Page 192 - 应用声学2019年第4期

652 2019 年 7 月

[13] He Y J, Sun G L, Han J Q. Spectrum enhancement with negative matrix factorization with Gaussian mixtures and
sparse coding for robust speech recognition[J]. Digital Sig- masking model for speech enhancement[J]. Speech Com-
nal Processing, 2015, 43: 59–70. munication, 2017, 87: 18–30.
[14] Zhang L, Bao G Z, Zhang J, et al. Supervised single- [20] Jeon K M, Kim H K. Local sparsity based online dictio-
channel speech enhancement using ratio mask with joint nary learning for environment-adaptive speech enhance-
dictionary learning[J]. Speech Communication, 2016, 82: ment with nonnegative matrix factorization[C]. INTER-
38–52. SPEECH, 2016.
[15] Lee D D, Seung H S. Algorithms for non-negative matrix [21] Kwon K, Shin J W, Kim N S. NMF-based speech en-
factorization[J]. Advances in Neural Information Process- hancement using bases update[J]. IEEE Signal Processing
ing Systems, 2001: 556–562. Letters, 2015, 22(4): 450–454.
[16] Févotte C, Idier J. Algorithms for nonnegative matrix fac- [22] Rangachari S, Loizou P C. A noise-estimation algorithm
torization with the β-divergence[J]. Neural Computation, for highly non-stationary environments[J]. Speech Com-
2011, 23(9): 2421–2456. munication, 2006, 48(2): 220–231.
[17] Grais E M, Emad H E. Hidden Markov models as priors [23] Aharon M, Elad M, Bruckstein A, et al. K-SVD: an algo-
for regularized nonnegative matrix factorization in single- rithm for designing of overcomplete dictionaries for sparse
channel source separation[C]. INTERSPEECH, 2012. representation[J]. IEEE Transactions on Signal Process-
[18] Wilson K W, Raj B, Smaragdis P. Regularized non- ing, 2006, 54(11): 4311–4322.
negative matrix factorization with temporal dependencies [24] Lee D D, Seung H S. Learning the parts of objects by
for speech denoising[C]. INTERSPEECH, 2008. nonnegative matrix factorization[J]. Nature, 1999, 401:
[19] Chung H, Plourde E, Champagne B. Regularized non- 788–791.