Projection based hashing that balances robustness and sensitivity of media fingerprints

A method comprising:

• extracting a plurality of N features from media content that comprises one or more of audio or video content, wherein N comprises a first positive integer;
• computing a set of K optimal projection matrices from the N media content features, wherein K comprises a second positive integer;
• and
• deriving media fingerprints from the media content with the computed K optimal projection matrices.

A method, comprising:

• identifying a plurality of candidate features that are components of one or more of audio or video content and a plurality of candidate hash functions, wherein each of the candidate hash functions comprise an array of coefficients that relate to the candidate features;
• selecting a subgroup of the hash functions, based at least in part on an optimized combination of at least two characteristics of the candidate features or the hash functions;
• and
• deriving fingerprints that uniquely identify the audio or video content from the selected optimized hash functions subgroup.

A method comprising:

• extracting a plurality of N features from media content that comprises one or more of audio or video content, wherein N comprises a first positive integer;
• computing a set of K optimal hash functions from the N media content features, wherein K comprises a second positive integer;
• and
• deriving media fingerprints from the media content with the computed K optimal hash functions.