REPET (REpeating Pattern Extraction Technique)
- 技術優勢
- Entirely automated – no userinput neededNo system “training” neededSimple processing – nocomplex frameworks needed
- 詳細技術說明
- REPET uses a simple but reliable technique for music/voice separation.
- *Abstract
-
Currently availablemusic separation methods are more demanding and complex, demanding system“training,” user designation of special audio features, and extensive processingtime to support their complex frameworks. REPET demands none of this, and is simple, fast and completelyautomatable. Evaluationof 1,000 song clips showed that REPET achieves better separation performance than existing automaticapproaches, and in much simpler fashion.
REPET exploitscore principle in music– repetition. This especially applies to popular songs. REPET separates music from voice, simply by extracting the repeating musical structure. It first findsthe period of the repeating structure. Then, the spectrogram is segmentedat period boundaries and the segments are averagedto create a repeating segment model. Finally, each time-frequency bin in a segment is compared to the model, and the mixture is partitionedusing binary time-frequency masking by labeling bins similar to themodel as the repeatingbackground. Because REPET utilizes“self-similarity” between repeating segments, it works on a variety of audiosignals having one or more repeating patterns within a recording.
- 國家/地區
- 美國
