标题: | 基于情绪和潜在因素之影片配乐推荐系统 Mood Based Video Background Music Recommendation via Latent Factor |
作者: | 陈映全 李嘉晃 刘建良 Chen, Ying-Chuan Lee, Chia-Hoang Liu, Chien-Liang 多媒体工程研究所 |
关键字: | 背景音乐推荐系统;基于情绪的;潜在因素;Background music recommendation system;mood based;latent factor model |
公开日期: | 2016 |
摘要: | 随着科技的日新月异,越来越多移动装置都具有拍摄影片的功能,使用者也乐于随 时记录生活中发生的事物;而社群网站的兴起,也让许多人热衷于在网路上分享自己的生活片段。许多使用者会将日常拍摄的片段加上背景音乐,使其更有吸引力,希望能得到更多关注,然而,找寻适当背景音乐的过程往往非常旷日废时,因此我们希望能建构一套系统,辅助使用者更快速地找到适合的背景音乐。本论文提出基于情绪和潜在因素的影片配乐推荐系统,以情绪为影片和音乐的关系依据,并使用潜在因素模型,找出潜藏在影片和音乐特征中的潜在因素,藉由这些潜在因素,推荐使用者适合的背景音乐,并提升推荐的准确度。除了一般推荐系统常用的准确率指标以外,本研究也征求了受测者来进行量化研究和质性研究,透过两种不同的实验,进行问卷研究和访谈研究,希望从不同的角度来分析问题,并验证系统的有效性。而从实验的结果可以发现,本研究所提出的系统在准确度和问卷调查方面,都优于比较的方法,但在质性研究中我们也发现,过半的受测者(60%)是以影片内容和音乐节奏做为配乐适当与否的依据。 The popularity of mobile devices allows people to easily record the splendid moments of their lives and share the videos on the social media website instantly. Embedding appropriate background music to make the video more attractive is a typical use case, but it is a time-consuming and labor-intensive task to find music to fit the video. Inspiring by the requirement, this thesis proposes a background music recommendation system based on mood and latent factors. The proposed system uses mood tags to associate videos with music, and use latent factors between them to recommend appropriate music to users. In the experiments, the evaluation involves accuracy, quantitative research, and qualitative research. The experimental results indicate that the proposed system is promising in accuracy and quantitative research. Furthermore, the investigation shows that over half of the participants (60%) mainly use video contents and music rhythms to judge whether the music is appropriate for the video. |
URI: | http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070256632 http://hdl.handle.net/11536/139222 |
显示于类别: | Thesis |