Automatic segmentation of infant cry signals using hidden Markov models

EURASIP Journal on Audio, Speech, and Music Processing

Table 5 Comparison of the performance of the original segmentation system with the adapted system on Cape Town database

	Accuracy	F score (%)
	(%)	Insp	Exp
Original system	58.3 ±1.4	20.7 ±0.7	65.4 ± 1.6
With feature	80.0 ±1.4	36.5 ±2.0	71.2 ±1.6
Norm. alone
With feature norm.	80.7 ±1.4	38.5 ± 1.8	73.2 ±1.7
and SSL
System trained with Cape Town data	85.2 ±0.7	39.2 ±2.3	78.0 ± 1.1

The expiratory and inspiratory phase classes are denoted by Exp and Ins, respectively