Direction-of-arrival estimation under noisy condition using four-line omni-directional microphones mounted on a robot head

Tetsuji Ogawa, Kosuke Hosoya, Kenzo Akagiri, Tetsunori Kobayashi

Research output: Contribution to journalConference articlepeer-review

Abstract

We propose a new direction-of-arrival (DOA) estimation method suitable for autonomous mobile robots. Autonomous mobile robots have to meet physical constraints of signal processing devices, such as a space-saving microphone arrangement and few computational resources. In addition, DOA estimation of the robots needs to be robust to noise around the robots. In order to cope with the physical constraints, we used four-line omni-directional micro mechanical systems (MEMS) microphones. DOA estimation was conducted using statistical pattern recognition in which normalized spectral amplitudes, which were free from sound sources, were used as DOA features. In the proposed method, strict head related transfer function estimation, which is not practically feasible, is not needed. In addition, unlike many conventional methods, phase information is not explicitly used because the phase information is unreliable in the situation that we deal with, i.e., situations in which the microphone spacings are small, or strong reflections and diffractions occur around the microphones. The feature vectors we used can cope with these problems. Effectiveness of the proposed method was experimentally investigated in recognition of 19 DOAs in the presence of diffuse noise: the proposed method achieved a DOA correct of approximately 99% at a SNR of 0 dB.

Original languageEnglish
Pages (from-to)879-883
Number of pages5
JournalEuropean Signal Processing Conference
Publication statusPublished - 2009 Dec 1
Event17th European Signal Processing Conference, EUSIPCO 2009 - Glasgow, United Kingdom
Duration: 2009 Aug 242009 Aug 28

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Direction-of-arrival estimation under noisy condition using four-line omni-directional microphones mounted on a robot head'. Together they form a unique fingerprint.

Cite this