Interesting section extracting device, interesting section extracting method

An interesting section extracting method for an interesting section extracting device, comprising:

• storing anchor models that respectively represent features of a plurality of types of sound pieces;
• obtaining a specified time;
• calculating likelihoods for feature vectors that represent features of an audio signal, in one-to-one correspondence with unit sections of the audio signal by using the anchor models, and generating likelihood vectors having the calculated likelihoods as components thereof;
• and
• calculating a candidate section, which is a candidate for an interesting section to be extracted from the video file, by using the likelihood vectors and extracting, as the interesting section, a part or a whole of the candidate section including the specified time,
• wherein the interesting section extracting method extracts the interesting section interesting to a user from the video file with reference to the audio signal included in the video file such that the specified time is included in the interesting section.

A non-transitory computer-readable recording medium storing a program causing an interesting section extracting device to perform steps comprising:

• storing anchor models that respectively represent features of a plurality of types of sound pieces;
• obtaining a specified time;
• calculating likelihoods for feature vectors that represent features of an audio signal, in one-to-one correspondence with unit sections of the audio signal by using the anchor models, and generating likelihood vectors having the calculated likelihoods as components thereof;
• and
• calculating a candidate section, which is a candidate for an interesting section to be extracted from the video file, by using the likelihood vectors and extracting, as the interesting section, a part or a whole of the candidate section including the specified time,
• wherein the interesting section extracting device extracts the interesting section interesting to a user from the video file with reference to the audio signal included in the video file such that the specified time is included in the interesting section.

An interesting section extracting integrated circuit for an interesting section extracting device, comprising:

• an anchor model storage unit storing anchor models that respectively represent features of a plurality of types of sound pieces;
• a specified time obtaining unit configured to obtain a specified time;
• a likelihood vector generating unit configured to calculate likelihoods for feature vectors that represent features of an audio signal, in one-to-one correspondence with unit sections of the audio signal by using the anchor models, and generate likelihood vectors having the calculated likelihoods as components thereof;
• and
• an interesting section extracting unit configured to calculate a candidate section, which is a candidate for an interesting section to be extracted from the video file, by using the likelihood vectors and extract, as the interesting section, a part or a whole of the candidate section including the specified time,
• wherein the interesting section extracting device extracts the interesting section interesting to a user from the video file with reference to the audio signal included in the video file such that the specified time is included in the interesting section.