In this paper, a method to detect driver emergency-steering intention by using Electromyography (EMG) signals is presented. An emergency driving situation, where a driver tries to avoid collisions by steering alone, is anticipated. EMG signals are acquired from the upper limb, while the driver steers a vehicle under normal and emergency driving situations. A linear discriminant analysis (LDA) and multilayer perceptron (MLP) classifiers are constructed to detect the emergency steering from normal steering during driving. Our experimental results show that the emergency steering situation can be detected from EMG signal at about 607ms after the onset of visual stimuli and before a collision happens with an accuracy of greater than 86%.