Research on computer-aided music generation based on user-tag-media semantic mining

In order to simulate the generation of a given music by computer, we analyzed the time series of music signals, explored the representation methods of some of its features, defined the corresponding characteristic parameters, and discussed the mathematical modeling of the time series of music signals. A computer-aided music generation technology based on user-tagmedia semantic mining based on a comprehensive reasoning model is proposed. First, the inference source is constructed according to the underlying characteristics of the user-tag-media object.. Influence source is constructed according to the symbiotic relationship of the user-tag-media object. Then, it fielded to perform comprehensive reasoning and construct the user-tagmedia semantic space; then for different retrieval examples, according to pseudo-relevance feedback. Different retrieval methods are adaptively selected for each retrieval music generation example. In order to deal with retrieval examples, training is not required In the case of the collection, a two-stage learning method is proposed to complete retrieval; at the same time, a log-based long-range feedback learning algorithm is proposed to improve system performance. Experimental results prove that this technology can accurately mine user-tag-media semantics. User-tag-media document retrieval and user-tag-media retrieval are accurate and stable.