eduzhai > Applied Sciences > Engineering >

Visual Attention for Musical Instrument Recognition

  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: In the field of music information retrieval, the task of simultaneouslyidentifying the presence or absence of multiple musical instruments in apolyphonic recording remains a hard problem. Previous works have seen somesuccess in improving instrument classification by applying temporal attentionin a multi-instance multi-label setting, while another series of work has alsosuggested the role of pitch and timbre in improving instrument recognitionperformance. In this project, we further explore the use of attention mechanismin a timbral-temporal sense, à la visual attention, to improve theperformance of musical instrument recognition using weakly-labeled data. Twoapproaches to this task have been explored. The first approach appliesattention mechanism to the sliding-window paradigm, where a prediction based oneach timbral-temporal `instance is given an attention weight, beforeaggregation to produce the final prediction. The second approach is based on arecurrent model of visual attention where the network only attends to parts ofthe spectrogram and decide where to attend to next, given a limited number of`glimpses .

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...