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

Table 6 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 - 0.1 s, 100 chnf

 

MU

SP

SM

SN

Total

One U per class

27.9

37.9

32.4

40.9

34.8

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