Identifying Multimedia Objects Based on Multimedia Fingerprint

An apparatus for training a model for identifying multimedia objects, comprising:

• a fingerprint calculator configured to, for each of one or more samples including a training query multimedia object, a training reference multimedia object and a mark indicating whether the training query multimedia object matches the training reference multimedia object or not, derive training query fingerprints from the training query multimedia object according to fingerprint algorithms F1 to FG respectively, where the fingerprint algorithms F1 to FG are different from each other, and G>1, and derive training reference fingerprints from the training reference multimedia object according to the fingerprint algorithms F1 to FG respectively;
• and
• a training unit configured to:
• for each fingerprint algorithm Ft, generate at least one candidate classifier based on the training query fingerprints and the training reference fingerprints derived according to the fingerprint algorithm Ft, the candidate classifier being adapted to calculate a decision for any two fingerprints derived according to the fingerprint algorithm Ft, which indicates a possibility that the two fingerprints are not derived from the same multimedia content;
• and
• generate the model including a weighted sum of classifiers selected from the candidate classifiers and a threshold THc for evaluating the weighted sum such that the identifying error obtained by applying the model to the training query fingerprints and the training reference fingerprints derived from the samples is minimized.

A method of identifying a multimedia object, comprising:

• acquiring query fingerprints fq,1 to fq,T which are derived from the multimedia object according to fingerprint algorithms F1 to FT respectively, where the fingerprint algorithms F1 to FT are different from each other, and T>1;
• for each fingerprint algorithm Ft, calculating decisions through each of at least one classifier corresponding to the fingerprint algorithm Ft based on the query fingerprint fq,t and reference fingerprints derived from a plurality of reference multimedia objects according to the fingerprint algorithm Ft, each of the decisions indicating a possibility that the query fingerprint and the reference fingerprint for calculating the decision are not derived from the same multimedia content;
• for each of the reference multimedia objects, calculating a distance D as a weighted sum of the decisions relating to the reference fingerprints derived from the reference multimedia object according to the fingerprint algorithms F1 to FT respectively;
• and
• identifying the multimedia object as matching the reference multimedia object with the smallest distance which is less than a threshold THc.

A method of training a model for identifying multimedia objects, comprising:

• for each of one or more samples including a training query multimedia object, a training reference multimedia object and a mark indicating whether the training query multimedia object matches the training reference multimedia object or not,
• deriving training query fingerprints from the training query multimedia object according to fingerprint algorithms F1 to FG respectively, where the fingerprint algorithms F1 to FG are different from each other, and G>1;
• deriving training reference fingerprints from the training reference multimedia object according to the fingerprint algorithms F1 to FG respectively;
• and
• for each fingerprint algorithm Ft, generating at least one candidate classifier based on the training query fingerprints and the training reference fingerprints derived according to the fingerprint algorithm Ft, the candidate classifier being adapted to calculate a decision for any two fingerprints derived according to the fingerprint algorithm Ft, which indicates a possibility that the two fingerprints are not derived from the same multimedia content;
• generating the model including a weighted sum of classifiers selected from the candidate classifiers and a threshold THc for evaluating the weighted sum such that the identifying error obtained by applying the model to the training query fingerprints and the training reference fingerprints derived from the samples is minimized.