The single-channel blind dereverberation approach we present in this paper is an extension to the one based on Constant Modulus Algorithm (CMA) we proposed in previous work. By substituting the Modified CMA algorithm for the original CMA algorithm, we demonstrate a more suitable approach for blind deconvolution of reverberant audio signals with super-Gaussian distribution. To further improve the performance, the Modified CMA is applied to the LP residual instead of the time domain signal because of the flatter spectrum provided by the Linear Prediction (LP) residual approach. In real implementations, a Delayless Subband Adaptive Filtering (DSAF) architecture is also combined with CMA to further reduce the computational complexity. Experimental results show that our modified method outperforms previous approaches in audio signal blind dereverberation.
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.