Commercial Detection Based On Audio Fingerprint

With the development of the multimedia technologies, the broadcast and TV media of china develop quickly. As the component of the broadcast and TV program, broadcast and TV commercial experiences germination, rapid development, the scale of broadcast and TV commercial expands constantly. Although high quality of broadcast and TV commercials play an important part in promoting consumption and market prosperity, some illusive commercials cause inestimable harm to society and pubic interest, and badly affect the impression and credible of media in society. Therefore, it is urgent to solve the problem of the commercial detection. TV commercial includes kinds of digital media such as character, video, audio and so on. As digital fingerprint technology can resolve the accuracy and real time problems in TV commercial detection, it can be used to detect the commercial. In this paper, we study the broadcast and TV commercial detection based on audio fingerprint technology, and focused on studying robust of audio fingerprint extraction and efficient algorithm of commercial detection.Based on the analysis of existing methods, this paper mainly research work are as follows:1. A fast and robust of commercial audio fingerprint extraction method is proposed. Firstly, it obtains the frequency domain information of the commercial audio data for each frame by the Fourier transform. Moreover, frequency domain which is most related to human auditory perception is divided into 33 equal bands. Finally, the summation values of the neighbor frame bands energy are calculated. According to relationship between the values and 0, the 32 bits 0/1 values are extracted for each frame. Experiment results show that the speed of the audio fingerprint extraction and robust for the method are improved much more than those in relevant literatures.2. Twice matching commercial detection strategy based on audio fingerprint is proposed. The first second fingerprint of commercial comparison is adopted in the first match to search the start point frame position of the commercial in the be detected broadcast and TV stream . The whole fingerprint of commercial comparison is used to confirm the start point frame position which is correct or not in the second match. If the second match is successful, output the correct detected commercial. The broadcast and TV stream can move the detected commercial length where is the start point frame for the next march. The improved algorithm can accelerate the speed of commercial detection.