EURASIP Journal on Audio, Speech, and Music Processing

Table 4 Word accuracy rates of Aurora 2.0 clean-train tasks for the 0-20 dB SNR test data, subtracting 1 (MVN+EMD1) or 2 (MVN+EMD2) IMFs

From: Noise-robust speech feature processing with empirical mode decomposition

	MVN+EMD1	MVN+EMD2	Diff.
Clean	98.4	98.2	+0.2
20 dB	96.3	96.0	+0.3
15 dB	93.3	92.9	+0.4
10 dB	85.8	85.0	+0.8
5 dB	68.6	68.6	+0.0
0 dB	39.6	41.9	-2.3
-5 dB	16.6	18.8	-2.2

Back to article page