EURASIP Journal on Audio, Speech, and Music Processing

Table 2 Eightfold cross-validation result of the fingering detection system

From: Detecting fingering of overblown flute sound using sparse feature learning

		SVM		Random forest
Note	Feature (dim.)	P _f ± σ ² _Pf	P _f ± σ² _Pf (Mp)	P _f ± σ² _Pf	P _f ± σ ² _Pf (Mp)
C5	MFCCs (39)	9.87 ± 1.57	8.07 ± 1.00	14.91 ± 1.91	12.20 ± 1.05
	SF (39)	0.55 ± 0.47	0.20 ± 0.27	0.68 ± 0.52	0.13 ± 0.24
	SF (1024)	0.26 ± 0.24	0.00 ± 0.00	0.36 ± 0.24	0.07 ± 0.19
D5	MFCCs (39)	7.43 ± 1.30	8.21 ± 0.82	10.16 ± 1.59	10.20 ± 2.42
	SF (39)	2.20 ± 0.78	1.41 ± 1.12	1.48 ± 0.54	0.40 ± 0.51
	SF (1024)	1.67 ± 0.72	1.18 ± 0.68	1.00 ± 0.36	0.39 ± 0.38
C6	MFCCs (39)	8.47 ± 0.86	6.91 ± 1.48	11.78 ± 1.59	8.91 ± 0.82
	SF (39)	2.08 ± 0.88	1.50 ± 0.80	1.30 ± 0.50	0.67 ± 0.40
	SF (1024)	0.76 ± 0.42	0.13 ± 0.17	0.56 ± 0.34	0.04 ± 0.12
C#6	MFCCs (39)	7.01 ± 0.62	5.47 ± 0.90	9.31 ± 1.01	6.02 ± 0.98
	SF (39)	1.39 ± 0.72	1.07 ± 0.82	1.26 ± 0.52	0.56 ± 0.33
	SF (1024)	0.31 ± 0.29	0.08 ± 0.15	0.29 ± 0.27	0.04 ± 0.11
E6	MFCCs (39)	6.36 ± 0.78	4.56 ± 0.90	7.43 ± 0.96	4.80 ± 1.27
	SF (39)	2.35 ± 0.79	1.74 ± 0.76	2.03 ± 0.46	1.31 ± 0.51
	SF (1024)	1.06 ± 0.48	1.11 ± 0.30	1.02 ± 0.37	0.34 ± 0.24
G6	MFCCs (39)	13.80 ± 1.14	12.77 ± 1.28	13.76 ± 1.12	10.63 ± 0.96
	SF (39)	5.71 ± 1.30	4.49 ± 0.76	4.62 ± 1.26	3.22 ± 0.58
	SF (1024)	1.53 ± 0.53	1.06 ± 0.54	1.55 ± 0.36	1.06 ± 0.44

Mean and standard deviations of detection error (P _f ± σ ² _Pf) probabilities (%) using MFCCs and sparse filtering (SF) for 39 and 1024 units. SVM and random forest were used for the classifications, and probabilities are given for features without and with max-pooling (Mp)

Back to article page