Abstract
In recent years, the use of video clip sharing sites and applications has become popular. It is desirable to add appropriate
background music to the video when submitting a video. However, in the past, when searching for a sound source that can be used as
background music, one would have to rely on meta-information such as language information subjectively attached to the database by an owner,
or actually listen to a number of songs to confirm. we think that the video contributors would improve production efficiency if there were a
system that automatically lists up suitable background music for the video, and I implemented such as search system.
First, we constructed a website for the questionnaire survey. Through the questionnaire survey, the impressions of some videos and musics
were scored by using language to describe each. At this time, we also asked the participants to select the top five most suitable background
music for the video. In the proposed method, features for videos are formulated using color histograms and optical flow histograms, and
MFCCs, which are often used in recent deep learning research, are used for music features. A function that converts these features to linguistic
evaluation values was obtained by multiple regression analysis.
To evaluate the proposed system, we calculated the top five music files with the highest similarity to the input video, and checked whether
they were compatible with the respondents' ranking. In addition, MAP and NDGC were calculated to evaluate the system as a retrieval system.
