Abstract: Movie background music plays an irreplaceable role in strengthening film affection,heightening the dramatic and rendering atmosphere．If movie background music can be automatically classified by affection,it will be remarkable to improve the accuracy of movie affective content analysis undoubtedly．In view of this,movie background music movie background music affection feature vector and classifier were proposed,so that the annotation of movie background music was improved．Movie background music affection vector is consisted of bar-long rhythm patterns,bar-long baseline patterns,Mel Frequency Cepstrum Coefficients (MFCC) and interval features extracted from music audio signal．Compared with other features,rhythm pattern and baseline pattern features are able to demonstrate rhythm structure over movie background music clip．Probability latent semantic analysis is used to classify the movie background music into excitement,tension,relaxation and sadness．Experimental results show that the movie background music affection feature vector and the PLSA classifier effectively improve the accuracy of classes than the state of the art.