Skip to main content

Table 2 Average PESQ scores (%) comparisons of different clean speech states and basis vectors (\(\ddot{J} = 1,{\ddot{K}}=70\))

From: A speech enhancement algorithm based on a non-negative hidden Markov model and Kullback-Leibler divergence

Parameters

\({\overline{K}}=5\)

\({\overline{K}}=10\)

\({\overline{K}}=25\)

\({\overline{K}}=50\)

Noisy

2.02 (\(\pm \,{0.03 }\))

NMF-HMM, \({\overline{J}}=1\)(T-NMF)

2.12 (\(\pm \,{0.03}\))

2.18 (\(\pm \, 0.03)\)

2.21 (\(\pm \, 0.02\))

2.18 (\(\pm \, 0.02)\)

NMF-HMM, \({\overline{J}}=5\)

2.27 (\(\pm \,{0.03}\))

2.31 (\(\pm \,0.03 )\)

2.32 (\(\pm \,0.02\))

2.29 (\(\pm \, 0.02)\)

NMF-HMM, \({\overline{J}}=10\)

2.31 (\(\pm \,{0.03 }\))

2.35 (\(\pm \,0.03 )\)

2.35 (\(\pm \, 0.03\))

2.30 (\(\pm \, 0.02)\)

NMF-HMM, \({\overline{J}}=20\)

2.36 (\(\pm \,{0.03 }\))

2.39 (\(\pm \,0.02 )\)

2.36 (\(\pm \, 0.02\))

2.32 (\(\pm \, 0.02)\)

NMF-HMM, \({\overline{J}}=40\)

2.38 (\(\pm \,{0.02 }\))

2.41 (\(\pm \, 0.02)\)

2.39 (\(\pm \,0.02\))

2.33 (\(\pm \, 0.02)\)