You are here

Acoustic event detection and classification

Acoustic Event Detection (AED) and Classification consists in detecting the occurrence of an event embedded in one or more audio streams,
in classifying the nature of the corresponding sound source, and establishing the related time boundaries.

The audio sequences generally derive from a real-time acquisition system, based on either a single or a multi-channel microphone set-up. Otherwise, the sequences are available in a uncompressed or compressed recorded audio/multimedia format.

In speech recognition and speech communication tasks, the detection problem is typically referred to as Voice Activity Detection (VAD) or Speech Activity Detection (SAD).

Related keywords: acoustic event modeling, speech activity detection, speech/non-speech segmentation.