APPLICATION OF GPU-CUDA PARALLEL COMPUTING TO THE SMITH-WATERMAN ALGORITHM TO DETECT MUSIC PLAGIARISM

This study introduces the Smith-Waterman algorithm because the advantage of this algorithm is that it can determine the similarity from any position that corresponds to music plagiarism, considering that song similarities can occur in any part of a song. Plagiarism detection can be done by comparing the melody notes of 2 songs to determine whether or not there are similarities. Songs that are identified as plagiarism have similar melodies of 8 bars. However, the Smith-Waterman algorithm has a weakness, namely that the speed of this algorithm is relatively slow, so parallel computing is required to speed up the detection process. Parallel computing relies on the capabilities of multi-core GPUs that can be programmed using CUDA. Therefore, the innovation raised in this study is to speed up the computing process in detecting music plagiarism by applying parallel computing to the Smith-Waterman algorithm. The methodology stages begin with melody extraction, namely taking the song melody from the MIDI file along with the melody's tempo in the MIDI file and then transposing it to the basic tone of C. The study's results showed that using the GPU can speed up the execution time by up to 5.7 times compared to using the CPU. In addition, validation was carried out with real music plagiarism cases and validation of the results using the MIPPIA website. This shows that parallel computing has been successfully applied to the Smith-Waterman algorithm in detecting music plagiarism.