In this letter, we propose an online estimated local dictionary based single-channel speech enhancement algorithm, which focuses on low-rank and sparse matrix decomposition. In the proposed algorithm, a noisy speech spectrogram can be decomposed into low-rank background noise components and an activation of the online speech dictionary, on which both low-rank and sparsity constraints are imposed. This decomposition takes the advantage of local estimated exemplar’s high expressiveness on speech components and also accommodates nonstationary background noise. The local dictionary can be obtained through estimating the speech presence probability (SPP) by applying expectation–maximal algorithm, in which a generalized Gamma prior for speech magnitude spectrum is used. The proposed algorithm is evaluated using signal-to-distortion ratio, and perceptual evaluation of speech quality. The results show that the proposed algorithm achieves significant improvements at various SNRs when compared to four other speech enhancement algorithms, including improved Karhunen–Loeve transform approach, SPP-based MMSE, nonnegative matrix factorization-based robust principal component analysis (RPCA), and RPCA.
Sun, Pengfei and Qin, Jun. "Low Rank and Sparsity Analysis Applied to Speech Enhancement via Online Estimated Dictionary." IEEE Signal Processing Letters 23, No. 12 (Fall 2016): 1862-1866. doi:10.1109/LSP.2016.2627029.