Acoustic analysis of infant cry signal towards automatic detection of the cause of crying

Shivam Sharma, Pruthvi Raj Myakala, Rajasree Nalumachu, Suryakanth V. Gangashetty, V. K. Mittal

Abstract: Infant cry signal carries enough clues about the cry cause. Using these clues, a mother or a paediatrician can easily identify the probable cry-cause, and emotional or physiological state of the baby. In our earlier studies, differences between infant cries due to pain and discomfort were examined using features derived from the cry signal. In this paper, acoustic analysis of the infant cry signal is carried out for 4 cry-causes: discomfort, environmental factors, pain and strangers anxiety. Acoustic features are derived from the infant cry signal as expressive media. Features F 0 , signal energy and 5 formants are derived using autocorrelation, short-term Fourier transform and linear prediction analysis, respectively. An infant cry signal database (ICSD 2), having total 1531 cries, is analyzed. Analysis results show notable differences in their features and definite trends in their contours. Study lays groundwork for automatic detection of the cry-cause, towards assistive healthcare.