Enhanced content-based multimedia recommendation method
AssigneePolitecnico Di Milano
InventorsPaolo Cremonesi, Mehdi Elahi, Yashar Deldjoo
Life of this patent
Estimating the term…
Abstract
There is disclosed a method for generating movie recommendations, based on automatic extraction of features from a multimedia content, wherein the extracted features are visual features representing mise-en-scène characteristics of the movie defined on the basis of Applied Media Aesthetic theory, said extracted features being then fed to content-based recommendation algorithm in order to generate personalized recommendation.
Claims9
A method implemented by a multimedia recommender system for automatically generating movie recommendations from multimedia content, comprising:
- identifying, within a movie stream of a video of the multimedia content, a plurality of shots;
- defining a single key frame within each one of said shots;
- automatically extracting, from the key frames of the plurality of shots, visual features representing mise-en-scène characteristics of the video defined on the basis of Applied Media Aesthetic theory;
- and
- feeding said extracted visual features to a content-based recommendation algorithm that generates personalised recommendation information concerning said multimedia content from the extracted visual features,
- wherein said visual features extracted from the video belong to a visual modality Mv defining the characterization of a movie M=M(MV), where
- MV=MV(fv)
- and
- fv=(Lsh,μcv,μm,μσm2,μlk),
- where
- Lsh is the average shot length in the video,
- μcv is the mean colour variance over the defined key frames of plurality of shots of the video,
- μm and μσm2 are, respectively, the mean motion average and standard deviation across all frames of the video, and
- μlk is a mean lighting key over said key frames,
- wherein the generated personalised recommendation information is output to a user by way of any of a display, a web service, and via plain http for rendering by a client application.
A multimedia recommender system for automatically generating movie recommendations from multimedia content, configured to:
- identify, within a movie stream of a video of the multimedia content, a plurality of shots;
- define a single key frame within each one of said shots;
- automatically extract, from the key frames of the plurality of shots, visual features representing mise-en-scene characteristics of the video defined on the basis of Applied Media Aesthetic theory;
- and
- feed said extracted visual features to a content-based recommendation algorithm that generates personalised recommendation information concerning said multimedia content from the extracted visual features,
- wherein said visual features extracted from the video belong to a visual modality Mv defining the characterization of a movie M=M(MV), where
- MV=MV(fv)
- and
- fv=(Lsh,μcv,μm,μσm2,μlk),
- where
- Lsh is the average shot length in the video,
- μcv is the mean colour variance over the defined key frames of plurality of shots of the video,
- μm and μσm2 are, respectively, the mean motion average and standard deviation across all frames of the video, and
- μlk is a mean lighting key over said key frames,
- wherein the generated personalised recommendation information is output to a user by way of any of a display, a web service, and via plain http for rendering by a client application.
Dependencies are inferred from the claim text.