A wireless sensor system for quantification of infant feeding behavior

Abstract

Research shows that rapid weight gain in infancy is associated to the development of obesity at a later stage in life. Feeding behavior in infants contributes to the rapid weight in early life. Sucking counts can be used to quantify the feeding behavior in infants. This paper presents a new signal processing algorithm to estimate sucking counts in infants from the data collected by a wireless jaw motion sensor. Meals for both breast-fed and bottle-fed infants were videotaped and synchronized with the sensor signal. Sensor signals were normalized and divided into 10 second segments. A percentile-based peak detection algorithm was used to estimate sucking count for each segment. The proposed approach was able to achieve a mean absolute error rate of 7.11% compared to human annotated sucking count with an average intra-class correlation of 0.92 between the algorithm and human raters.