wirtz
/
OpenAudio_ArduinoLibrary
mirror of https://github.com/chipaudette/OpenAudio_ArduinoLibrary.git
1
0
Fork 0
Code Issues Releases Wiki Activity

spectral: Add SpectralNoise to the GUI tool

Browse Source 
Add the SpectralNoise component to the UI/GUI Audio
Design Tool.

Signed-off-by: Graham Whaley <graham.whaley@gmail.com>
pull/13/head
Graham Whaley 3 years ago
parent 0d481fc332
commit 762df65cad
1 changed files with 109 additions and 0 deletions
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Show Stats Download Patch File Download Diff File
  1. 109
      docs/index.html

109
docs/index.html
Unescape View File

@ -386,6 +386,7 @@ span.mainfunction {color: #993300; font-weight: bolder}
{"type":"AudioLMSDenoiseNotch_F32","data":{"defaults":{"name":{"value":"new"}},"shortName":"LMS","inputs":"1","output":"0","category":"filter-function","color":"#E6E0F8","icon":"arrow-in.png","outputs":"1"}},
{"type":"AudioSpectralDenoise_F32","data":{"defaults":{"name":{"value":"new"}},"shortName":"Spectral","inputs":"1","output":"0","category":"filter-function","color":"#E6E0F8","icon":"arrow-in.png","outputs":"1"}},
{"type":"AudioFilterFreqWeighting_F32","data":{"defaults":{"name":{"value":"new"}},"shortName":"freqWeight","inputs":"NaN","output":"0","category":"filter-function","color":"#E6E0F8","icon":"arrow-in.png","outputs":"NaN"}},
{"type":"AudioFilterTimeWeighting_F32","data":{"defaults":{"name":{"value":"new"}},"shortName":"timeWeight","inputs":"1","output":"0","category":"filter-function","color":"#E6E0F8","icon":"arrow-in.png","outputs":"1"}},
{"type":"AudioMathAdd_F32","data":{"defaults":{"name":{"value":"new"}},"shortName":"mathAdd","inputs":"2","output":"0","category":"math-function","color":"#E6E0F8","icon":"arrow-in.png","outputs":"1"}},
@ -899,6 +900,114 @@ See Compressor and Compressor2 for complete, ready to use classes.</p>
</div>
</script>
<div>
<script type="text/x-red" data-help-name="AudioSpectralDenoise_F32">
<!-- ============ AudioSpectralDenoise_F32 ========= -->
<h3>Summary</h3>
<div class=tooltipinfo>
<p>Spectral Noise Reduction</p>
<p>Remove random noise, such as background noise or machine generated noise, and try
to leave voice intact.</p>
<p>Spectral NR does not try to remove constant tones or clicks and pops.</p>
<p>Spectral NR is generally not useful for cleaning up 'pure tone' sources, such as morse code.</p>
<h3>Audio Connections</h3>
<table class=doc align=center cellpadding=3>
<tr class=top><th>Port</th><th>Purpose</th></tr>
<tr class=odd><td align=center>In 0</td><td>Signal to be filtered</td></tr>
<tr class=odd><td align=center>Out 0</td><td>Filtered Signal Output</td></tr>
</table>
<h3>Functions</h3>
<p class=func><span class=keyword>setup</span>(<strong>AudioSettings_F32 &</strong>settings, <strong>int</strong> n_fft);</p>
<p class=desc>Setup the filter. Must be called if parameters not passed in to the
constructor. Can be re-called if other parameters have been changed to have
the changes take effect.
The maximum FFT size is dictated by the underlying FFT_OA code.
</p>
<p class=func><span class=keyword>enable</span>(<strong>bool</strong> start);</p>
<p class=desc>Turn on or offthe filter. The filter is enabled by default upon
creation.
</p>
<p class=func><span class=keyword>enabled</span>();</p>
<p class=desc>Return the enabled state of the filter.</p>
<p class=func><span class=keyword>getAsnr</span>();</p>
<p class=desc>Return the current ASNR value.</p>
<p class=func><span class=keyword>setAsnr</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the ASNR value. Active SNR.</p>
<p class=desc>Call <span class=keyword>setup()</span> for the change to take effect.</p>
<p class=func><span class=keyword>getVADHighFreq</span>();</p>
<p class=desc>Get the VAD band high frequency cutoff value.</p>
<p class=func><span class=keyword>setVADHighFreq</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the VAD High Frequency cutoff value. Sets the high
