Abstract: This paper analyses problem of mouth state detection and this problem is great significance in facial expression studies, and driving state research areas. The precise positioning human face is given by using HSV color model and the face area is caculated. Then threshold value is given based on R-B color different between the lip portion and the skin portion and binarization of the photo is given by the determined threshold. Using the ratio of the lip portion and the facial skin area can accurately determine the state of the mouth. The experimental results show that using this methodology can enhance the computing speed compared with single red method and the result will not affected by the distance changes from head to the camera.

Key words: mouth state detection, Hue Saturation Value（HSV）, RGB, image binarization, threshold

摘要： 嘴巴状态的检测对于人脸的表情研究，驾驶员驾驶状态的研究等领域都有着重要的意义。首先采用HSV肤色模型对人脸进行精确的定位，并计算出人脸的面积，根据嘴唇部分与皮肤部分的R-B的色差确定阈值；根据所求的阈值进行二值化，准确地计算出嘴唇部分的面积；根据嘴唇部分与面部皮肤部分的比值，可以准确地判断出嘴巴的状态。与单独的红色排除法相比，计算速度快，不会因摄像头离人的距离产生变化而影响检测结果。

关键词: 嘴巴状态检测, 色度（H）、饱和度（S）、亮度（V）, 彩色模型（RGB）, 图像二值化, 阈值