Skip to main content

Table 2 Comparison of proposed scheme with existing work under noisy condition

From: Paralinguistic and spectral feature extraction for speech emotion classification using machine learning techniques

SNR

Method

ACC

TPR

TNR

PPV

NPV

5dB

Method of [39]

82.55%

91.68%

93.72%

82.60%

97.36%

 

SEC-SVM using paralinguistic feature

66.17%

97.87%

88.22%

72.44%

99.24%

 

SEC-SVM using spectral feature

57.38%

69.40%

100.00%

100.00%

76.27%

10dB

Method of [39]

84.52%

93.29%

94.52%

84.62%

97.89%

 

SEC-SVM using paralinguistic feature

67.48%

99.04%

91.49%

81.10%

99.37%

 

SEC-SVM using spectral feature

71.59%

82.47%

100.00%

100.00%

90.46%

15dB

Method of [39]

84.68%

92.58%

94.62%

84.94%

97.68%

 

SEC-SVM using paralinguistic feature

74.39%

94.40%

96.89%

92.91%

97.56%

 

SEC-SVM using spectral feature

83.55%

92.70%

99.00%

97.27%

96.97%

  1. The bolded data indicates the best results