Table 9.

MLP-based DSD cross-validation performance vs. feature subsets–RLDD and RODeCAR.

Feature Subset	Subset Size	RLDD		RODeCAR
		UA [%]	WA [%]	UA [%]	WA [%]
all	2260	54.0	54.0	51.0	53.4
APF	26	53.0	53.0	52.4	53.4
FDF	1842	55.3	55.3	52.4	52.9
MFCC	78	52.0	52.0	52.5	52.9
MBF	312	58.6	58.6	53.8	54.8
UPF	2	49.8	49.7	51.4	59.6
APF + FDF	1868	55.5	55.5	52.7	52.8
APF + MFCC	104	52.5	52.5	53.6	54.2
APF + MBF	338	57.5	57.5	53.9	54.2
APF + UPF	28	51.8	51.8	52.6	53.0
FDF + MFCC	1920	54.9	54.9	52.6	53.2
FDF + MBF	2154	56.5	56.5	52.8	52.1
FDF + UPF	1844	55.1	55.2	51.9	52.5
MFCC + MBF	390	57.9	57.9	52.7	55.0
MFCC + UPF	80	53.4	53.4	52.0	53.6
MBF + UPF	314	56.8	56.8	53.4	55.5
APF + FDF + MFCC	1946	55.7	55.7	53.1	54.7
APF + FDF + MBF	2180	56.0	56.0	51.6	51.3
APF + FDF + UPF	1870	57.4	57.5	52.4	53.5
APF + MFCC + MBF	416	57.4	57.5	53.8	54.6
APF + MFCC + UPF	106	54.8	54.8	52.4	53.4
APF + MBF + UPF	340	58.1	58.1	54.9	55.0
FDF + MFCC + MBF	2232	57.2	57.2	51.6	52.4
FDF + MFCC + UPF	1922	54.9	55.0	52.4	54.0
FDF + MBF + UPF	2156	55.5	55.5	52.7	53.8
MFCC + MBF + UPF	392	55.4	55.4	53.1	55.1
APF + FDF + MFCC + MBF	2258	56.8	56.8	52.1	54.7
APF + FDF + MFCC + UPF	1948	55.7	55.7	52.0	54.2
APF + FDF + MBF + UPF	2182	55.5	55.5	52.1	53.6
APF + MFCC + MBF + UPF	418	56.0	56.0	53.0	55.6
FDF + MFCC + MBF + UPF	2234	56.2	56.2	52.3	51.7
top-10 selected by the KS-based algorithm	10	59.3	59.2	51.6	53.6
top-20 selected by the KS-based algorithm	20	58.1	58.2	52.3	56.5
top-50 selected by the KS-based algorithm	50	58.3	58.3	52.5	53.3
top-100 selected by the KS-based algorithm	100	58.0	58.0	52.0	53.4
top-200 selected by the KS-based algorithm	200	55.0	55.1	52.3	51.5