CALLAS

The component takes input from live audio, classifyes audio streams by different sound classes such as speech, music, silence, clapping, whistling and applause, nonlinguistic vocalizations which, additional to audio speech, can give emotional cues of behavior. It provides in output a corresponding audio class for each audio frame. This audio analysis is performed in real time and in live environments, against usual offline content classification of video clips, by dividing them to contain e.g. speech, music, noise or song.

---

Availability of the component: Reference contact is Johannes Peltola of  VTT Technical Research Centre of Finland.
Notes: the component can be made available for non commercial use upon request

---

Besides functional tests, extensive experimentation in CALLAS is done in Proof-of-Concepts:

• the EUCLIDE : interactive installation the component was used through the MultiKeyword Spotting component to detect laughter from live audio