Emotion Detection from Infant Cries

Infant cry is a way in which baby tries to communicate specific messages. As crying is the only way to express discomfort, so this signal carries a lot of information about infants physical and psychological condition. Therefore, the aim to extract and classify the state of an infant based on the patterns exhibited by the crying sound signal. More specifically we propose a methodology able to distinguish among the following five states: (a) hungry, (b) uncomfortable, (c) need to burp, (d) in pain, and (e) need to sleep. There are many feature extraction techniques evolved such as Linear Predictive Codes, Perceptual Linear Prediction, Mel Frequency Cepstral Coefficients (MFCC), RASTA Filtering etc evolved over past few years. This paper is to study MFCC technique to detect the emotion of crying infant. Among all MFCC is the most prevalent and dominant method used to extract spectral feature. The recognition accuracy is high. That means the performance rate of MFCC is high. So here we compare the performance of various feature classification techniques that are used with MFCC and other extraction techniques.