We present a new method for dereverberating speech, based upon a multichannel Wiener Filter and a microphone array. We demonstrate the effectiveness of this method under real, reverberant conditions and show that the method may be described as a self-steering beamformer. Furthermore, we investigate the performance of the method under simulated conditions, designed to closely match the acoustic characteristics of the real room environment. These simulations yield significantly inferior results to those obtained using real recordings and we show that this is as a result of the failure of simulated impulse responses to accurately model real impulse responses, in certain critical respects.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.