多特征融合技术应用于中医舌象分析的初步研究

Primary Study of Tongue Manifestation Multi-Feature Fusion Application

本文对中医舌象信息的融合方案进行了初步探讨.对于舌象多特征融合,通常认为可采用特征层、决策层或特征层与决策层联合融合等不同方案.本文进行有关算法实验的结果表明,由于诸多因素影响,这些方案对于舌象多特征融合难以获得理想的效果.我们认为基于粗糙集理论的方法在舌象多特征确定中医证候方面可以获得比较好的结果,并进行了初步探讨.采用粗糙集理论可根据舌象特征确定部分证候,对于不能确定证候的样本可获得可能的结果,这在一定程度上避免了误判,对于中医辅助诊断是非常重要的.

Several means of tongue manifestation information fusion are analyzed and compared and the means based on feature layer and decision layer and both are available in principle. The experiment results show the correct recognition rate based on them is low. It is very difficult to acquire a good result because of some factors such as much difference between two features. Rough set theory is a valid means to solve this problem. Some symptoms may be defined by the tongue manifestation features and some symptoms cannot be defined by the means based on the rough set theory, which avoids the error judgement in some sense and is crucial to the Chinese medical diagnosis. Rough set theory is hopeful to become an important means in the standardization of the symptom of Traditional Chinese Medicine （TCM）.