Abstract:
There is a disadvantage of the feature reduction based on granular computing. It is not suitable for the accurate data. This paper presents a method to avoid this disadvantage that is using arithmetic average. The time complexity of this method is polynomial. It uses Support Vector Machine(SVM) to do pattern identification in every character. Then try to compare every recognition rate with the average recognition rate. When the decision table becomes simplified, it can be reduced in granular computing. Comparing the result with the original features, experiment shows that this method can remove redundancy, and it is not effected by adding noise.
Key words:
granular computing,
feature reduction,
decision table,
arithmetic average,
human shap recognition,
adding noise
摘要: 为将基于粒度计算的属性约简方法应用于人形特征的筛选,避免传统方法难以准确消除冗余的缺点,提出一种基于算术平均数的粒度计算方法。采用支持向量机对目标图片进行人形识别,记录不同特征参数下的识别率,求出其算术平均数,并与单个数据做比较,简化决策表后通过粒度计算的方法约简人形特征。将得到的约简特征与原始特征在相同测试集下做性能比较,实验结果表明,该方法能消除冗余,提高识别的性能,且鲁棒性较好。
关键词:
粒度计算,
特征约简,
决策表,
算术平均数,
人形识别,
加噪
CLC Number:
HE Nian, DAN Yong-Zhao, CHENG Ke-Yang. Human Shape Feature Reduction Based on Granular Computing of Arithmetic Average[J]. Computer Engineering, 2012, 38(3): 193-195,199.
何念, 詹永照, 成科扬. 基于算术平均数粒度计算的人形特征约简[J]. 计算机工程, 2012, 38(3): 193-195,199.