Optimized CNN Model with Derived Kernels for Apnea Classification Application

The sleep related disorders are common in both men and women, irrespective of their age. In recent years, there is an abrupt increase of interest in the prediction of sleep related disorders among the researchers. The most common data analysis approach utilize a large data set to predict the sleep apnea episode. The computational complexity of such predictions is high and most of the techniques are failing to select the optimal features. The recent trends show that the convolution neural network gaining the popularity because of its ability to select the optimal features and the dimension reduction property. On the other way, it is also important to address the noise issues in the acquired physiological data. The Variational Mode Decomposition (VMD) is an adaptive way of decomposing the different frequency levels, such that the noise level can be easily separated into the corresponding levels. This paper presents a novel technique for the optimal feature extraction from the Variational Mode Function (VMF) levels by using the concept of the Convolution Neural Network (CNN).