In an online study, N = 517 participants rated 48 very short musical stimuli comprised of well-known pop songs with regard to arrangement parameters and cross-modal variables. Identification rates for songs and artists ranged between 0-7%. We observed associations between increasing stimulus durations as well as structural sections (chorus or verse) and detection rates. Analyses of the cross-modal variables revealed a main factor, representing the perceived ‘orderliness' of a plink as a strong predictor for title recognition. When psychoacoustic low-level features were entered, Spectral Entropy became the main predictor. The presence of a singing voice additionally seemed to facilitate recognition processes.

中文翻译：

---

持续时间、歌曲部分、熵：对快速音乐识别过程模型的建议

在一项在线研究中，N = 517 名参与者就编曲参数和跨模态变量对 48 首由著名流行歌曲组成的非常短的音乐刺激进行了评分。歌曲和艺术家的识别率在 0-7% 之间。我们观察到增加刺激持续时间以及结构部分（副歌或诗​​歌）与检测率之间的关联。对跨模态变量的分析揭示了一个主要因素，将感知到的 plink 的“有序性”作为标题识别的强预测因子。当输入心理声学低级特征时，谱熵成为主要预测指标。歌声的存在似乎也促进了识别过程。