| Abstract [eng] |
The goal of this work is to propose a new algorithm for music indexing and searching. To achieve this, objectives and requirements were formulated for the newly proposed algorithm. State of the art algorithms for music indexing and searching were also examined. Following that, an algorithm that uses Taylor coefficients to distinguish music features was suggested. To compare music features, two algorithm versions were suggested: a quick version th at uses hash keys, and a slow version, using more data to compare the music. The suggested algorithms are tested in the experimental part. Noise immunity and their dependence on the length of the query are checked. The results are compared with those of th e state of the art algorithms. They show that the algorithm is able to recognize a music that has thirty - five decibel noise and only from a thirty seconds query. |
