary maps. Meanwhile, we also propose a new learning algorithm for the Hebbian like neural network to do the character matching. The main advantages of the proposed approach are stated here. First, it possesses high robustness to the variation of handwriting characters from the different people. Second, it needs only one iteration to complete leaning or recognition. Finally, it can not only overcome some problems of on-line Chinese character recondition (OLCCR), such as connected stroke, stroke shape distortion and constraint of stroke order but also provide a new technology in developing the OLCCR/OCR mixed system. |
本系統中英文摘要資訊取自各篇刊載內容。