Comparing Two Methods Based on Time-Frequency Analysis to Estimate the Instantaneous Mixing Matrix in Blind Audio Source Separation
This paper explores the feasibility of using two methods, based on time-frequency analysis, to estimate the instantaneous mixing matrix in Blind Audio Source Separation applications. These methods are: 1) the LOST line orientation separation technique algorithm; and 2) a method based on histograms. They exploit sparsity and independency of sources, and what is of key importance, work properly for under-determined cases. The comparative study shows that only when a) the number of mixtures is small, and b) the number of mixture samples is low, the histogram-based approach works better than the LOST algorithm. In the remaining problems, the LOST algorithm performs better than the histogram-based one, and is especially useful when the system has sucient number of samples to obtain an estimate of the mixing matrix.
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 temporarily free for AES members.