Music mood can express inherent emotional meaning of a music clip. It's used in music recommendation, music information retrieval, and music classification. In this paper, we follow the Thayer's emotion plane, and extract three different features sets to apply the Chinese popular music mood-detection. We find that the distribution of music moods is quite different from west popular music. Moreover, some feature extract tools which are developed for west popular music aren't suitable for Chinese popular music. In our experiment, we show that the valence dimension is harder to classification (best average precision: 64%) than arousal dimension (best average precision: 86%). Finally the support vector machine, k-nearest neighbors and Naïve Bayes algorithm are used to classifier the music mood. The performance of `exuberance' mood is totally satisfactory, while the `depression' and `contentment' mood are hard to distinguish.