摘要
分析了增强典型相关方法(Enhanced Canonical Correlation Analysis,ECCA)的不足,提出ECCA与偏最小二乘方法(Partial Least-Square,PLS)相结合的多特征线性融合识别模型.为解决多特征非线性融合问题,将增强典型相关方法推广到核空间中,得到了核增强典型相关方法(Kernel ECCA,KECCA).最后,将KECCA+PLS模型用到水下底质回声识别.四种底质回声识别实验表明,采用KECCA+PLS模型,识别效果得到进一步改善.
The shortage of Enhanced Canonical Correlation Analysis（ECCA） is analyzed and a kind of multiple features linear fusion and recognition model is proposed by combining ECCA and Partial Least-Square（PLS） method.In order to solve multiple features nonlinear fusion problem,Kernel ECCA（KECCA） method is proposed by spreading ECCA into kernel space.Finally,the proposed KECCA＋PLS model is used on underwater echoes recognition.The experimental results on four kinds of underwater materials show that using the proposed KECCA＋PLS model improves the recognition effectiveness.
出处
《湘潭大学自然科学学报》
CAS
CSCD
北大核心
2011年第1期94-97,112,共5页
Natural Science Journal of Xiangtan University
基金
国家自然科学基金项目(50875265)
中国博士后科学基金项目(20080440992)
湖南省科技支撑计划项目(2009SK3159)
