Amazon.com, Inc.
Device arbitration using audio characteristics

Abstract:

Described herein is a system for device arbitration using characteristics of audio data captured by multiple devices. The system determines a feature vector corresponding to each device that captured an audio input, where the feature vector includes the audio energy levels, spectral data corresponding to the audio data and centroid data corresponding to the audio data. The feature vectors are processed using a trained component to select a device for further processing.