Télécom Paris

Single Channel Blind Dereverberation Based on Auto-Correlation Functions of Frame-wise Time Sequences of Frequency Components
  • Kenko Ohta, (Doshisha University)
  • Masuzo Yanagida, (Doshisha University)
  • Adaptive filtering algorithms and structures for echo and noise control
  • Noise reduction techniques
  • Sound enhancement and sound separation
Get the paper in PDF format
Acrobat Reader (version 5 minimum) is necessary to read this document.

This paper proposes a new blind dereverberation method based on auto-correlation function of frame-wise time series of each frequency component. This method can handle frequency characteristics of sound paths. To realize dereverberation, it is necessary to estimate the delay time and frequency characteristics of reflection rate though most conventional methods assume flat frequency characteristics. The proposed method estimates these parameters based on auto-correlation function of frame-wise time sequence of each frequency component or a sequence of a spectral component in frequency spectra obtained every frame-shift interval. Comparing with methods that assume flat frequency characteristics, the proposed method improves dereverberation performance. The proposed method raises average segmental SNR 3.4dB (from -2.4dB to 0.9dB) and reduces reverberation time from 390ms to 109ms.

©2006 Télécom Paris/TSI
Edition : Télécom Paris -- 2006