HKU956: Dataset in "Detecting Music-Induced Emotion Based on Acoustic Analysis and Physiological Sensing: A Multimodal Approach"

Introduction HKU956 is a multimodal dataset for analyzing listeners’ emotions and physiological responses induced by music. Five kinds of peripheral physiological signals (i.e., heart rate, electrodermal activity, blood volume pulse, inter-beat interval, and skin temperature) of 30 participants (18 females) were recorded as they listened to music within a period of 40 minutes. Each participant listened to 10 or more songs, and their physiological signals were aligned with the 956 listening records. Participants reported their emotions induced by each song in the arousal and valence dimensions in a scale of [-10, 10]. In addition, participants' personality traits were measured by the Ten Item Personality Measure (TIPI). The raw score in each of the five personality dimensions were presented, together with “high” or “low” categories derived from the TIPI norms provided by Gosling (https://gosling.psy.utexas.edu/scales-weve-developed/ten-item-personality-measure-tipi/). Last but not least, the original audio files of a total of 592 unique music pieces were presented in this dataset. These audio files were originally obtained from Jamando.com with CC-BY Licenses. <br> Update ├─14.09.2022(GMT+8) - First online date ├─31.10.2022(GMT+8) - Posted date └─07.11.2022(GMT+8) - (1) Add the "play_duration" (i.e., Length of time a user plays a song) column into the file "3. AV_ratings.csv"; (2) Fill the missing value of the "valence_rating" column for the record with participant_id=hku1929, song_no=7, song_id=1119024 in the file "3. AV_ratings.csv" <strong>Please cite the following paper when using this dataset</strong> Hu, X.; Li, F.; Liu, R. Detecting Music-Induced Emotion Based on Acoustic Analysis and Physiological Sensing: A Multimodal Approach. Applied Science. 2022, 12, 9354. https://doi.org/10.3390/app12189354 (free text available through Open Access)

数据集介绍 HKU956是一款用于分析音乐诱发听众情绪与生理反应的多模态数据集。本数据集采集了30名参与者（其中18名为女性）在40分钟音乐聆听过程中的5类外周生理信号，分别为心率（heart rate）、皮肤电活动（electrodermal activity）、血容量脉搏（blood volume pulse）、心搏间期（inter-beat interval）与皮肤温度（skin temperature）。每名参与者聆听至少10首乐曲，其生理信号与956条音乐聆听记录完成对齐匹配。参与者需针对每首乐曲诱发的情绪，在唤醒度（arousal）与效价（valence）两个维度上以[-10, 10]的量表进行自评。
此外，本数据集采用十项人格量表（Ten Item Personality Measure, TIPI）对参与者的人格特质进行测量，公开了5大人格维度的原始得分，并附带基于Gosling提供的TIPI常模划分的「高」「低」类别（数据来源：https://gosling.psy.utexas.edu/scales-weve-developed/ten-item-personality-measure-tipi/）。
最后，本数据集包含共计592首独立乐曲的原始音频文件，这些音频最初从Jamando.com获取，采用CC-BY许可协议。
更新记录
├─2022年9月14日（GMT+8）：首次上线日期
├─2022年10月31日（GMT+8）：发布日期
└─2022年11月7日（GMT+8）：(1) 在「3. AV_ratings.csv」文件中新增「play_duration」字段（即用户播放单首乐曲的时长）；(2) 补全「3. AV_ratings.csv」中参与者ID为hku1929、乐曲编号为7、乐曲ID为1119024的记录的「valence_rating」（效价评分）字段缺失值

**使用本数据集时请引用以下论文**：
Hu, X.; Li, F.; Liu, R. 基于声学分析与生理感知的音乐诱发情绪检测：一种多模态方法. 应用科学（Applied Science）. 2022, 12, 9354. https://doi.org/10.3390/app12189354（可通过开放获取获取免费文本）