浏览全部资源
扫码关注微信
北京理工大学电子工程系
纸质出版:1995
移动端阅览
[1]黄德双.三种新的学习子空间模式识别方法的研究[J].电子学报,1995(07):25-30.
HuanR Deshuang. A Study on Three Kinds of New Learning Subspace Methods for Pattern Recognition[J]. Acta Electronica Sinica, 1995, (7).
[1]黄德双.三种新的学习子空间模式识别方法的研究[J].电子学报,1995(07):25-30. DOI:
HuanR Deshuang. A Study on Three Kinds of New Learning Subspace Methods for Pattern Recognition[J]. Acta Electronica Sinica, 1995, (7). DOI:
本文提出了最小模、检错平均、前后向平滑三种新的神经网络学习子空间模式识别方法.这些方法在识别率和收敛速度等整体性能上,接近或超过E.Oja的平均学习子空间方法;特别是,前后向平滑学习子空间方法是目前最好的一类学习子空间方法,在模式识别领域,特别是语音识别方面具有广泛的应用前景,本文就舰船目标与箔条杂波利用这些方法进行了若干分类与识别实验,计算机模拟结果证实了这些方法的有效性.
This paper presents three kinds of new Neural Networks
Learning Subspace Methods(LSM) for Pattern Recognition such as Mininum Norm(MN)
Detecting Error Averaged(DEA)
Forward and Backward Smoothing(FBS).Their total performances in recognition rate and convergence speed approach or surpass ALSM’s by E.Oja.Especially
the FBS-LSM is by now best method in LSMs.They have been widely used in Pattern Recognition
especially
in speech classfication.Finally
these methods are applied to the classification of targets-ships and clutters-chaffs
the results of the computers simulation are given
also.
0
浏览量
138
下载量
5
CSCD
关联资源
相关文章
相关作者
相关机构