WAVESSENCE

A unique technology of fingerprints extraction

The core of the technology is a patented digital signal processing method that allows the extraction of digital fingerprints out from the time variations of a waveform, then the location of similarities between signals by simple comparison of their fingerprints.

Compare to watermarking techniques, the method then imposes no particular operation constraints due to insertion and extraction of ID markers within signals.

In the particular case of audiovisual signals, the fingerprint acts just as an additional digital track that can be processed like a continuous index for tracking and identifying similarities between signals.

In practical, the fingerprints once computed can be saved with additional meta data (title, history …) as regular computer files. It is then no more mandatory to keep the original audio files to manage search operations.

For sound recognition purposes, a standard profile used to implement the method allows the identification of program contents by the only use of few seconds’ extracts, and their localization with a “picture frame accuracy”. With such a profile, each fingerprint data characterises a 1/25 second time interval (one TV picture), which leads to extremely compact fingerprints volumes: less than 100 Kbyte per hour of recorded program, less than 1Gbyte for a program database of one year cumulative duration.

The fingerprint comparison processes lies on sophisticated fuzzy logic algorithms, and offers many options for customizing searches, such as choice of false detection levels, noise tolerance …

The results of a comparison characterize common audio extracts. They are given under the form of lists comprising at least the description, the duration and the time offsets of each founded program extract.