Skip to main content

Table 1 Music genre classification accuracies for the GTZAN, ISMIR, Homburg, 1517-Artists, and Unique datasets

From: Music classification by low-rank semantic mappings

Method

Features

GTZAN

ISMIR

Homburg

1517-Artists

Unique

LRSMs

Fusion cmc

87.00 (2.62)

82.99

62.40 (3.65)

54.91 (2.54)

72.90 (1.26)

 

Fusion cm

86.80 (2.85)

82.30

62.29 (4.04)

54.74 (2.68)

72.84 (1.11)

 

Cortical

85.50 (2.79)

81.62

61.71 (4.02)

54.43 (2.58)

72.35 (1.05)

 

MFCCs

50.60 (5.35)

59.08

43.26 (2.30)

23.45 (1.96)

54.60 (1.87)

 

Chroma

17.6 (4.03)

43.90

26.93 (1.39)

9.77 (1.13)

24.71 (1.87)

SRC

Fusion cmc

84.40 (2.27)

82.85

59.64 (3.24)

53.08 (2.83)

72.61 (1.18)

 

Fusion cm

84.40 (2.71)

80.50

58.10 (4.15)

50.78 (2.41)

71.97 (1.85)

 

Cortical

84.10 (3.04)

79.97

57.52 (3.98)

50.72 (2.61)

67.48 (1.14)

 

MFCCs

63.60 (5.01)

70.50

38.10 (2.74)

30.12 (1.87)

56.59 (1.06)

 

Chroma

36.80 (5.67)

47.73

26.61 (2.58)

17.01 (1.31)

31.20 (2.94)

SVMs

Fusion cmc

86.80 (2.82)

82.99

62.61 (3.22)

53.30 (3.19)

75.15 (1.48)

 

Fusion cm

86.40 (2.98)

73.93

61.07 (3.32)

53.08 (3.38)

73.54 (1.87)

 

Cortical

86.00 (2.83)

73.79

60.92 (2.83)

53.71 (3.18)

68.89 (2.22)

 

MFCCs

54.90 (3.14)

52.67

43.95 (2.05)

26.16 (2.96)

53.22 (1.06)

 

Chroma

16.90 (4.02)

48.42

34.99 (1.96)

12.16 (2.27)

39.87 (2.67)

NN

Fusion cmc

81.40 (3.20)

78.64

50.26 (4.21)

44.87 (2.21)

64.68 (2.31)

 

Fusion cm

81.10 (3.31)

79.02

50.21 (3.48)

44.90 (2.43)

64.68 (2.31)

 

Cortical

80.70 (3.26)

79.69

49.78 (2.98)

44.84 (2.55)

64.43 (2.57)

 

MFCCs

57.60 (5.05)

67.76

29.79 (3.13)

26.57 (1.84)

48.82 (2.17)

 

Chroma

34.10 (4.67)

42.24

23.64 (1.93)

14.40 (1.80)

25.32 (2.96)

 

[20] 90.60

[20] 86.83

[16] 61.20

[16] 41.10

[16] 72.00

 

[22] 84.30

[10] 83.50

[60] 57.81

[61] 35.00

 
 

[56] 82.50

[22] 83.15

[61] 55.30

  
 

[16] 82.00

[62] 82.30

[54] 53.23

  
 

[57] 77.20

 
  1. The numbers within the parentheses indicate the standard deviations obtained by 10-fold cross-validation. The best results are indicated in italics.