Syllabic Voice Squelch
Improvement suggestion: Add a syllabic voice squelch for SSB.
This is the kind of voice squelch commonly found in commercial land mobile HF transceivers.
It helps to alleviate the constant background noise when monitoring an HF SSB channel, such as: aeronautical, maritime, amateur, NGO, government, or utility.
Some suggested details:
1. User sensitivity adjustment.
2. One second delay of audio output, and the use of cadence for voice recognition, with pre-delay squelching to prevent missing the first syllable or two.
3. On/Off button, plus an Enable/Disable button which includes the delay audio output, so that the operator may manually unsquelch without losing the second of delay.
4. Visual indication of voice squelch opening with blinking tab or tray icon; this would provide a feature for many of us who are using a tabbed browser kiwiSDR for multichannel monitoring.
This is the kind of voice squelch commonly found in commercial land mobile HF transceivers.
It helps to alleviate the constant background noise when monitoring an HF SSB channel, such as: aeronautical, maritime, amateur, NGO, government, or utility.
Some suggested details:
1. User sensitivity adjustment.
2. One second delay of audio output, and the use of cadence for voice recognition, with pre-delay squelching to prevent missing the first syllable or two.
3. On/Off button, plus an Enable/Disable button which includes the delay audio output, so that the operator may manually unsquelch without losing the second of delay.
4. Visual indication of voice squelch opening with blinking tab or tray icon; this would provide a feature for many of us who are using a tabbed browser kiwiSDR for multichannel monitoring.
Comments
In Australia and NZ, it is a very common feature in Land Mobile HF radios.
Icom's voice squelch is found in the IC-746, IC-F7000, IC-F8101 series, and perhaps some others.
In the Icom IC-F8101, voice squelch sensitivity is adjustable, it is a real pleasure to use.
Codan's voice squelch is quite effective and works well, even for weak signals.
Voice squelch also helps to keep the radio squelched during selcalls, interference, or co-channel digital signals.
http://blog.shriphani.com/2015/11/14/voice-activity-detection-in-python-and-swig/
https://dsp.stackexchange.com/questions/2386/libraries-for-voice-activity-detection-not-speech-recognition
https://dsp.stackexchange.com/questions/37059/are-there-any-realtime-voice-activity-detection-vad-implementations-available
Even quite simple methods can work quite well.
Many years ago a built an analogue circuit using filtered and then peak rectified audio that was passed to a comparator with a long time constant on one input and a short time constant on the other, the output was used to trigger a timer which was held on, as long as it kept on being re-triggered. The circuit tracked the average audio level (noise) and triggered when short duration peaks were present (usually speech).
Having a browser tab that indicates activity is a great idea.
Regards,
Martin - G8JNJ
If I remember correctly, one of the older voice squelch circuits utilized a bandpass filter for sensing peaks of human speech energy in the 350Hz to 1kHz range, and that went into a wide lock range PLL detector in the .3 to 5 hertz range.
I visited an Aeronautical MWARA facility where each of the 12 operators had to monitor 10 or more HF SSB and VHF-AM receive channels. They made extensive use of voice squelch with activity indicator LEDs for each circuit.
Here is an interesting page on a voice squelch circuit (albeit not designed for HF SSB) with a description of how it works:
http://www.repeater-builder.com/projects/jpl-vox-sq/ssb-squelch.html
http://websdr.ewi.utwente.nl:8901/
http://www.pa3fwm.nl/technotes/tn16e.html
http://www.pa3fwm.nl/technotes/tn16f.html
Regards,
Martin - G8JNJ
http://homepages.ihug.com.au/~daz2002/tech/FT747-clip.ogg
This recording is a friend's FT747 with the squelch modified to be linear. It made the radio much nicer to listen to!
From 1m 05s to 1m 15s I turned the squelch off to show the difference.
This would be easy to do in DSP - A "noise gate" driven from the amplified AGC signal.
Daz
Over on Simon Brown's SDR radio IO Group, he's just mentioned the Voice Activity Detection (VAD) component of Rnnoise, and thinks that it may be worthy of further investigation.
https://people.xiph.org/~jm/demo/rnnoise/
It's written in C and is available under a BSD license, and uses Recurrent neural networks (RNN) to provide voice detection and noise suppression.
Regards,
Martin - G8JNJ