A Lung Sound Category Recognition Method Based on Wavelet Decomposition and BP Neural Network.

Area of Science:

• Medical Signal Processing
• Biomedical Engineering
• Acoustics

Background:

• Lung sound analysis is crucial for diagnosing respiratory conditions.
• Traditional methods face challenges with noise and high-dimensional data.
• Efficient characteristic extraction and recognition are needed for accurate diagnosis.

Purpose of the Study:

• To develop an effective method for characteristic extraction and recognition of lung sounds.
• To address the issue of high-dimensional feature vectors in lung sound analysis.
• To improve the accuracy and efficiency of lung sound classification.

Main Methods:

• Wavelet denoising was employed to reduce noise in collected lung sounds.
• Wavelet decomposition extracted characteristic coefficients, which were then transformed into signal energy.
• Linear Discriminant Analysis (LDA) was used for dimensionality reduction.
• A Backpropagation (BP) neural network was utilized for lung sound recognition.

Main Results:

• The proposed method achieved a lung sound recognition accuracy of 82.5% with high-dimensional vectors.
• Dimensionality reduction using LDA and signal energy transformation improved accuracy to 92.5%.
• The combination of wavelet transform, signal energy, LDA, and BP neural networks proved effective.

Conclusions:

• The developed method offers an efficient and accurate approach for lung sound recognition.
• Signal energy transformation and LDA are valuable techniques for handling high-dimensional lung sound features.
• This technique has potential applications in non-invasive respiratory disease diagnosis.