The Feature Analysis And Recognition Of Infant Cry

The cry of infants is a form of communication,which is very limited,but similar to the way of communication of adults.It is also a biological alarm,which conveys the physical and psychological needs of infants to the outside world.Both the physical condition of infants can be sure and disease can be detected based on the information carried by the wave of infants'cry.Therefore, the effective identification of cry and"translating"cry into"adult language"successfully can make us understand the meaning of infant cry,which has great practical significance.As infant cry is associated with voice signal,nowadays more and more researchers have started to use the technology of voice processing and method of pattern recognition into the classification of infant cry.The rapid development of technology of voice processing also provides a strong support to the recognition of infant cry.In this paper,a certain degree of research work of infant cry recognition have been explored as follows:The infant cry and voice signal have been compared to find the difference through phonation structure and principles of phonation.Using the appropriate technology to analysis infant cry.The database is established. Since there is no unified standardized infant cry database and researchers from different areas also focus on different aspects.Therefore, the infant cry is categorized by a variety of types.Considering all factors, the infant cry is sorted by painful and unpainful in experiments. After preprocessing,the focus of infant cry is parameter extraction.Through analyzing the advantage and disadvantage of commonly used parameters, which include linear prediction coefficient and linear prediction cepstrum coefficient and mel frequency cepstrum coefficient etc,the choice of parameter for infant cry is mel frequency cepstrum coefficient,and select different order to observe effect of recognition rate on the painful and unpainful infant cry. Then taking the unique infant track structure and the high pitch frequency of infant cry into account, the information extraction of infant cry in low frequency is better than in high frequency by using mel frequency cepstrum coefficient,a inverted mel frequency cepstrum coefficient is mentioned. The new parameter has a great advantage on the high frequency part to obtain information,is a complementary merit to mel frequency cepstrum coefficient. Eventually mel frequency cepstrum coefficient and inverted mel frequency cepstrum coefficient will be used as a combination,that is the formation of combined parameters,also select different order to observe effect of recognition rate on the painful and unpainful infant cry. The experiment results show,compared with using a single parameter,the recognition rate of combination of the two parameters has been eventually increased.Three commonly used algorithms of recognition have been analyzed.But according to the state of a few samples and the feature of infant cry and the size of computation,the DTW recognition algorithm which is broad in the recognition of isolated word has been determined to use in the painful and unpainful infant cry.