An infringement detection system for videos based on audio fingerprint technology

The system designs and implements a copyright detection system for videos based on audio fingerprint technology, which consists of two parts: the construction of the fingerprint dataset and the detection of infringing videos. In the audio fingerprint extraction algorithm, the audio fingerprint is constructed by using the peak point extraction method based on the auditory mechanism and the method of using the adjacent peak point pairs, which ensures that the fingerprint feature has a high differentiation and robustness. In the audio fingerprint retrieval algorithm, the stability and efficiency of fingerprint retrieval are ensured by using the combination of hash function and inverted indexing method and the method of flexible matching of fingerprints within a certain threshold range. The validity of the system method is proved by verification in two datasets.