Robust sound event classification with bilinear multi-column ELM-AE and two-stage ensemble learning

EURASIP Journal on Audio, Speech, and Music Processing

Table 4 Multi-condition (MC) classification results of different algorithms with the mean probability value ensemble learning (-b) at the second-stage TsEL (unit: %)

	Clean	20 dB	10 dB	0 dB	Mean
ELM	95.20 ± 0.59	94.06 ± 0.38	93.29 ± 0.69	89.42 ± 0.93	92.99 ± 0.65
ELM-AE	96.65 ± 0.85	96.03 ± 0.85	95.88 ± 0.78	94.35 ± 0.83	95.73 ± 0.83
B-ELM-AE	98.15 ± 0.24	98.03 ± 0.41	97.48 ± 0.31	95.95 ± 0.35	97.40 ± 0.33
MC-ELM-AE-C	97.20 ± 0.49	96.80 ± 0.72	96.43 ± 0.40	95.10 ± 0.72	96.38 ± 0.58
MC-ELM-AE-TsEL-V	96.93 ± 0.50	96.50 ± 0.64	96.15 ± 0.44	95.10 ± 0.89	96.17 ± 0.62
MC-ELM-AE-TsEL-W	97.95 ± 0.53	97.43 ± 0.47	97.55 ± 0.52	96.05 ± 0.77	97.25 ± 0.57
B-MC-ELM-AE-C	98.08 ± 0.54	97.65 ± 0.45	97.40 ± 0.47	96.08 ± 0.50	97.30 ± 0.49
B-MC-ELM-AE-TsEL-V	98.20 ± 0.35	97.9 ± 0.40	97.55 ± 0.45	96.03 ± 0.59	97.42 ± 0.45
B-MC-ELM-AE-TsEL-W	98.65 ± 0.39	98.23 ± 0.45	98.18 ± 0.19	96.90 ± 0.59	97.99 ± 0.41