Photoplethysmogram Signal Denoising Method

PPG Denoising Using Two-Stage Morphological Filtering and Dual-Tree Complex Wavelet Transform

This technology is a denoising method for photoplethysmogram (PPG) signals. It cleans baseline components using two-stage morphological filtering, then removes high-frequency noise and reconstructs the signal by separating real and imaginary parts via Dual-Tree Complex Wavelet Transform (DTCWT).

Pulse wave signals are often compromised by low-frequency respiratory drift and high-frequency noise from power lines or electromyography (EMG), hindering accurate heart function assessment. Adaptive filtering requires additional hardware for reference signals, while discrete wavelet transforms suffer from shift sensitivity and frequency aliasing, leading to the loss of valid signal components.

This technology applies a two-stage morphological filter that averages the results of closing-opening and opening-closing operations, then performs threshold-based denoising in the complex wavelet domain after calculating the difference. It maintains low computational complexity while preserving waveforms and improving SNR and RMSE metrics. It can be applied to wearable biosignal devices like smartwatches and pulse oximeters to obtain clean pulse waveforms without needing a separate reference sensor.

‍

‍

Key Features:

• A first-level morphological filtering step that averages the closing-opening and opening-closing operations of the PPG signal
• A second-level morphological filtering step that applies closing-opening and opening-closing operations to the first-level output signal and averages them
• A step of obtaining a morphologically filtered signal by subtracting the second-level output from the first-level output
• A step of processing the morphologically filtered signal using real and imaginary discrete wavelet transforms and reconstructing the PPG signal by removing noise from high-frequency coefficients

‍