The Problem Statement
For building a fully-fledged console application for advertisement recognition purposes, the client required a technology partner with experience in developing machine learning algorithms. In this desktop based application, the input to the application would be an audio clip along with a timestamp file of ad breaks and the output would be its matching advertisement clip information from the master database. Also, the client wanted to analyze the audio stream to report any new advertisement that is not there in the already existing master database.
KritiKal partnered with the client to design and develop an intelligent advertisement recognition engine that analyzes advertisements and can find any new advertisements during the advertisement breaks. Leveraging its expertise in machine learning, speech and pattern recognition, KritiKal Solutions developed an indigenous console application that processes a given audio clip recording of an FM radio stream in .wav or .mp3 format. During the processing, it automatically recognizes & detects the advertisements in an ad break and matches them against an advertisement master database.
The output of processing each advertisement break would be a list of advertisements including brand name, start time in input clip, end time in input clip, start time in recognized advertisement clip, end time in recognized advertisement clip, & matched audio clip file name from master database. Our talented team worked on MATLAB to deliver a full-fledged prototype with utmost accuracy. KritiKal’s developed audio analytics engine is capable of integration with various modules that interact with the processing engine, the user interface, the SQL database and other helper modules.
This module can prove to be a great tool for automatically gathering data on advertising spends by competitor radio channels. It can help in significantly slashing down the cost of audio recognition done on manual basis by the ad agencies & media companies. Moreover, it can deliver a quick turnaround time by automating the advertisement recognition tasks.