Article Preview
Top1. Introduction
The just-born babies or infants are not able to orally communicate with parents since they are not able to speak. They use to cry as their communication medium to express their physical and emotional needs. When they required more attention or care they cry. Thus, the parents are not able to understand the crying of the baby every time completely. The only solution to this problem is to study the acoustic speech pattern of the infant cry and determine the reason behind the cry. The acoustic speech pattern produced by infant cry is always different for different reasons. Thus, the different situations in the baby of hungry or sleepy or in pain generate different acoustic speech patterns. The infant crying pattern can be used as a biological alarm system to alert the parents. To differentiate the acoustic speech patterns using signal processing algorithms. Various types of features can be extracted to identify and analyze the cry signal patterns (R.P. Balandong, 2013; S. Bhattacharya, 2016). At the earliest stage the reason for infant cry helps the parents to take appropriate steps and for pediatricians guides them properly in treatment. The acoustic characteristics of the cry patterns are directly influenced by the infant’s physical and psychological state (D. Lederman, 2002). The acoustic signal of infant cry contains valuable information such as gender, health, identity, and emotions (R. Cohen, 2012). By using these properties (features) and analyzing them, we can detect the infant’s reason behind crying. The basic objective is to find unique features. As the infants are not able to speak their cry defines it all. We have to analyze their cry signal patterns to find similar and discriminating features. It will be a great help for pediatricians and parents if they know the reason behind cry. Proper treatment can be given to that baby.
There are many attempts made to detect the reason behind infant cry. Orozco in (J. Orozco, 2003) has used Neural networks primarily for this purpose. As a large amount of database is used, neural networks work best. They achieved good accuracy results (J. Orozco, 2003). Another research work related to Baby cry detection did the comparison between classical and new methods of acoustic analysis of Infant cry in (G.J. Varallyay, 2004). They used fundamental frequency detection and dominant frequency detection. In the cry of infants with normal hearing and hard of hearing, the ratio seems to be different between fundamental frequency and dominant frequency (G.J. Varallyay, 2004). When it comes to the processing of cry signal the main focus is in the extraction and analysis of the fundamental frequency (F0) and the first three formats F1, F2, and F3 of infant cry signal as implemented in [P. Pal, 2006]. These parameters contain important information regarding the emotional state of the infant. The Harmonic Product Spectrum (HPS) method has been used to obtain the values of the fundamental frequency (F0) as for infant cries, the fundamental frequency varies widely and rapidly (P. Pal, 2006). The detection is done in five categories such as pain, hunger, fear, sadness, anger.