Noise removal is a critical aspect of signal processing, yet existing algorithms often
exhibit limitations when applied to complex and highly corrupted signals. In particular,
the suppression of speckle noise from vibration velocity signals acquired t...
Noise removal is a critical aspect of signal processing, yet existing algorithms often
exhibit limitations when applied to complex and highly corrupted signals. In particular,
the suppression of speckle noise from vibration velocity signals acquired through Laser
Doppler Vibrometry (LDV) remains challenging due to high computational cost and
limited denoising efficiency. These methods frequently struggle to capture underlying
signal patterns while maintaining the processing speed required for real-time applications. To address these issues, a Total Variation (TV)-based denoising framework is
proposed for low-noise signals, while deep learning based generative models such as
the Bidirectional Generative Adversarial Network (BiGAN) and Variational Autoencoder (VAE) are investigated for higher noise distributions. The performance of TV
based desnoising is compared with their corresponding windowed variants, whereas
the learning-based frameworks are evaluated against existing learning based denoising
approaches reported in the literature. The results demonstrate that the proposed TVbased method effectively denoises low level noisy signals while preserving their temporal
structure, achieving signal-to-noise ratio (SNR) improvements of up to 25 dB significantly outperforming conventional counterparts. In contrast, the deep learning models,
trained on preprocessed noisy–clean signal pairs, successfully learned noise characteristics and error trends, enabling robust speckle denoising under highly noised conditions. A comprehensive comparative analysis further reveals that the BiGAN combined
with a low-pass filter (BiGAN+LPF) achieved the highest performance, improving the
SNR from 4.58 dB to 20.26 dB for higher noise level signals. In real-time environments, the BiGAN+LPF framework established itself as a state-of-the-art method for
high-frequency speckle noise suppression, achieving an average ∆SNR improvement of
16.9 dB compared to other models.