Fast matching for video/audio fingerprinting algorithms

For video and audio copy detection systems, a new windowing scheme and a fast matching algorithm are proposed. The scheme is applicable to all fingerprinting algorithms that are based on temporal windowing of video or audio assets. The proposed method requires a window overlapping strategy for the query different from that of the database. The proposed scheme, enables a state machine-based algorithm for matching the query fingerprints to the corresponding fingerprints of the original video or audio in the database. The proposed novel matching algorithm, decreases the retrieval time by 1 to about 3 orders of magnitude. It also increases the accuracy of localizing parts of the asset that have been copied.