Emotions have an important influence on decision-making. However, due to the limitation of acquisition methods for emotional state, there is little research currently about the linguistic multi-criteria decision-making method considering the influence of emotion state. With the development of Internet of things technology, wearable sensors can easily acquire decision makers' electroencephalogram(EEG) signals and then perceive their emotion states. Therefore, a linguistic multi-attribute decision-making method based on emotion perception is studied. Firstly, an emotion perception method based on the SVM probability output model is proposed. According to the EEG signals of decision makers, the emotional state probability distribution can be obtained accurately in real time. On this basis, a quantitative method of linguistic assessment based on a cloud model is proposed. Considering the fuzziness and randomness of linguistic assessment as well as emotional factors, the linguistic assessment is quantified. Then, the scheme sorting method based on the prospect theory is proposed. On the basis of affective generalization hypothesis, the comprehensive prospect values are calculated and the schemes are sorted. Finally, the feasibility and effectiveness of the proposed method are illustrated by examples.