Music Information Retrieval Trough Melodic Similarity Using Hanson Intervallic Analysis

In music information retrieval, the melodic similarity is used as main feature for the detection of relevant information. Among the possible applications there is a detection of plagiarism of ideas exposed by an artist in the past, the payment for copyrights using detection of musical pieces in radio transmissions, the computer-aided composition, etc. There exist several techniques exposed in melodic similarity that use diverse statistical and probabilistic analysis. The objective in this work is to establish a text words equivalent to musical notation using a representation based on intervallic relationships and duration, and to evaluate three textual information retrieval techniques applied to this representation, as well as to propose changes to improve the system's performance.