Study Of Critical Analysis And Intelligent Retrieval Of Emotional Music

Images, audio, video and other multimedia transmitting on the Internet has become an important part of the digital and network era, the retrieval technology based on keywords was unable to meet the digital multimedia information retrieval needs. Otherwise, the retrieval technique based on multimedia content have a high requirement on the equipments and the users themselves, so retrieval technology based on the music emotion rises.Human emotion and music are directly related, the professional emotion of music generated in its creation process, but the popular emotion of music expresses by the massive ordinary users through various forms. Interactive music reviews in forums are the typical music emotion expression forms. Extracting music attributes from the magnanimous web and giving the reasonable emotion results to the searcher are innovative solutions. In this thesis, the definition of music emotion, classification, music reviews analyse and features extraction is introduced mainly in the following aspects to the work:1. Based on the existent theory of the text mining pretreatment, enhancing the dictionary construction and the participle method, and improving the participle efficiency. Defining a synonym structure, and realized the similar computation and search function.2. Presents a fast template-based analysis method of reviews, which achieves a very high rate of accuracy at the same time retaines the broad scope of the application.3. Defines a series of atomic rules, and based on the music-centric clustering of content, proposes a rule-based method to extract music entities. The experiment indicates that the accuracy rate reaches a higher degree.4. Transforms the keywords wxtracted from the music reviews to vector pattern, and improves the computing method of Hevner Emotion Ring, then extracts the emotion features and scene features from the music reviews. Finally, assigns the music which is described in the reviews to the corresponding category. 5. Take the emotion and the scene computation as foundations, proposing an innovation retrieval solution, which can give the searcher the corresponding musics based on any combination of emotion or scene words.The thesis presents an intelligent retrieval solution on music emotion expressed by the music reviews and the development of the corresponding system. The experimental results show that, relative to the current system based on the music itself, this solution is more reasonable, and the results returned to the user are more convincing.