SCIENCE@home

Lung sound signal classification by using Cosine Similarity-based Multilevel Discrete Wavelet Transform Decomposition with CNN-LSTM Hybrid model

Khabat Hasan Abdullah,Mehmet Bilal Er

Published 2022 in International Joint Conference on the Analysis of Images, Social Networks and Texts

ABSTRACT

Lung sound signals can be used to diagnose the initial stages of respiratory ailments. Artificial intelligence techniques might be used to automatically identify respiratory illnesses including asthma, pneumonia, and bronchiolitis, or normal and abnormal utilizing this digital data. This study uses a deep learning technique for the classification process. A novel Convolutional Neural Network (CNN) structure with long short-term memory (LSTM) is proposed for extracting deep features and classifying signals into the main categories of normal and abnormal. The proposed CNN consists of (No. of Layers) including convolutional layers, Batch normalization, Activation Function, Pooling, Fully-connected, drop-out and also at the end non-linear function is used for classification such as (SoftMax). And the proposed model used for classification is called the CNN-LSTM hybrid model through coming feeding vectors to it and obtaining a good output of classification. During the preprocessing steps of data, the de-noising method is used called the Butterworth filter (Bandpass filter) eliminates noise from the lung sound. And 3-level DWT decomposition technique is used called CS-MDWTD for extraction and selection features. Non-linear adaptive filter based on (DWT and ANN). In the beginning, the data is preprocessed to convert the lung signal sounds into images. The suggested method was tested on the Respiratory Sound Database (ICBHI 2017). The proposed CS-MDWTD technique is assessed by using the CNN-LSTM hybrid model on the ICBHI dataset. The experimental results showed performing classification and diagnosis under the proposed method, and the results accuracy rates of the lung sound classification reached over 90% for most of the cases classification’s class improved approximately by (5%) in probability, against the other studies/benchmarks of the lung signal sound classification techniques.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-19 of 19 references · Page 1 of 1

CITED BY

Showing 1-8 of 8 citing papers · Page 1 of 1