[CQ-Contest] SSB Voice Keying

[Rudy Bakalov]

About a year ago some friends and I looked into this very topic. The goal- turn SSB contesting into CW contesting by revamping the N1MM+ voice keyer.

The first step was to find a good algorithm for text to speech. We analyzed the list of call signs from the check partial database and built a statistical model to determine the minimum number of 2 and 3 letter combinations that needed to be pre-recorded to voice the entire CP database. The number of such 2 and 3 character snippets turned out surprisingly low.

The next step was to rewrite the N1MM voice keyer module.  We had a team of professional developers, all hams, willing and excited to work on the project. However, the N1MM team turned down our request for access to the source code. At the time they were not interested in further development of the voice keyer module.

We are stubborn and looked at other ways of achieving integration. We prototyped a solution where we can "grab" a call from the entry window and run it thru the new voice keyer module. It worked OK, but was not as seamless as if the feature was built in into N1MM

As you could imagine the effort died as nobody wants to develop software that won't see much of an adoption.

By the way, while at it, we pre-processed all pre-recorded chunks of audio for maximum PEP power using commercial broadcast industry tools. As a result, you could not only run a SSB contest without ever saying a word, but also sound much louder and clearer due to the pre-recorded audio. Static preprocessing are not very effective as during a 48 hour period, or in a M/M environment, the operator's voice changes during the contest and the initial preprocessing settings need adjustment. By prerecording you do the preprocessing once and no further adjustments are needed.

Finally, while at it, we profiled extensively N1MM for CW and RTTY performance, memory leaks, and a bunch of other software performance tweaks. The software can certainly benefit from some tweaks.

To be clear, I do not mean to drag the N1MM team under the bus. Without a doubt they have and still are doing an incredible job "powering" the contesting community. I am sure the team is constantly bombarded with requests for all sorts of features and modifications. It is entirely their prerogative to decide how to manage their development efforts. So if a better voice keyer is not on the roadmap, that's OK and we should respect that.

Rudy N2WQ

Sent using a tiny keyboard.  Please excuse brevity, typos, or inappropriate autocorrect.


> On Mar 2, 2016, at 12:00 AM, Björn SM0MDG <bjorn at sm0mdg.com> wrote:
> 
> Hi,
> 
> I have for a few years used the voice keying features of N1MM, mainly when operating my station via remote from our city apartment. This has worked out fairly well and has been keeping my family asleep during the most intense moments of the night. But with the upgrade of N1MM to N1MM+ voice keying is not as smooth anymore. The biggest difference is that a slight delay is being introduced between letter and number files making the exchange sounding very robotic.
> 
> I’d like to hear what other solutions are out there. I am looking for something that lets me run a complete SSB contest without saying a word into the microphone. I want it to sound as natural as possible and it should handle sequential number exchanges or anything else contest promotors can throw at us as an exchange. It should be possible to use it with N1MM and/or Wintest. Software solution is preferred but I won’t rule out hardware if not software can provide this. 
> 
> Is there anything like this out there?
> 
> 73 de Björn,
> SM0MDG
> SE0X
> 
> 
> _______________________________________________
> CQ-Contest mailing list
> CQ-Contest at contesting.com
> http://lists.contesting.com/mailman/listinfo/cq-contest