The Aligned Continuo Realization Dataset (ACoRD)
收藏资源简介:
ACoRD是由查尔斯大学研究团队构建的首个数字低音即兴演奏符号化数据集,包含7位羽管键琴演奏者对5段乐谱各5次演绎的175条MIDI录音,每条录音平均含105个音符。数据集通过动态时间规整技术实现了演奏与乐谱的精确对齐,并创新性地采用griff表征来捕捉演奏中的和声结构特征。该数据集填补了巴洛克数字低音即兴演奏实证研究的空白,为音乐信息检索、演奏风格分析及历史演奏实践研究提供了重要基础,尤其有助于揭示约束性即兴创作中的个性化表达空间。
ACoRD is the first symbolic dataset of figured bass improvisation developed by the research team at Charles University. It comprises 175 MIDI recordings, with each of the 7 harpsichord performers delivering 5 renditions of each of the 5 musical scores, and each recording contains an average of 105 musical notes. The dataset achieves precise alignment between the recorded performances and the original scores via dynamic time warping (DTW) technology, and innovatively employs the Griff representation to capture the harmonic structural features embedded in the improvisations. This dataset fills a critical gap in empirical research on Baroque figured bass improvisation, serving as a foundational resource for music information retrieval, performance style analysis, and historical performance practice studies. It is particularly valuable for uncovering the individualized expressive spaces inherent in constrained improvisational creation.

- 1Beyond Rules: Towards Basso Continuo Personal Style Identification查尔斯大学 · 2026年



