OTT Content Recognition
March 1, 2011, OTTCon, San Jose, CA– we talked with Zeitera about their audio and video fingerprinting technology. Over the Top (OTT) TV devices are delivering content to viewers directly over the Internet. In many cases these devices must deliver content from various sources that have inconsistent or incomplete metadata causing problems with the viewer experience, audience metrics, advertising, and recommendation engine data.
Content fingerprinting analyzes unique features of audio and video frames to create a unique key or fingerprint for each piece of content. Fingerprints are generated at millisecond resolution, so it’s straight forward to match even very short sequences for identification. These fingerprints are unique to each sequence and are relatively immune against distortions such as re-encoding, cropping, or aspect ratio changes. Fingerprints are stored in a reference database and unidentified content can be checked against the references files and identified. The database is licensed so users don’t need to worry about ongoing fingerprinting and database costs.
The base system has support for large reference databases (100K+ hours), high speed ingestion and search, and few false positives. In addition to the Content ID appliance or standard servers, the code runs on embedded processors in mobile phones, TVs and set-top-boxes. A fingerprint can be created for either reference or query on any device.
Applications for the tools include the obvious anti-piracy and take-down as well as broadcast monitoring, archive or digital asset management de-duplication, copyright identification and enforcement, and audience metrics. In addition, providers could link specific ads to a segment or make special offerings to known programming viewers from the database matches. As content moves off of the fixed TV, more uses are possible on the many other devices capable of displaying video content.
