Analysis And Pattern Classification Of Infants' Cry

Cry is a main expression of infants to communicate with outside world, which implies lots of information about psychology and pathology. Previous research indicates that different babies under certain given state or requirement have some common characters in their cry. Therefore, identification of infants' cry modes has very important practical significance. Nowadays, more and more researchers abroad pay much attention to the study of the objective corresponding relationship between infants' cry acoustic features and their actual requirements. However, in China, the researches are still based on hearing apperception and subjective estimation.In this thesis, some interrelated researches have been explored as follows:1, Create the first domestic Database of Infants' Cry.2, Compare the vocal ism systems of infants with adults, analyze in detail their similarities and differences in phonation configuration and phonation principle. It shows the differences between infants' cry signals and voice signals. Based on the practical significance of this subject, there are three different kinds of classification modes of infants' cry.3, Choose MFCC and LPC, and attempt LPCC as identification characteristics, and compare the recognition effects of these three features in our experiment. It shows that LPC is the best feature in the classification; LPCC follows; and MFCC may be acceptable when we choose a certain order.4, Try to use DTW algorithm in the infants' cry classification system in consideration of the actual conditions. Compare it with the BP neural network in our experiment, which shows that DTW is better than BP neural network when data volume is small.5, Present nonlinear sampling method to resolve the problem of length difference of the feature vectors, which is caused by the difference of the cry's...