Deep CNN Architecture for Respiratory Sound Classification

Deep learning is a special method of machine learning that takes place in continuous layers of neural networks to retrieve data in a repetitive manner. An in-depth study is very useful when you are trying to find patterns from unstructured data. Deep learning complex neural networks are designed to mimic how the human brain works, so computers are often trained to solve undefined abstractions and problems. Recent improvements in AI, Big Data, and machine learning have increased the importance of image processing and biomedical signalling research. Biomedical signal processing requires periodic quantitative analysis and recording on a patient's chart to produce useful information that can be determined by clinics. The aim of this paper is to develop an in-depth study-based classification model for the identification of respiratory sounds for the diagnosis of orientation of lung and pulmonary diseases. In this project, we introduce a deep learning model CNN. This model use to classify respiratory sound. The model classifies the mel-spectrogram of respiratory sound. Here we propose 4 classifier models for classifying respiratory sounds (normal, wheeze, crackle, combination of wheeze and crackle). The main contribution of the paper is as follows: First, the proposed model is ready to identify a state-of-the-art score in the ICBHI’17 dataset. Second, compare the performance of the generalized models. Finally, the trained weight was calculated for memory optimization.