Enhanced Classification of Snoring Sounds Using Stacked Classifier Models of Machine Learning with SVM-KNN and Deep Learning with RNN-LSTM

Sleep disorders caused by snoring are a common problem that negatively affect the individual’s daily quality of life. For instance, poor sleep caused by snoring will induce important physical and mental issues. Given the fact that finding common criteria for all snoring sounds is very tough, this study aims to propose two models for snoring classification using AI-based learning methods. The first model is a Machine Learning (ML)-based built by stacking two classifiers namely, the SVM and KNN, that will learn the features extracted by applying the MFCC as a feature extraction technique. The second model is a Deep Learning (DL)-based where RNN and LSTM classifiers are stacked and where three feature extraction techniques (i.e. MFCC, STFT, ZCR) are applied. An online dataset consisting of .wav audio signals is used to implement the two models. Results show that the first and second models have achieved high accuracy scores of 98.5% and 80.8% respectively.