Underdetermined Separation of Speech Mixture Based on Sparse Bayesian Learning

Figure 3

The sparse coefficients of original sources solved from dictionary trained with K-SVD.

(a) Original source 1

(b) Original source 2

(c) Original source 3