Driver State Detection from In-Car Camera Images

A non-neglectable number of car accidents are caused by driver's loss of ability to drive the car, which may be caused by serious health problems, e.g. heart attack, stroke, drug or alcohol influence, as well as by drowsiness and other problems. In this paper, a method is presented for detecting the anomaly situations during driving. The method is based on detecting the particular parts of driver's body in the sequence of images obtained from an in-car camera. A feature vector containing the distances between the body parts and describing the situation in a chosen number of frames is computed and used for detection. For the detection itself, the neural network of the autoencoder type containing the LSTM units is used. The method is compared with some other methods; the results show that the method is useful. Moreover, the video sequences used for training and testing are presented, which may be regarded as an additional contribution.