VE7UWU/sdrangel
1
0
Fork 0
mirror of https://github.com/f4exb/sdrangel.git synced 2026-01-19 03:55:34 -05:00
Code Issues Projects Releases Wiki Activity

Denoiser feature: build only for Linux

Browse Source
This commit is contained in:
f4exb 2026-01-11 09:18:15 +01:00
parent 499b7bb0fe
commit 7e0892801c
2 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
external/CMakeLists.txt vendored
View File

@ -916,7 +916,7 @@ if(ENABLE_FEATURE_MORSEDECODER)
endif()
# For denoiser feature
if(NOT APPLE AND ENABLE_FEATURE_DENOISER)
if(LINUX AND ENABLE_FEATURE_DENOISER)
if (WIN32)
set(RNNOISE_LIBRARIES "${SDRANGEL_BINARY_LIB_DIR}/rnnoise.lib" CACHE INTERNAL "")
set(RNNOISE_ARGS "-DRNN_ENABLE_X86_RTCD=OFF")

4
plugins/feature/denoiser/readme.md
View File

@ -13,7 +13,7 @@ It connects to the "demod" stream of Rx channels similarly to the Demod analyzer
- WFM demodulator
- WDSP plugin (multimode)
Note that this plugin is not available on Mac O/S
Note that this plugin is available only in Linux
The following noise reduction schemes are covered. It can be selected via the (6) combo box:
@ -21,7 +21,7 @@ The following noise reduction schemes are covered. It can be selected via the (6
Noise reduction based on the RNnoise library originally from J.M. Valin. It uses a fork for easier integration in the build system (Cmake support with download of the model file): https://github.com/f4exb/rnnoise
The noise reduction is based on a mix of DSP functions and a recursive neural network (RNN). Basically the RNN helps the DSP functions to adjust the gain in various spectral bands thus very efficiently cancelling the background noise in many situations. Although the model was not particularly trained on radio transmissions it can do a pretty good job at AM, SSB, FM noise reduction however you will need a reasonable SNR to get something out of it else it will consider the audio is just noise. Do not expect it to dig signals out of the noise the goal is to reduce ear fatigue by removing background white noise and other noises e.g birdies.
The noise reduction is based on a mix of DSP functions and a recursive neural network (RNN). Basically the RNN helps the DSP functions to adjust the gain in various spectral bands thus very efficiently cancelling the background noise in many situations. Although the model was not particularly trained on radio transmissions it can do a pretty good job at AM, SSB noise reduction however you will need a reasonable SNR to get something out of it else it will consider the audio is just noise. Do not expect it to dig signals out of the noise the goal is to reduce ear fatigue by removing background white noise and other noises e.g birdies. Results for FM signals may vary.
You will find all the details about RNnoise here: https://jmvalin.ca/demo/rnnoise/