Pulse Signal Analysis Based on Deep Learning Network

Area of Science:

• Biomedical Signal Processing
• Cardiovascular Physiology
• Artificial Intelligence in Medicine

Background:

• Pulse signals are crucial physiological indicators reflecting cardiovascular health.
• Accurate analysis of pulse signals is vital for noninvasive diagnosis and remote patient monitoring.
• Traditional pulse signal analysis faces challenges in feature extraction and noise reduction.

Purpose of the Study:

• To develop an automated system for pulse signal analysis and recognition.
• To improve the accuracy of pulse signal classification using advanced signal processing and deep learning techniques.
• To combine time and frequency domain feature extraction with convolutional neural networks (CNNs) for robust pulse signal analysis.

Main Methods:

• Utilized wavelet transform and ensemble empirical mode decomposition (EEMD) for effective noise reduction in pulse signals.
• Implemented a differential threshold method for accurate detection and extraction of time-domain pulse signal features.
• Applied a one-dimensional convolutional neural network (CNN) for the classification of multiple pulse signal types.

Main Results:

• Successfully removed noise from pulse signals using wavelet transform and EEMD.
• Achieved accurate positioning and extraction of time-domain features through the differential threshold method.
• Demonstrated the effectiveness of the 1D CNN model in classifying diverse pulse signals.

Conclusions:

• A novel deep learning framework combining time-frequency feature extraction and CNNs is proposed for pulse signal classification.
• The developed method offers a significant advancement in automated pulse signal analysis for clinical applications.
• This approach holds promise for enhancing noninvasive diagnostic capabilities and remote cardiovascular monitoring.