Noise reduction combining time-domain epsilon-filter and time-frequency epsilon-filter

Area of Science:

• Signal Processing
• Digital Filtering
• Acoustics

Background:

• Traditional time-domain epsilon-filters (TD epsilon-filters) effectively reduce noise but struggle with large noise amplitudes.
• Existing methods are limited in handling non-stationary noise and signals with drastic variations.

Purpose of the Study:

• To develop advanced epsilon-filter techniques for enhanced noise reduction.
• To address the limitations of TD epsilon-filters in handling large amplitude noise.
• To improve signal preservation during noise reduction.

Main Methods:

• Introduction of the time-frequency epsilon-filter (TF epsilon-filter) applied to complex spectra.
• Development of a combined TD epsilon-filter and TF epsilon-filter approach.
• Proposal of a variable time-frequency epsilon-filter for adaptive noise reduction.

Main Results:

• The proposed TF epsilon-filter and combined methods effectively reduce noise across a wider range of amplitudes, including large noise.
• Experimental results demonstrate superior performance compared to conventional noise reduction techniques.
• The advanced filters preserve signals with drastic variations, such as speech signals, more effectively.

Conclusions:

• Advanced epsilon-filter techniques, including TF and variable methods, offer significant improvements in noise reduction.
• These methods overcome the amplitude limitations of traditional TD epsilon-filters.
• The proposed algorithms provide a robust solution for noise reduction in various signal types.