Smart Virtual Bass Synthesis algorithm based on music genre classification

The aim of this paper is to present a novel approach to the Virtual Bass Synthesis (VBS) algorithms applied to portable computers. The proposed algorithm employed automatic music genre recognition to determine the optimum parameters for the synthesis of additional frequencies. The synthesis was carried out using the non-linear device (NLD) and phase vocoder (PV) methods depending on the music excerpt genre. Classification of musical genres was performed utilizing the k-Nearest Neighbor algorithm and the extracted MPEG 7-based feature vectors. To confirm the relationship between the presented music excerpt genre and the listener's preferences, subjective tests were carried out. The pairwise comparison test was performed. Test material consisted of 18 pair samples belonging to six music genres: classical, pop, rock, rap, jazz, electronic. For comparison purposes music samples were prepared with the benchmark MaxxBass system and the Smart VBS algorithm proposed by the authors. On the basis of the listeners' opinions statistical tests were carried out to confirm the validity of adjusting low frequency synthesis settings according to the music content of audio files.