Skip to main content
EURASIP Journal on Audio, Speech, and Music Processing

Table 5 Error per class and total error for FA segmentation-by-classification systems

From: Audio segmentation-by-classification approach based on factor analysis in broadcast news domain

IoChD scoring: step - 0.5 s, 100 chnf

 

MU

SP

SM

SN

Total

One single U

40.3

76.9

60.5

64.3

60.5

One U per class

33.3

45.6

36.2

47.4

40.6

  1. The experiments are computed with one single U for all the classes and one U matrix for each class/non-class using IoChD scoring. No score combination or smoothing was carried out. The italicized numbers show the total error of each system.
Back to article page