Abstract: Wafer defect patterns usually indicate the abnormal sources existing in the manufacturing process. Therefore wafer defect recognition plays a crucial important role in finding the root causes of the out-of-control process. This paper develops a model for wafer defect detection and self-adaption recognition. First of all, feature extraction is applied to different wafer patterns. Then based on modeling the various defect patterns with corresponding Hidden Markov Model（HMM）, a dynamic ensemble scheme HMMs is proposed to detect and recognize defect patterns occurring in wafers. The proposed model is successfully applied to WM-811K wafer bin map database and the experimental results prove the effectiveness of the model.

Key words: semiconductor manufacturing, wafer defect, pattern recognition, hidden Markov model

摘要： 晶圆表面的缺陷通常反映了半导体制造过程存在的异常问题，通过探测与识别晶圆表面缺陷模式，可及时诊断故障源并进行在线调整。提出了一种晶圆表面缺陷模式的在线探测与自适应识别模型。首先该模型对晶圆表面的缺陷模式进行特征提取，基于特征集对每种晶圆模式构建相应的隐马尔科夫模型（Hidden Markov Model，HMM），并提出基于HMM动态集成的晶圆缺陷在线探测与识别方法。提出的模型成功应用于WM-811K数据库的晶圆缺陷检测与识别中，实验结果充分证明了该模型的有效性与实用性。

关键词: 半导体制造, 晶圆缺陷, 模式识别, 隐马尔科夫模型