关键词: 疲劳识别, 保局投影, 自适应邻域选择, 模糊k近邻

Abstract: Driver fatigue detection is attractive in Intelligent Transportation Systems（ITS）.It has great theoretical significance and practical application value.A new method is proposed for fatigue recognition based on adaptive locality preserving projections.Firstly，fatigue feature is extracted using locality preserving projections which adopt compression or expansion to select the neighborhoods.It can improve the correlativity of the data and maintain the local-geometric structure.Then，the fatigue expression is recognized with fuzzy k-nearest neighborhood classification.The result shows that it is powerful and effective for fatigue recognition.

Key words: fatigue detection, locality preserving projections, adaptive neighborhood selection, fuzzy k-nearest neighborhood