An artificial neural network based snore detector

A method to detect snores is presented that is based on the use of an artificial neural network (ANN). A network with 15 inputs, one output and two hidden layers with two adaline nodes per layer was used. The detector was tested using prerecorded data obtained from 13 male and 2 female patients (age 51.2/spl plusmn/6.5, weight 252/spl plusmn/31.15 lbs and height 5'9.4"/spl plusmn/2.49"). An overall accuracy of 75.2% in correctly detected snores showed an improvement of 12% over a previous method. The average number of false detections per hundred breaths was a 35% improvement.