Enhanced DWT for Denoising Heartbeat Signal in Non-Invasive Detection

Peibin Zhu, Lei Feng, Kaimin Yu, Yuanfang Zhang, Wen Chen, Jianzhong HaoSensors 2025 · DOI

TL;DR

This paper introduces an enhanced Discrete Wavelet Transform (DWT) method for denoising heartbeat signals from non-invasive sensors. By incorporating an objective quality assessment metric and an adaptive threshold function, the method significantly improves signal-to-noise ratio (SNR) and reduces artifacts, leading to more accurate cardiovascular disease diagnosis.

What Problem Does This Paper Address?

Non-invasive methods for monitoring heart rate, like optical sensors or accelerometers, are crucial for early detection of cardiovascular diseases. However, these signals are often plagued by various types of noise and artifacts (e.g., additive white Gaussian noise, baseline wander, electrode motion, muscle artifacts). Existing denoising techniques, whether traditional signal processing or AI-based, face challenges: traditional methods can be limited by noise estimation, while AI methods require extensive data and computational power and can be less adaptive. Specifically, wavelet-based methods struggle with selecting optimal thresholds and threshold functions, which can lead to residual noise or signal distortion. This paper aims to tackle these limitations by proposing a more robust and accurate wavelet denoising approach.

Methods

The core of this work is an "enhanced DWT" that improves upon traditional DWT by focusing on two key areas:

  • Objective Denoising Quality Assessment: Instead of relying solely on subjective visual inspection or standard metrics like SNR (which often require a clean reference signal), the authors introduce the Nonzero-Order Periodic Peak (NZOPP) of the normalized autocorrelation function (ACF). This metric leverages the inherent autocorrelation of periodic signals (like heartbeats) and the lack thereof in random noise, allowing for objective evaluation of denoising quality even without a clean reference.
  • Adaptive Threshold Function: The paper proposes an improved threshold function based on the hyperbolic tangent function. This function aims to overcome the limitations of traditional soft (introduces bias) and hard (can cause oscillations) thresholding. The proposed function offers a smoother transition, effectively mitigating pseudo-Gibbs phenomena and constant deviations, while maintaining signal continuity and improving differentiability.
  • Fast Algorithm for Parameter Optimization: To efficiently find the optimal wavelet threshold and the tuning factor for the new threshold function, a bisection fast algorithm is employed, guided by the NZOPP metric.

The overall process involves:

  1. Wavelet decomposition of the signal.
  2. Applying the enhanced thresholding and threshold function.
  3. Wavelet reconstruction.

Datasets

The paper utilizes several publicly available ECG datasets for evaluation:

  • MIT-BIH arrhythmia database: Contains 30-minute, two-channel ECG recordings.
  • Clayton University ventricular arrhythmia database: Focuses on severe ventricular arrhythmias.
  • PTB diagnostic ECG database: Provides high-resolution, 15-lead ECG signals.
  • MIT-BIH noise stress test database: Used to simulate specific noise types.

Additionally, the authors validate their method on heartbeat signals from a fiber-optic micro-vibration sensing system, which represents a real-world non-invasive detection scenario.

Key Insights and Ideas

  • Objective Noise Assessment is Key: The reliance on NZOPP for evaluating denoising quality is a significant contribution, as it bypasses the need for a clean reference signal, which is often unavailable in real-world non-invasive monitoring.
  • Beyond Soft and Hard Thresholds: The proposed hyperbolic tangent-based threshold function offers a compelling alternative, balancing the strengths of soft and hard thresholds to achieve better noise reduction without introducing significant artifacts or bias.
  • Adaptive Parameters for Better Performance: The combination of an objective metric (NZOPP) and an adaptive threshold function, optimized via a fast algorithm, allows the DWT to be tailored to specific noise conditions and signal characteristics.
  • Real-World Applicability: The successful application to fiber-optic sensor data demonstrates the practical utility of the enhanced DWT for non-invasive heartbeat monitoring.

Key Results

  • AWG Noise: The enhanced DWT achieves a 1-5 dB SNR improvement over traditional DWT for Gaussian white noise.
  • Real Noise: For baseline wander, electrode motion, and muscle artifact noise, the proposed method provides SNR gains that are tens to hundreds of times higher than state-of-the-art techniques, with significantly lower RMSE and PRD.
  • Visual Comparison: Visualizations show that the enhanced DWT produces cleaner heartbeat signals with fewer artifacts and better preservation of signal morphology compared to other methods, especially in challenging real-noise scenarios.
  • Computational Cost: While slightly more computationally intensive than some basic wavelet methods, the enhanced DWT is still efficient, with a demonstrated cost of 0.2597s for denoising a signal.

Limitations and Future Work

  • Non-Periodic Signals: The current method relies on the NZOPP metric, which is based on the autocorrelation of periodic signals. This means it may not perform optimally for highly irregular heart rhythms (e.g., severe arrhythmias) where periodicity is lost.
  • Future Research Directions:
    • Developing algorithms for processing non-periodic signals, particularly for complex arrhythmias.
  • Accurately separating and extracting heart rate and respiration from complex physiological data after noise reduction.
  • Embedding the proposed method into IoT devices for continuous, real-time physiological monitoring.

Summary updated

Abstract

Achieving both accurate and real-time monitoring heartbeat signals by non-invasive sensing techniques is challenging due to various noise interferences. In this paper, we propose an enhanced discrete wavelet transform (DWT) method that incorporates objective denoising quality assessment metrics to determine accurate thresholds and adaptive threshold functions. Our approach begins by denoising ECG signals from various databases, introducing several types of typical noise, including additive white Gaussian (AWG) noise, baseline wandering noise, electrode motion noise, and muscle artifacts. The results show that for Gaussian white noise denoising, the enhanced DWT can achieve 1-5 dB SNR improvement compared to the traditional DWT method, while for real noise denoising, our proposed method improves the SNR tens or even hundreds of times that of the state-of-the-art denoising techniques. Furthermore, we validate the effectiveness of the enhanced DWT method by visualizing and comparing the denoising results of heartbeat signals monitored by fiber-optic micro-vibration sensors against those obtained using other denoising methods. The improved DWT enhances the quality of heartbeat signals from non-invasive sensors, thereby increasing the accuracy of cardiovascular disease diagnosis.
BibTeX
@inproceedings{zhu2025enhanced,
  author = {Peibin Zhu and Lei Feng and Kaimin Yu and Yuanfang Zhang and Wen Chen and Jianzhong Hao},
  title = {{Enhanced DWT for Denoising Heartbeat Signal in Non-Invasive Detection}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25061743},
}
Figure