Related Experiment Videos
Robust time delay estimation of bioelectric signals using least absolute deviation neural network.
Zhishun Wang1, Zhenya He, Jiande D Z Chen
1Department of Child Psychiatry and Brain Imaging, Columbia University and NYSPI, New York, NY 10032, USA. zw2105@columbia.edu
IEEE Transactions on Bio-Medical Engineering
|March 12, 2005
Summary
This study introduces a novel time delay estimation (TDE) algorithm using a least absolute deviation neural network (LADNN) for robust biomedical signal analysis. The LADNN-based TDE method outperforms existing techniques in complex and noisy environments.
Area of Science:
- Signal Processing
- Biomedical Engineering
- Machine Learning
Background:
- Time delay estimation (TDE) is crucial for biomedical signal analysis, but challenging due to signal non-stationarity, instability, chaos, and noise contamination.
- Existing TDE algorithms often rely on assumptions (e.g., non-Gaussian signals, Gaussian noise) that limit their applicability in real-world biomedical scenarios.
Purpose of the Study:
- To present a novel, robust time delay estimation (TDE) algorithm for biomedical signals using a least absolute deviation neural network (LADNN).
- To evaluate the performance of the LADNN-based TDE method against traditional algorithms, particularly higher-order spectra (HOS)-based methods, under various noise conditions and in a real biomedical application.
Main Methods:
- A new TDE algorithm is proposed, utilizing the least absolute deviation neural network (LADNN), which is a neural implementation of the L1-norm optimization model.
- The algorithm models signals using a moving average (MA) model, estimating MA parameters via LADNN. The time delay is identified at the peak of the MA coefficients.
- The LADNN-based TDE method is compared with existing wavelet-domain correlation and higher-order spectra (HOS)-based TDE algorithms through simulations and a real-world application.
Main Results:
- The LADNN-based TDE method demonstrates superior robustness compared to wavelet-domain correlation and HOS-based TDE algorithms, especially in non-Gaussian noise and chaotic environments.
- The proposed method is free from assumptions of non-Gaussian signals and Gaussian noise, making it more applicable to real-world biomedical data.
- Experiments show the LADNN-based TDE effectively extracts time delay information from gastric myoelectrical activity (GMA), assessing spatial propagation characteristics during migrating myoelectrical complex (MMC) phases.
Conclusions:
- The LADNN-based TDE algorithm offers a more robust and widely applicable solution for time delay estimation in complex and noisy biomedical signals.
- This novel approach overcomes limitations of conventional methods, paving the way for improved analysis of spatial propagation in biomedical data.
- The successful application to GMA analysis highlights the potential of LADNN for advancing biomedical signal processing and understanding physiological processes.