English

A Deep Reinforcement Learning Approach to Audio-Based Navigation in a Multi-Speaker Environment

Sound 2021-05-17 v1 Machine Learning Audio and Speech Processing

Abstract

In this work we use deep reinforcement learning to create an autonomous agent that can navigate in a two-dimensional space using only raw auditory sensory information from the environment, a problem that has received very little attention in the reinforcement learning literature. Our experiments show that the agent can successfully identify a particular target speaker among a set of NN predefined speakers in a room and move itself towards that speaker, while avoiding collision with other speakers or going outside the room boundaries. The agent is shown to be robust to speaker pitch shifting and it can learn to navigate the environment, even when a limited number of training utterances are available for each speaker.

Keywords

Cite

@article{arxiv.2105.04488,
  title  = {A Deep Reinforcement Learning Approach to Audio-Based Navigation in a Multi-Speaker Environment},
  author = {Petros Giannakopoulos and Aggelos Pikrakis and Yannis Cotronis},
  journal= {arXiv preprint arXiv:2105.04488},
  year   = {2021}
}

Comments

To be published in ICASSP 2021

R2 v1 2026-06-24T01:57:17.627Z