An Accurate and Rapid Advertisement Broadcast Monitoring Method

Abstract

The invention belongs to the technical field of digital audio processing, and provides an accurate and fast advertising broadcast monitoring method, which can be applied to the advertising monitoring of broadcasting stations; Among them, a low-complexity, error-insensitive straight line fitting method is used, combined with two features of audio fingerprint and Mel frequency cepstral coefficient, to realize the rapid positioning of the head position. In the detection of the tail position, frame feature comparison is used The long and short two dynamic time warping methods respectively deal with the alignment and non-alignment of advertisements and slow-recorded audio, and further precisely locate the detected head position during this period, and finally realize the crossover by cleaning and merging detection items. Correct detection of advertisements of two adjacent slow recording blocks; the invention is fast and accurate in processing, has robustness to time stretching, and can be widely used in advertisement monitoring and management of radio stations.

Description

technical field [0001] The invention belongs to the technical field of digital audio processing, relates to an audio feature matching method, and specifically provides an accurate and fast advertisement broadcast monitoring method, which can be applied to advertisement broadcast monitoring of radio stations. Background technique [0002] At present, audio broadcasting has become the most popular mass publicity and entertainment media, and many businesses also use this platform to play advertisements of corresponding products on the radio. In the radio advertising business, radio stations and merchants are very concerned about whether each advertisement required to be broadcast is broadcast on time. Not only that, but the integrity of advertising broadcast is also an important indicator. In this regard, the radio station itself usually needs to monitor the advertisements, monitor the start time and end time of each advertisement broadcast every day, and determine whether the...

AI Technical Summary

Problems solved by technology

One of the biggest problems with this method is that there are too many comparisons, resulting in too long detection time; in addition, this method assumes that the advertisement and the slow-recorded audio clips are aligned, but if one of the two identical audios has a certain difference in time scale Even if the amount of stretching is very small, the comparison will fail; another way is to use the audio fingerprint technology in fast music retrieval to retrieve advertisements in slow recordings. Although this method can solve the problem of detection speed, but Existing technologies are difficult to accurately detect the actual broadcast position of the advertisement, especially when the advertisement is not completely broadcast, this type of technology cannot accurately give the real end position of the advertisement broadcast

On the other hand, radio stations usually need to count the broadcasting of advertisements for a day. In this case, the slow recording file has a duration of 24 hours. If it is directly processed as a whole, the time offset in the audio fingerprint will exceed the range of data representation and The algorithm fails, so the slow recording data needs to be divided into blocks, but the new problem brought about by this is that when the advertisement spans two adjacent blocks, it is easy to cause missed detection or multiple detection errors

[0004] In short, the current advertising monitoring software has not yet solved the detection speed, the accuracy of detecting the broadcast position (start and end time) and the missing or multiple detection of advertisements across slow recording blocks in the block processing.

Images