EURASIP Journal on Audio, Speech, and Music Processing

Table 1 Total misclassification error difference between adopting the fixed combination (10, 30) and utilizing the best ( M , R ) combination

From: Efficient voice activity detection algorithm using long-term spectral flatness measure

Noise type	−10 dB	−5 dB	0 dB	5 dB	10 dB
White	2.85	0.69	0	0	0.87
Pink	4.49	0.64	0	0	1.19
Tank	2.32	0	1.59	2.20	2.17
Military vehicle	2.36	2.59	2.62	2.82	3.31
Jet cockpit	0	0	0	0	1.36
HF channel	1.15	0	0.57	1.02	2.27
F-16 cockpit	0	0	0.10	1.94	2.30
Factory floor	6.04	7.35	7.59	7.04	6.25
Car interior	2.33	2.82	3.16	2.33	2.21
Machine gun	12.45	11.33	9.83	8.74	7.56
Speech babble	2.11	0	2.068	0	0
Speech shaped	3.80	0	0	0	1.84

The numbers are the points that worse (higher error value) than utilizing the best (M, R) combination for each noise type and SNR level shown in the subfigures.

Back to article page