Convert voice to text message

[radiotrack1]

I have a project I am working on. I would like to get all of my county dispatch calls in the form of text messages or emails.

I currently am a feed provider and I have a scanner connected to my pc. My county uses two tone encoders for call paging. They also utilize rip-run pager fax.

Is there something out there that will detect the two tone page, listen for about 30 seconds, convert the 30 seconds of voice to text and then send me an email or text?

Or is there an easier way to do this?

 

[DC31]

I have a project I am working on. I would like to get all of my county dispatch calls in the form of text messages or emails.

I currently am a feed provider and I have a scanner connected to my pc. My county uses two tone encoders for call paging. They also utilize rip-run pager fax.

Is there something out there that will detect the two tone page, listen for about 30 seconds, convert the 30 seconds of voice to text and then send me an email or text?

Or is there an easier way to do this?

Converting voice to alpha-numeric text as you are proposing is extremely challenging. There are just so many variables involved that the reliability of the conversion becomes quite poor. It is all affected by the quality of the audio signal, the length of the message, the differences in dispatchers' voices, and so on. Those who have tried it get some quite hilarious results.

However, there is a program that does exactly what you are describing without the text conversion, it texts you the audio recording. TwoToneDetect. Will that suit your needs?

 

[DC31]

Take a look at this previous discussion:

http://forums.radioreference.com/st...etect-works-python-based-128.html#post2309570

 

[tulsascan]

check this out and maybe work this angle may get you there fairly easy deaf people use it all the time. just a thought.

"http://www.optus.com.au/shop/mobilephones/features/voicetotext"

 

[ffryan76]

Take a look at this previous discussion:

http://forums.radioreference.com/st...etect-works-python-based-128.html#post2309570

DC31,
Did you ever get consistent results using what was discussed? I couldn't figure out how it was done.

 

[DC31]

DC31,
Did you ever get consistent results using what was discussed? I couldn't figure out how it was done.

No, i never got consistent results at all. That is a big part of the challenge

 

[mnewmanstc]

I've been looking into a similar system for some time as well with less than encouraging results.
Speech recognition just can't cope with the poor quality of the radio comms. I've run the audio straight into a site like https://dictation.io/ which uses Google's speech recognition back end (which in my experience has been pretty spot-on), and the results were inconsistant at best, and usually just wrong.

 

[kc2kvy]

You might be better off writing software to parse the rip n run sheet. Set up a fax machine to save it as a pdf, run text recognition on that, read the relevant data, and send that out.

Probably a lot of work, but more reliable.


Sent from my iPhone using Tapatalk

 

[ffryan76]

I've been looking into a similar system for some time as well with less than encouraging results.
Speech recognition just can't cope with the poor quality of the radio comms. I've run the audio straight into a site like https://dictation.io/ which uses Google's speech recognition back end (which in my experience has been pretty spot-on), and the results were inconsistant at best, and usually just wrong.

I've have a html script that does something like that. It worked really well until Dispatch stopped sending the rip and run.[emoji19]

 

[mnewmanstc]

As I've looked more into the speech transcription capabilities of HTML5 and looked at various implimentations of it like at that Dictation.io and Speechnotes.co I'm starting to think this might actually be a viable system.

I'm thinking it might be possible to augment the javascript in Trunking Recorder to submit each call against the detection and then submit the results back to another database.

Right now I'm routing one instance of Chrome into another via VBCable and got the following transcription:

"I'm from the front door but I want to do in an emergency living room what about in Philly number it with a 93 year old female patient please have about 20 minutes now it has clotted at this time patient vital signs are solid BP is 160 over 88 Colorado 63 frustrating 69 labor to set on 99 percent room hypertension shorty Patriot about 2 minutes and it question to order coffee room"

Here's the actual call for comparison:
http://sumterscanner.from-sc.com/me...S Ambulance to Hospital_4022-Tuomey__722D.mp3

 

[ffryan76]

I think these Speech recognition engines recognize normal speech patterns as it try to put out what a person is saying. When dispatchers talk it isn't a in a normal sentence structure.

 

[mnewmanstc]

It's a little of both, actually. That "speechnotes" exposes part of the 'initial guess" detection framework in the JS console. It builds up it's interpretation one word at a time and revises the whole thing as the sentence goes on, but either way all of the in-between guesses are programmatically accessible.

For my own purposes I had envisioned a 'hotword' notification system for various street names, businesses, family names, etc that I was interested in... I would image it possible to keep a log of all the various interpretations logged per call for mass inspection, since at this time a spot-on transcription isn't reasonable to expect.

 

[a417]

capturing the OTA text pages would be infinitely easier, if they broadcast them...I'd start there. There are several remarkably competent programs out there that will even dump them into files for post-processing, data recall...etc.

 

[mnewmanstc]

Are the OTA messages sent as part of trunking systems? I've seen PDW for pager messages, but I'm unfamiliar with this from an emergency service standpoint.

 

[radiotrack1]

Our county sends text out to members of the department being paged. But I want a text for each call that is sent out. I have unlimited text messaging so the number of calls is not an issue. Some departments use active 911 but you have to be a member of each one you want text for. I have used carolinas fire page but we do not have an active submitter in my area so the pages are git and miss.

 

[nated1992]

No, i never got consistent results at all. That is a big part of the challenge

What where you attepting to do with it?

Im pretty skilled with TTD i also have a commercial contract with the developer of TTD for other applications.

 

[nated1992]

I have a project I am working on. I would like to get all of my county dispatch calls in the form of text messages or emails.

I currently am a feed provider and I have a scanner connected to my pc. My county uses two tone encoders for call paging. They also utilize rip-run pager fax.

Is there something out there that will detect the two tone page, listen for about 30 seconds, convert the 30 seconds of voice to text and then send me an email or text?

Or is there an easier way to do this?

TTD will send you the audio of dispatches provided you know the tone sets for what your wanting the here.

Also taking the dispatch audio and converting it into a text is going to be impossible, all kinds of things from back ground noise to static will distort what is heard and there isn't any way to cancel all that out .