提出一种利用乐音频率变化特征进行音符起始点检测的算法,该算法基于连续小波变换(CWT)对音乐信号进行分解,分解后的信号频率和乐音频率分布相同。从分解后的信号中提取音符起始点特征,使用双边指数光滑和移动窗口归一化算法优化检测函数,产生音符起始点。
This paper proposes a new algorithm for note onset detection,which is based on the Continuous Wavelet Trans-form(CWT) and the music knowledge.The musical signals are decomposed into the frequencies whose distribution are identi-cal to music tones,and note onset features are extracted.The double side exponential smoothing method and normalization techniques based on moving window are employed to detect the onsets.