Sleep classification in infants by a Learning Vector Quantizer (LVQ3)

The study reports on the possibility of classifying sleep stages in infants with a Learning Vector Quantizer. The polygraphic data recorded over 8 hours from 4 babies aged 6 months were subject to classification. From each baby 22 signals were recorded, digitized and stored on an optical disc. Subsets of these signals and additionally calculated parameters were used to form data vectors in intervals of 30 sec. The teaching input was formed by a human expert. For classification a Learning Vector Quantizer (LVQ3) was trained on the labelled data vectors from 3 babies after which it was tested on the remaining fourth baby. Correct classification rates varied between 40 and 74%.