Type-2 fuzzy pattern matching for classifying hunger and pain cries of healthy full-term infants

Abstract

Crying is the first way through which infants communicate with others. Various cries of infants have different meanings and origins, such as hunger, pain, etc. Therefore, the analysis of infant cries could help adults to earlier understand its needs, and diagnose its diseases. For this purpose, this paper uses the type-2 fuzzy pattern matching as a method for classifying hunger and pain cries respectively recorded from healthy full-term infants in Imam Khomeini hospital and Shahid Rajaee clinic, which are located in Noor city, Mazandaran Province, Iran. The Features fed into classifier are Mel Frequency Cepstral Coefficients (MFCCs) extracted from the database. Results on one-second segments of cry signals show that type-2 fuzzy classifier has higher accuracy in comparison with Support Vector Machines (SVM) and Logistic Regression (LR) classifiers, while results on cry signals show 100% accuracy in all three classifiers.