基于时间序列分析的工尺谱自动翻译研究

Research on Automatic Interpretation of Gongchepu Based on Time Series Analysis

工尺谱是现存数量最多的中国传统乐谱.今人解读工尺谱的最大难点在于处理不确定的节奏型,即工尺谱只规定了节拍的起始位置,并未对音符的具体时值进行分配.本文探讨了工尺谱自动翻译成五线谱的必要性与合理性,并提出了一种基于时间序列分析的自动翻译模型.实验中,该模型的正确率为88.25%,与隐马尔可夫模型和条件随机场模型相当,并且能够避免OOV(Out Of Vocabulary)问题.本文的自动翻译模型,并非旨在代替人工翻译,而是希望将人的经验通过数据的形式,建立起可重复使用并不断更新的统计模型,从而使得中国传统音乐的实践经验得以更便捷地传承下去.

Gongchepu is the most existing Chinese traditional musical score.The major obstacle for today＇s researchers is the uncertain rhythm pattern,which means that the score only specifies the starting position of the beat but does not specify the duration of the notes.This paper discusses the necessity and rationality of the automated interpretation of Gongchepu into staves,and proposes an automatic translation model based on time series analysis.In the experiment,the precision of this model is 88.25%,which almost reached the precision of Hidden Markov Model and Conditional Random Field Model,and can avoid the issue of OOV（Out Of Vocabulary）.The model of this paper is not intended to replace human interpretation,but to preserve the human wisdom that Chinese people condense from generation to generation in the oral practice in the form of data,and through the use of statistical models and computers.