• DocumentCode
    127413
  • Title

    Direction of arrival estimation of kiwi call in noisy and reverberant bush

  • Author

    Gray, Charles ; Hioka, Yusuke

  • Author_Institution
    Dept. of Electr. & Comput. Eng., Univ. of Canterbury, Christchurch, New Zealand
  • fYear
    2014
  • fDate
    18-20 Feb. 2014
  • Firstpage
    258
  • Lastpage
    262
  • Abstract
    This paper outlines the steps taken to accurately perform acoustic source localisation of kiwi calls using microphone arrays in the presence of noise and reverberation. A generalised cross correlation exploiting the reverberation resistance of the PHAT weighting function was investigated and implemented as part of the system. To reduce correlated noise, spectral subtraction and Wiener filtering were investigated, and their suitability in the system was analysed. A novel algorithm exploiting the harmonic structure of the kiwi call is proposed to determine whether or not a kiwi call is present. In a reverberant and noisy environment, the proposed method outperformed the previous study in terms of the estimation accuracy, which had an average deviation of 4.18°.
  • Keywords
    Wiener filters; acoustic radiators; correlation methods; direction-of-arrival estimation; microphone arrays; signal denoising; PHAT weighting function; Wiener filtering; acoustic source localisation; correlated noise reduction; direction of arrival estimation; generalised cross correlation; harmonic structure; kiwi call; microphone array; noisy environment; reverberant bush; reverberation resistance; spectral subtraction; Direction-of-arrival estimation; Estimation; Harmonic analysis; Microphones; Noise; Power harmonic filters; Reverberation; DOA estimation; STFT frame selection; generalised cross correlation; harmonic structure; kiwi call; microphone array;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Sensors Applications Symposium (SAS), 2014 IEEE
  • Conference_Location
    Queenstown
  • Print_ISBN
    978-1-4799-2180-5
  • Type

    conf

  • DOI
    10.1109/SAS.2014.6798957
  • Filename
    6798957