i need to detect face and eyes and mouth to detect a drowsy person. i found face and eyes with xml files of opencv. but it seems finding the mouth using haarcascade is not possible.because it doesnt work good.thats why i Decided to use an algorithm to do it.but using canny and sobel etc didn t help me to find mouth.Specially in images with low Quality.creating a new xml file also is hard. then now i dont know how can i find mouth and detect yawning??i use opencv c++ .may i ask you to help me for it?