Sleep stage classification using neuro-fuzzy intelligence
Neurofuzzy systems find their applications in many areas, medical diagnosis being one of many areas. About 50 million adults, i.e. about one third of all adults in America, have complaints about their sleep. Finding a way to study sleep data and getting valuable information from the data is a challenging task. Classifying this data can help medical professionals to diagnose sleep related disorders efficiently and accurately. The main objective of this thesis was to classify the sleep stages using Neuro-Fuzzy intelligence. The research mainly consists of two parts. The first part was using a neural network to learn and classify the sleep stage patterns accurately. The sleep feature data were pre-processed and fed to a Multilayer Preceptron, with Backpropagation algorithm. Once the patterns were classified, they were tested with a set of data and the corresponding accuracies were recorded.
The second part of my research included extracting concise fuzzy rules from the fuzzy network. The weights corresponding to the input features were recorded. From these recorded weights, rules were extracted. Some of these rules were eliminated using the pruning algorithm. The rules were then used to create a Fuzzy Inference System for classifying sleep stages.