C library for audio noise reduction and other spectral effects

• Background
• De-noise Algorithms
• Build
• Installation
• Usage Examples
• Development
• Contributing
• License

This library is a standalone, modular spectral processing engine originally based on noise-repellent. It decouples the DSP algorithms from any specific plugin API (like LV2), allowing for easy integration into various applications.

The core architecture uses a unified spectral processor concept. The library currently implements advanced spectral denoising using efficient circular buffering (SbSpectralCircularBuffer) and modern STFT processing. It is designed to be extensible, supporting future additions like de-crackle or de-click algorithms.

The library implements a sophisticated spectral gating algorithm enhanced by several psychoacoustic and statistical techniques:

The fundamental noise reduction method uses spectral subtraction with proprietary framing and windowing to minimize artifacts.

A Non-Local Means (NLM) algorithm filters the spectrogram in both time and frequency domains simultaneously. This preserving structural details of the signal while reducing musical noise and "burbling" artifacts often associated with simple spectral subtraction. Note: This feature is computationally intensive and requires SIMD optimization (enabled in Release builds).

To preserve transients and prevent over-processing, a psychoacoustic masking model estimates the auditory masking threshold. If the signal components are strong enough to mask the noise naturally, the "veto" system prevents unnecessary noise reduction, preserving the natural character of the audio.

Specialized handling for tonal noise components allows for more aggressive reduction of static hums and whines without affecting broadband characteristics.

The whitening feature (noise floor recovery) has been refined to be transparent at 0dB reduction, ensuring that the noise floor texture is natural and consistent with the reduction amount.

In addition to manual noise profile capture, the library supports adaptive noise usage for changing noise environments.

If you wish to compile yourself and install the library you will need:

• A C compiling toolchain (GCC or Clang)
• Meson build system (0.60.0 or newer)
• Ninja build tool
• FFTW3 library (float version)
• OpenMP for parallel processing
• libsndfile (optional, for examples)

git clone https://github.com/lucianodato/libspecbleach.git
cd libspecbleach
meson setup build --buildtype=release
meson compile -C build
sudo meson install -C build

You can configure the build using -Doption=value:

• enable_examples: Build example applications (default: false). Requires libsndfile.
• enable_tests: Build unit and integration tests (default: false). Requires libsndfile.
• static_deps: Link internal dependencies (like FFTW3) statically (default: false). Useful for creating self-contained libraries.
• custom_warning_level: 0-3 (default: 2). Controls compiler warning verbosity.
• treat_warnings_as_errors: Treat compiler warnings as errors (default: false).
• enable_sanitizers: Enable sanitizers in debug builds (default: false).
• sanitize_address: Enable AddressSanitizer (default: false).
• sanitize_undefined: Enable UndefinedBehaviorSanitizer (default: false).
• lv2dir: Install directory for LV2 bundles (absolute path or relative to prefix) (default: '').

Example for a static build with examples:

meson setup build -Dstatic_deps=true -Denable_examples=true
meson compile -C build

Simple console apps examples are provided to demonstrate how to use the library. It needs libsndfile to compile successfully. You can build them with:

meson setup build --buildtype=release -Denable_examples=true
meson compile -C build

To process a file using a manually captured noise profile (first N frames):

./build/example/denoiser_demo --learn-frames 10 input.wav output.wav

To use the adaptive noise estimator:

./build/example/denoiser_demo --adaptive input.wav output.wav

./build/example/denoiser_demo \
  --adaptive \
  --reduction 20 \
  --whitening 50 \
  --smoothing 0.0 \
  input.wav output.wav

It will recognize any libsndfile supported format.

For development builds with debugging symbols:

meson setup build --buildtype=debug
meson compile -C build

The project uses clang-format for code formatting. To format the code:

meson compile format -C build

If tests are enabled:

meson setup build -Denable_tests=true
meson test -C build

To generate coverage reports locally, you will need gcovr or lcov installed.

meson setup build --buildtype=debug -Db_coverage=true
meson compile -C build
meson test -C build
ninja -C build coverage-html

The report will be available in build/meson-logs/coveragereport/index.html.

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version.

See LICENSE for more details.