Table 7 Performance of models using standardized view.

From: A benchmark for domain adaptation and generalization in smartphone-based human activity recognition

Model

Time

Frequency

KH

MS

RW-T

RW-W

UCI

WDM

Mean

KH

MS

RW-T

RW-W

UCI

WDM

Mean

KNN

49.3%

66.6%

43.6%

53.4%

66.7%

60.7%

56.7%

86.8%

90.9%

65.2%

74.5%

81.7%

89.8%

81.5%

Random Forest

80.7%

89.1%

62.5%

67.2%

88.1%

85.5%

78.8%

82.6%

92.9%

80.5%

74.7%

92.8%

89.5%

85.5%

SVM

61.1%

76.7%

64.7%

65.2%

78.6%

74.6%

70.2%

71.5%

81.9%

74.1%

73.6%

79.7%

78.7%

76.6%

CNN (1D)12

78.1%

92.2%

69.8%

73.4%

94.9%

90.3%

83.1%

73.5%

91.1%

74.7%

81.6%

94.0%

90.7%

84.3%

CNN (2D)12

80.7%

94.2%

74.0%

75.8%

93.7%

87.6%

84.3%

77.2%

91.9%

74.6%

80.9%

91.0%

89.9%

84.3%

CNN PF34

79.2%

94.9%

69.9%

79.3%

95.2%

85.7%

84.0%

80.6%

90.7%

65.6%

83.8%

95.3%

91.0%

84.5%

CNN PFF34

80.0%

94.0%

67.4%

80.6%

96.9%

87.5%

84.4%

78.2%

91.7%

64.3%

83.0%

95.7%

90.2%

83.8%

ConvNet13

78.5%

95.8%

63.6%

77.5%

96.9%

87.3%

83.3%

81.2%

92.4%

81.5%

84.5%

94.4%

91.3%

87.6%

IMU CNN14

78.2%

87.7%

59.9%

69.7%

91.9%

83.6%

78.5%

80.7%

93.6%

65.3%

81.1%

95.5%

91.7%

84.6%

IMU Transf.14

73.5%

64.3%

63.0%

73.1%

62.8%

45.9%

63.8%

70.8%

77.6%

63.5%

77.4%

78.7%

57.9%

71.0%

MLP (2 Layers)

75.8%

84.3%

57.3%

62.5%

79.7%

81.7%

73.5%

86.9%

91.2%

74.3%

80.9%

92.6%

90.3%

86.0%

MLP (3 layers)

79.4%

84.1%

57.6%

64.1%

81.8%

81.4%

74.7%

86.0%

90.9%

74.8%

79.7%

93.5%

90.5%

85.9%

ResNet15

81.4%

79.5%

67.6%

74.6%

91.0%

79.6%

78.9%

71.9%

85.9%

67.2%

80.6%

90.7%

85.0%

80.2%

ResNetSE67

80.8%

83.0%

69.6%

74.7%

90.7%

76.9%

79.3%

67.4%

84.6%

70.0%

74.5%

84.6%

85.0%

77.7%

ResNetSE-567

82.6%

84.9%

74.1%

69.1%

92.1%

82.0%

80.8%

71.3%

88.5%

68.6%

78.6%

91.3%

81.0%

79.9%

Max

82.6%

95.8%

74.1%

80.6%

96.9%

90.3%

84.4%

86.9%

93.6%

81.5%

84.5%

95.7%

91.7%

87.6%

  1. The best results for each dataset and for each domain (time and frequency) are highlighted in bold. Mean column represents the average performance of the model in the datasets.