Music information retrieval research area involves new methods for classification, indexation and retrieval of musical audio signals. One of the main open problems of this area is the estimation of music similarity between - musical pieces. This problem requires multidisciplinary expertise in various - fields such as musical theory, audio signal processing, algorithmic for data - structure comparison, etc. Existing music similarity systems essentially - consider timbre similarity. Our project team aims at investigating the adaptation of algorithms applied in bioinformatics or string recognition to - the musical context. Such algorithms compute a similarity score between two - sequences. First studies led to promising results. Nevertheless the - adaptation to musical audio data raises new problems in both algorithmic and - computer music fields. During this project, several methods that take into - account musical characteristics such as melody, harmony, rhythm, etc. will be - investigated and evaluated according to listening experiments and existing - symbolic and audio databases. Applications of such algorithms are numerous: query-y-humming, query-by-example, automatic summaries, evaluation of musical performances, etc. ...