Skip to content

example_usage.md

Latest commit

 

History

History
113 lines (88 loc) · 3 KB

example_usage.md

File metadata and controls

113 lines (88 loc) · 3 KB
Raw

Example Usage

Building the Program

Windows (Visual Studio)

# Run the build script
build.bat

# Or manually with Visual Studio Developer Command Prompt:
cl /EHsc /W4 /O2 /Iinclude src\main.cpp src\wav_reader.cpp src\audio_processor.cpp /Fe:audio_cleaner.exe

Linux/macOS (GCC/Clang)

# Make build script executable and run
chmod +x build.sh
./build.sh

# Or manually:
g++ -std=c++17 -Wall -Wextra -O3 -Iinclude src/main.cpp src/wav_reader.cpp src/audio_processor.cpp -o audio_cleaner -lm

Using CMake

mkdir build && cd build
cmake ..
cmake --build .

Running the Program

Basic Noise Reduction

# Remove background noise from a recording
audio_cleaner -i noisy_recording.wav -o clean_recording.wav

Echo Cancellation

# Remove echo using a reference signal (e.g., the speaker output)
audio_cleaner -i microphone_input.wav -r speaker_output.wav -o clean_output.wav

Advanced Processing

# Custom learning rate for better echo cancellation
audio_cleaner -i input.wav -r reference.wav -o output.wav -l 0.005

Typical Use Cases

1. Voice Recording Cleanup

# Clean up a voice recording with background noise
audio_cleaner -i voice_with_noise.wav -o clean_voice.wav

2. Conference Call Enhancement

# If you have the original speaker output as reference
audio_cleaner -i microphone.wav -r speakers.wav -o enhanced_audio.wav

3. Audio Restoration

# Restore old recordings (noise reduction only)
audio_cleaner -i old_recording.wav -o restored_recording.wav

Parameters Explained

  • Learning Rate (-l): Controls how quickly the adaptive filter learns
    • Lower values (0.001): Slower but more stable
    • Higher values (0.1): Faster but potentially less stable
    • Default (0.01): Good balance for most cases

Tips for Best Results

  1. For Echo Cancellation:

    • Use a clean reference signal recorded from the speakers
    • Ensure reference and input have the same sample rate
    • Start with default learning rate, adjust if needed

  2. For Noise Reduction:

    • Works best with stationary noise (hiss, fan noise, etc.)
    • May need multiple passes for very noisy audio
    • Some musical artifacts may appear with aggressive settings

  3. File Format:

    • Only 16-bit PCM WAV files are supported
    • Convert other formats to WAV first
    • Use same sample rate for reference and input files

Troubleshooting

"Invalid WAV file format"

  • Ensure the file is 16-bit PCM WAV
  • Try converting with Audacity or ffmpeg

"Reference signal size mismatch"

  • Make sure reference and input files have the same duration
  • Check that sample rates match

Poor echo cancellation

  • Try adjusting the learning rate
  • Ensure reference signal is clean and synchronized
  • Check that reference actually contains the echo source

Musical noise after processing

  • This is normal with spectral subtraction
  • Try reducing the aggressiveness (would need code modification)
  • Use higher quality input recordings