查詢結果分析
來源資料
相關文獻
- 以類神經網路為基礎之線上手寫中文字辨識系統
- Neural Network Procedures for Taguchi's Dynamic Problems
- A Fast and Efficient Competitive Learning Design Algorithm Based on Weight Vector Training in Transform Domain
- 專家系統振動訊號圖型判別之研究
- 反傳遞模糊類神經網路於流量推估之應用
- 類神經網路(Neural Networks)的種類及其在影像處理上的應用
- C++Fuzzy類神經網路物件導向發展系統之建立
- 臺灣汽保費率之估計--對數線性費率模式與類神經網路之比較
- 運用類神經網路於股價指數之套利--以日經225指數為例
- 使用類神經網路預估碳化鎢材料放電加工性能
頁籤選單縮合
題名 | 以類神經網路為基礎之線上手寫中文字辨識系統= |
---|---|
作者 | 周義昌; |
期刊 | 電信研究 |
出版日期 | 19920600 |
卷期 | 22:3 1992.06[民81.06] |
頁次 | 頁307-316 |
分類號 | 312.84 |
語文 | chi |
關鍵詞 | 線上手寫中文字辨識系統; 類神經網路; |
中文摘要 | 這篇論文主要是介紹:如何利用夕層認知類神經網路來辨?線上手寫中文字(OLHCCR)。這套系統能辨識教育部所公佈的5401個常用字。輸入字型大小不受限制但必需正常書寫。系統的第一部份為原始筆畫座標特徵抽取處理,第二及第三部份是利用類神經網路辨認筆畫種類及字型,最後則是混淆字集的檢查。而信用類神經網路來做分類的理由在於它有非線性切割的能力。這個系統的另一特點提出了一個新的特徵抽取方法,這個方法不同的地方在於沒有「量化」這個步驟。因此特徵向量既可以保留輸入字型的絕大部份特徵,又適合類神經網路分類器的應用。實驗的資料庫收集了8個人的手寫字型,辦識率為98.4%。 |
英文摘要 | In this paper, an on-line handwritten Chinese character recognition (OLHCCR) system based on perception neural network is proposed. The system will be used to recognized 5401 handwritten Chinese characters, where the characters are size flexible, but within the constraint of normal handwriting. On the other hand, an algorithm of recognition system is first proposed. The first part of this algorithm is designed to extract features of stroke from primitive stroke data. And the second part of this algorithm, a stroke classifier is to identify the stroke type by neural network model. The third part of the algorithm, a character classifier is to recognize character by neural network. The lat part of this algorithm is to check confusion set. The main reason of using neural network as classifier is due to its capability of nonlinear discrimination. Novelly, a new feature extraction algorithm is also proposed which does not contain quantization procedure, such that it can both preserve the most information of the input data and suit for the classifier. The database of simulation are collected from 8 people and the recognition rate of inside test is 98.4%. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。