Adaptive audio noise filter -Free.

[admin]

Adaptive audio filters are noise reducers that continually measure noise and subtract this constant from the signal containing wanted audio with unwanted noise. They tend to work much better than static filters based on just removing frequencies not essential to successful copying of the signal. A simple example is the automatic notch filter, which detects a constant tone and then subtracts it. The notch filter is not so good with more than one or two tones for subtraction.

While much work has been done with such filters to clean up medical ECG recording: https://en.wikipedia.org/wiki/Adaptive_ ... e_filters ,
the principles are the same for audio processing, including echo suppression.

The basic premise behind an adaptive filter is that the noise is constant or repetitive, while the wanted signal is a varying waveform such as speech of a CW stream. The effectiveness of the filter depends on how well it can distinguish noise from signal, and continually learn and change its current definition of these to provide the best filtering. It seems there are ways to achieve good adaptive filtering free. It is not clear why the NR control on many tranceivers is typically not a modern adaptive filter, but a traditional design that does not include adaptive learning. Why not in a SDR today???

Commercial but expensive examples are pricey and include:

clrdsp.png

Also available with speaker and stereo.

bhi_DSP.png

Xiegu_GNR1.png

There are others such as the discontinued Radio Shack and MFJ units, still sometimes available at online markets and swap.

To do Adaptive Noise Filtering Free:

Ham radio noise is typically different from microphone meeting-room background noise, but some available tools work surprisingly well. The aim for the ham is to restore intelligible communication when buried in QRM. These are all, AFAIK, Open Source and free for personal use. It seems much less easy to modify Windows sound free, though if someone knows how, please post a howto. Mac works like Linux.

To do noise reduction free in software, the most commonly used app is Audacity, a comprehensive sound recording and editing package. While easy to learn, it has many options for cleaning up a recorded audio stream, -not quite the same thing as live adaptive filtering. Very good, but maybe not exactly what's needed for the ham? We need real-time, live stream, and able to clobber heavy QRM.

gr-adapt
Available at https://github.com/karel/gr-adapt/tree/master with some documentation.

NoiseTorch
Available at https://github.com/noisetorch/NoiseTorch with documentation. Designed really for microphones with bad noise, NoiseTorch lets you direct the audio to NoiseTorch which cleans it up and then gives a "virtual microphone" which may be used to record or hear the cleaned-up audio.

RNNoise
Available at https://hackernoon.com/enabling-instant ... n-on-linux with a detailed howto. The instructions are detailed and clear, though familiarity with basic command line use expected. RNNoise is really intended for removing background noise in meetings, so as they say, your ham mileage may vary. RNNoise is an audio processing library designed for use with Pulseaudio or Pipewire.

Other links how to Noise Filter:
Basic explanation of how to add noise filtering to current Linux sound. https://askubuntu.com/questions/18958/r ... pulseaudio

Explanation of Pipewire audio filtering with detailed howto. https://medium.com/@gamunu/linux-noise- ... f997f6764d

Quick YouTube howto add Pulseaudio filter. https://youtu.be/lTodCeVAfpI?si=igiUWT87leYbH9ov

A thought: most sane people use these filters with microphones, while we typically use USB devices to get audio out of our rigs. Lots to experiment with here, to find which and how to use these devices instead of the example microphones.