论文部分内容阅读
音符起点检测似中的一个棘手问题是:检测门限不仅因乐曲的不同而不同,而且对同一首乐曲的不同段落也不一样。此前提出的基于音符平均能量NAE(NoteAverageEnergy)的时域方法,虽然摆脱了门限选择问题,但它要求乐曲功率络的音符模式,要么是硬模式(快起慢落)要么是软模式(慢起快落)。智能检测方法首先对整首乐曲功率包络的变化按不同模式划分为若干段落,然后对不同的段落施以不同的检测准则,这使它能胜任更加复杂的混合模式。在成功识别音符模式的基础上,漏检音符的查找策略使智能检测方法的检测率显著提高。对各种乐器和曲风的乐曲所做的大量实验表明:智能检测方法能够准确检测出80%以上的音符。结合了多通道处理技术的智能检测方法,使检测率又提高了10%。
One of the thorny problems in note starting detection is that the detection threshold varies not only by the song, but also by the different paragraphs of the same song. The previously proposed time-domain approach based on the note-average energy NAE (NoteAverageEnergy), while getting rid of the threshold selection problem, requires that the note power mode be either hard (soft slow) or soft Fast drop). The intelligent detection method first divides the power envelope change of the entire music piece into several paragraphs according to different modes and then applies different detection criteria to different paragraphs, which makes it capable of more complicated mixed modes. Based on the successful recognition of the note mode, the search strategy of missing note detection significantly improves the detection rate of the intelligent detection method. A large number of experiments on various instruments and genres of music show that: intelligent detection method can accurately detect more than 80% of the notes. The combination of multi-channel processing technology, intelligent detection method, the detection rate and increased by 10%.