Creating Recognizers to Detect Vocalizations

Song Scope Recognizers are used to compare vocalizations found in long field recordings against a specific vocalization of interest using unique and sophisticated digital signal processing algorithms (patents pending) developed by Wildlife Acoustics.

Recognizers are built from training data, a collection of recorded vocalizations representative of the vocalization of interest. In other words, you need to start with confirmed recordings of the species you are interested in finding to build a recognizer. First, you annotate these recordings to indicate which vocalizations belong to the class and/or subclass corresponding to the vocalization of interest. Once built, you can then use the recognizer to quickly search long field recordings for matching vocalizations.

From the "File" menu, select "New Recognizer" to start a new recognizer. Alternatively, you can open a previously generated recognizer .ssr file.

There are a number of issues to consider when choosing the parameter settings and training data to use in building a recognizer discussed in the next sections, Pattern Recognition Basics and Signal Detection.

You can then Import and Review Training Data from your annotated recordings, and Generate and Save your recognizer.

Finally, you can use your recognizers to Search Recordings for Vocalizations.

Previous Page Next Page