HFU HF Underground
General Category => General Radio Discussion => Topic started by: ChrisSmolinski on December 23, 2015, 1356 UTC
-
Synthesized IDs are quite popular with stations, but are extremely difficult to decipher under poor, or even sometimes average, listening conditions.
There's an interesting paper here that discusses the problems older listeners have with understanding synthesized speech. Much of it has to do with missing frequency content:
http://www.sparc.ac.uk/media/downloads/executivesummaries/exec_summary_wolters.pdf (http://www.sparc.ac.uk/media/downloads/executivesummaries/exec_summary_wolters.pdf)
I suspect the same problem plagues shortwave radio, due to interference, and the narrow bandwidths we sometimes have to employ. (Not to mention the fact that there's a lot of older listeners)
This is not a criticism, just a tip - if you want listeners to be able to figure out your station name, make it as easy as possible. This includes having a clear ID, spoken by a real person. :)
-
Many a radio station I have known over the years got a 3 party to do ID's for radio... even ham radio at one time guys would get some lady friend to do recorded ID's a great kick in some cases.
And Chris you are very right... whenever my computer comes on with it's own voice... I shut the speakers off... most irritating...
A very Merry Christmas to you and yours and a very Happy New Year filled with great DX :-)
-
Bumping this back to the top... I think it's worth a re-read.
-
It is worth a re-read and I fully agree.
-
Looks like the original link is dead, but archive.org to the rescue! https://web.archive.org/web/20180528075210/http://www.sparc.ac.uk/media/downloads/executivesummaries/exec_summary_wolters.pdf
-
I recall reading comments from a number of pirates that they deliberately obscured their ids just to annoy the listeners. It seems to have worked.
-
I recall reading comments from a number of pirates that they deliberately obscured their ids just to annoy the listeners. It seems to have worked.
I'm tempted to start posting obscured logs ;D
-
I can understand the desire to use a speech synthesized ID. Not everyone wants to offer on air clues to their identity. What can be helpful if you really want to use one is:
- Repeat it a couple of times. Sometimes the second or third time heard is the eureka moment. Slight fading can quickly wipe out the intelligibility of the generated speech.
- Add some small pauses between words. This allows the brain to analyze what is being heard if it is difficult to understand and catch up before the next word is sent. About a 1/2 to 3/4 of a second between words is a good place to start.
- Be mindful of the accent used. Surprisingly, I do better with certain British female voices vs. American ones. This will likely be different for other listeners.
- Use the text generated speech as an interval signal. Broadcast it several times on sign on and sign off in addition to periodically during the broadcast so that the ID can be nailed down.
- Make sure the audio level is normalized with the rest of your content. Often the difficulty understanding it comes from under or over modulated generated speech.
Just some thoughts as a listener...
-
Σ had some very good suggestions. I'll also add: Send a CW or SSTV ID. Especially SSTV, as not everyone knows CW, and CW decoding software often fails under poor conditions. As a bonus, listeners can use the frequency offset information provided by my Black Cat SSTV app to retune stations that are off frequency :)
https://www.blackcatsystems.com/software/sstv.html
-
My problem with SSTV is I forget to turn on the decoder most of the time. :) But SSTV does help with positive ID quite a bit, I agree.
CW if sent slow enough will allow you to copy the characters individually and then do a look up on a Morse chart.
Will all these suggestions give someone an idea to start a "no content but ID in 50 different ways" pirate station? :D
-
Surprisingly, I do better with certain British female voices vs. American ones.
Me too Especially when it's Elizabeth Hurley. And I'm not a Real Person just an incredible simulation.. 8)
tAC
-
Synthesized IDs are quite popular with stations, but are extremely difficult to decipher under poor, or even sometimes average, listening conditions.
There's an interesting paper here that discusses the problems older listeners have with understanding synthesized speech. Much of it has to do with missing frequency content:
http://www.sparc.ac.uk/media/downloads/executivesummaries/exec_summary_wolters.pdf (http://www.sparc.ac.uk/media/downloads/executivesummaries/exec_summary_wolters.pdf)
I suspect the same problem plagues shortwave radio, due to interference, and the narrow bandwidths we sometimes have to employ. (Not to mention the fact that there's a lot of older listeners)
This is not a criticism, just a tip - if you want listeners to be able to figure out your station name, make it as easy as possible. This includes having a clear ID, spoken by a real person. :)
I did an awful lot of audio processing on the few that I had made, to make them more copyable-- running them thru the amplifiers, and high and low pass filters on "Wavepad" and then testing them through a transmitter.
I had used an ATT website to create the originals, but I remember now that they were very muffled and unreadable the first time they ran on the air, until I had worked them over.
You are certainly right, none of the voice synthesizers I used made fit copy for broadcast without considerable reworking of the audio.