Learning the precedence effect: initial real-world tests

Abstract

Audio source localization in reverberant environments has proved difficult for automated microphone array systems. Certain features observable in the audio signal, such as sudden increases in audio energy, provide cues to indicate time-frequency regions that are particularly useful for audio localization, but previous approaches have not systematically exploited these cues. We give an overview of a system that we have designed that exploits these cues by learning a mapping from reverberated signal spectrograms to localization precision. We then describe initial tests of the system that demonstrate improved source localization on real audio data using the generalized cross-correlation (GCC) framework. We also relate the system's learned mappings to the well-known precedence effect from psychoacoustic studies.