frequency value of the Voice Activity Detector (VAD) window.
</p>
<p class=desc>Call <span class=keyword>setup()</span> for the change to take effect.</p>
<p class=func><span class=keyword>getVADLowFreq</span>();</p>
<p class=desc>Get the VAD band low frequency value.</p>
<p class=func><span class=keyword>setVADLowFreq</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the VAD band low frequency value.</p>
<p class=desc>Call <span class=keyword>setup()</span> for the change to take effect.</p>
<p class=func><span class=keyword>getNRAlpha</span>();</p>
<p class=desc>get the NRalpha value. NR alpha is the time smoothing constant. It's
range is clipped between 0.9 and 0.9999 in the code. Settings >0.95
generally recommended, or you can get strong reverb and 'watery' effects.
</p>
<p class=func><span class=keyword>setNRAlpha</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the NRalpha value.</p>
<p class=desc>The change takes effect immediately.</p>
<p class=func><span class=keyword>getSNRPrioMin</span>();</p>
<p class=desc>Get the SNR prior value.</p>
<p class=func><span class=keyword>setSNRPrioMin</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the SNR prior value.</p>
<p class=desc>The change takes effect immediately.</p>
<p class=func><span class=keyword>getNRWidth</span>();</p>
<p class=desc>Get the NR width. Used for the measuring 'span' in the musical note
reduction code.
</p>
<p class=func><span class=keyword>setNRWidth</span>(<strong>int16_t</strong> value);</p>
<p class=desc>Set the NR width.</p>
<p class=desc>The change takes effect immediately.</p>
<p class=func><span class=keyword>getPowerThreshold</span>();</p>
<p class=desc>Get the power threshold. Used to limit the effects of the musical
note reduction code.
</p>
<p class=func><span class=keyword>setPowerThreshold</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the power threshold.</p>
<p class=desc>The change takes effect immediately.</p>
<p class=func><span class=keyword>getTaxFactor</span>();</p>
<p class=desc>Get the noise output smoothing factor.</p>
<p class=func><span class=keyword>setTaxFactor</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the noise output smoothing factor. Typical values are around 0.8.</p>
<p class=desc>Call <span class=keyword>setup()</span> for the change to take effect.</p>
<p class=func><span class=keyword>getTapFactor</span>();</p>
<p class=desc>Get the speech probability smoothing factor.</p>
<p class=func><span class=keyword>setTapFactor</span>(<strong>float32_t</strong> value);</p>
<p class=desc>Set the speech probability smoothing factor. Typical values are around 0.9.</p>
<p class=desc>Call <span class=keyword>setup()</span> for the change to take effect.</p>
<h3>Examples</h3>
<p class=exam>File &gt; Examples &gt; OpenAudio_ArduinoLibrary &gt; SpectralDenoise.ino
</p>
<h3>Notes</h3>
<p> The defaults work OK, but it is worth experimenting with the adjustable
parameters for your specific situation. Try adjusting NRAlpha, NRWidth and Asnr.
</p>
<p> The default FFT size is 256. You may wish to make this larger, which will allow
a finer granularity of 'buckets' to potentially improve the noise reduction, but,
will cost both processor time and memory overheads.
</p>
<p> The maximum FFT size is set by the underlying <strong>FFT_OA</strong> code, which at
time of writing limits the maximum to 2048.
</p>
<p> The code is written to account for the <strong>sample_rate_Hz</strong> passed in with
the audio settings, <strong>but</strong>, this has not been tested. If you do test,
either with a higher or lower sample rate, please report back and consider fixing
this documentation.
</p>
<h3>References</h3>
<p>The best reference on how the Spectral code was designed and works can be found on the
<a href="https://github.com/df8oe/UHSDR/wiki/Noise-reduction"> UHSDR wiki </a>
</p>
</script>
</div>
<div>
<script type="text/x-red" data-help-name="AudioFilter90Deg_F32">

Write Preview
Loading…
Cancel
Save