Title :
Estimation of babies´ emotion by frequency analyses of their cries
Author :
Kikuchi, Kazuro ; Arakawa, Kazuki
Author_Institution :
Meiji Univ., Japan
Abstract :
Summary form only given. A method for estimating how a baby is feeling when he cries is proposed using frequency analyses of the voice. In particular, two causes, hunger and sleepiness, are considered. It is clarified that the frequency spectrum of babies´ cries keeps large values in the high frequency region when the baby is hungry, while the values become small in the low frequency region when the baby is sleepy. Using these tendencies, a rule-based system is proposed, which can classify cries of babies from 3 to 12 months old in a hungry state and in a sleepy one. Since the tendency of the frequency spectrum changes depending on the month age, rule based systems are separately proposed for younger and older babies. Computer simulations show that about 80% of the babies´ cries are correctly classified.
Keywords :
audio signal processing; emotion recognition; knowledge based systems; parameter estimation; signal classification; spectral analysis; speech processing; baby cry classification; baby emotion estimation; frequency analyses; frequency spectrum analysis; hunger; rule-based system; sleepiness; Computer simulation; Face; Filter bank; Frequency estimation; Image processing; Knowledge based systems; Nonlinear filters; Pediatrics; Skin; Speech analysis;
Conference_Titel :
Nonlinear Signal and Image Processing, 2005. NSIP 2005. Abstracts. IEEE-Eurasip
Conference_Location :
Sapporo
Print_ISBN :
0-7803-9064-4
DOI :
10.1109/NSIP.2005.1502217