Skip to main content

Table 4 Average PESE scores (%) comparisons of different noise states and basis vectors (\(\overline{J} = 40 ,{\overline{K}}= 25\))

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

Parameters

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

\({\ddot{K}}=20\)

\({\ddot{K}}=40\)

\({\ddot{K}}=70\)

Noisy

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

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

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

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

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

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

NMF-HMM, \({\ddot{J}}=2\)

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

2.33 (\(\pm \, 0.04)\)

2.37 (\(\pm \, 0.04\))

2.40 (\(\pm \, 0.03)\)

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

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

2.34 (\(\pm \,0.04 )\)

2.39 (\(\pm \, 0.03\))

2.40 (\(\pm \, 0.03)\)

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

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

2.36 (\(\pm \, 0.03)\)

2.40 (\(\pm \, 0.02\))

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