一种结合端点检测可检错的DTW乐谱跟随算法

A DTW ALGORITHM FOR SCORE FOLLOWING COMBINING ONSET DETECTION BEING ERROR NOTES DETECTABLE

针对乐谱跟随中动态时间规整算法(DTW)时间精确度不高并且无法检测偏差音符的不足,提出一种基于端点检测与DTW算法的乐谱跟随系统。其前端结合端点检测算法,提取基于恒Q变换(CQT)的色度特征值,再将两个音频信号进行比较,最终利用DTW算法实现演奏音频与乐谱对齐功能,确定演奏音频每一时刻具体演奏内容。实验结果表明,提出的算法在时间精确度上较传统方法提高7.64%,并且不受节拍变化的影响,能够检测出演奏音频中是否有偏差音符。

To solve the problem that the dynamic time warping（ DTW） algorithm in score following has low time accuracy and can＇t detect error notes, a score following system based on onset detection and DTW algorithm is proposed. Its front end combines onset detection algorithmand extracts the constant Q transform（ CQT）-based chroma feature,and then compares the two audio signals. Finally,it uses DTW algorithm to realise the function of alignment between the playing audio and the score,and determines the specific playing content of the playingnotes at each moment. Experimental results show that the proposed algorithm improves the precision of alignment time by 7. 64% compared with traditional method with no impact from the beat variation,and can detect whether there are the error notes in playing audio.