This paper proposes a noise reduction technique that applies a priori information to unmixing matrix estimation in ICA; it offers fast and accurate convergence. We formulate the parameter estimation stabilized by the a priori information as a Bayesian framework?@of maximum a posteriori (MAP) estimation, and show its robustness in mobile phone environments, where the position of the microphone relative to the mouth is almost constant. We use the transfer function of mouth to microphone for one row of the unmixing matrix. Using these estimated parameters as initial values, the unmixing matrix can be updated with high efficiency in the framework of MAP estimation. Experimental results confirm that the proposed method achieves high performance, especially in high SNR noise conditions.
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.