Adding the patient's preferred music to tinnitus treatment can effectively improve the effect of tinnitus treatment.To meet the personalized music preferences of patients,we proposed a novel method for music automatic tagging based on the fusion of logMel and Hpcp features.Firstly the acoustic and music theory related features of the music were extracted and input to the two-branch Effi-cientNetV2-s network for music annotation.The tested results showed that the ROC-AUC value of this method on the MTAT dataset reached 0.911 9.Compared with other music annotation methods,the annotation performance has been improved to some extent.This study has a certain reference value for tinnitus preference tone selection.
维普
万方数据