Quick look at my "massively parallel police scanner" The beginning of an event early warning system?

[@DoogieLabs]

How much Computing Horsepower do YOU think it will take feed the firehose through speech to text engine? Ideas on feasible ways to do this are welcome.

[@billyjoejimbob75]

A local firefighter told me they're going to go digital soon, so I looked at a few dongles that started off around 25mhz. Haven't decided yet, but I'm sure I'll end up with one eventually.

[@slugcult1973]

{Me, just before watching this video:: "Hmm, the title sounds interesting. Lemme watch a little bit. Seems a little too technical for me, but I'm relatively clever and sure I can figure it out." {Me while watching:: "....umm....what the f....??" {Me after watching for ten minutes :: "Who am I kidding?? I can watch and/or read further info about massively parallel police scanners for the next 20 years, and I STILL wouldn't have a clue about them."

[@Gmtail]

What do you do about the encrypted traffic which is ever increasing now that it’s software based in the various trunked radio systems.

[@RaymondArias]

Where's the rabbit?

[@jaymzx0]

Very nice setup. I happened across this vid after finding your hovercraft vid while coming up ideas for my new HP C7000 fans. I have a similar hardware setup with four RTL-SDRBlog dongles plugged into a 4 channel Channelmaster distribution amp, and it all lives in a fan-cooled box as they get toasty. This is all fed into a little Core i5 Intel NUC machine running Unitrunker and Trunking Recorder. The CPU pain is real! I'm only running nine voice VFOs across three SDRs, with the fourth being dedicated to a control channel which is running at 0.96MS vs 2.4MS on the voice receivers to help with CPU. I can run the little thing harder but I don't want to listen to the fan all day :). I live in a condo, so I have to keep the antennas I have indoors for the most part (I have an APRS digi and ADS-B rx on the patio, but those are whole other topics). I made a little 1/4 wave ground plane with some 3D printed parts and 10ga ground wire and it works really well. I've recently been getting some intermod from some data transmissions around here somewhere, so I'm working on that at the moment. I can't be happy until I have full quieting. You know how it goes. Anyway, just subscribed. Lots of fun stuff going on here. Keep it up and good luck!

[@PhattyMo]

Any plans to incorporate ADS-B,for keeping an eye on planes?

[@kix5457]

Is "calculon" a Futurama reference? :D

[@DoogieLabs]

Yes it is!

[@Shiunbird]

I’m starting to do some work on speech to text solutions (video conference records get processed after the call is done, and then keywords are identified and people can look for stuff using a web page) at work, so here go my 2 cents: System bandwidth is a must. We are using some closed source stuff and developing open source because license. The video conference is recorded as it is (H264 with mp3 audio) encapsulated on MOV. Our flow then converts the audio to WAV, sends it to the speech to text engine, feeds the database with the results and then discards the WAV file. The thing is, we can afford a slowdown. It’s not time sensitive, so the backlogs are cleared overnight. There’s a flag we can set to prioritise certain jobs (usually big boss conferences)’. It just rearranges the queue. MP3 conversion takes enough CPU time to be worth not having to do it if you are extracting speech from tons of records at the same time. In your case, since you want to go straight from audio to text, you will need tons of system bandwidth. In parallel you would need to encode to MP3 (or whatever you choose) for storage. In our experience, CPU usage decreases and accuracy increases the clearer the audio is. That can be a problem for your use case. To deal with the volume, you can have certain keywords have priority. For example, if 15 seconds into a transmission you don’t hear something specific, then that job can be discarded, so you are sure there’s CPU time to deal with bombs, tornados or whatever your priorities are. You are probably not interested in cats stuck on top of trees. I’ll try to pull some logs and see what’s the resource usage for the transcription, compare it to our hardware and let you know.

[@erlendse]

Got any way for GPU offloading part of the processing? OpenCl e.t.c Just seems to me like something that it should be possible to speed up/offload quite much.

[@DoogieLabs]

Yes. There is a 1050Ti installed in the server as well. I have compiled CUDA successfully into ffmpeg and can leverage the GPU for encoding tasks. This hasn't been integrated into the scanner yet though, but ultimately you may be right in that GPU acceleration and something CUDA or OpenCL may be best to attack the issue.

[@erlendse]

@@DoogieLabs Totally. I don't really consider CUDA since it's a single vendor thing. For a private jig it may be ok tho.

[@aavalos163]

Why do you need a police scanner?

[@DoogieLabs]

@@aavalos163 Why not?

[@ObliterationOfMankind]

Have you think about installing a recent xanmod rt kernel and see if you can pull more out of those cores? I mean, I would do it just for the sake of listening to Brighton PD calls LOL. Btw, bashtop > glances

[@ytrewq6789]

I wished I had discovered your channel sooner?... But I hope I am not offensively late to the party?... Cheers from a new subscriber from Canada!!!...

[@rutgerjonaker5422]

Please make a update

[@grahamr6545]

I would love something like this..

[@DoogieLabs]

I think I will put together instructions potentially on getting it setup, but it is a bear indeed. It took two of us the better part of a week to get things setup and tuned in.. Then the whole breaking the site or database getting it all setup. It is a great hacker puzzle indeed.

[@254DEUS254]

That's really impressive.

[@Conmega1]

Great video on the project Doog! Well covered, and great job delivering it! Really looking forward to this project blossoming into something you can offer as a service or sell as a product! You definitely deserve some return on your massive time investment :)

[@aavalos163]

Why would someone need one of this?

[@Neojhun]

Not a person. It for serving a database to a community to use it as reference.

[@DoogieLabs]

Neojhun is right, it is a tool where the community can both understand what is happening within that community, as well as review it later for extra evidence or accounts of a given situation. In the case of the Police, the transmissions are recorded anyways, but in the case of the event center traffic, schools, etc. That stuff is not recorded, but now can be and later reviewed. Making something like this available to the authorities, media, and city staff can aid in incident response and training.

[@malcore1184]

I dont understand anything but nice hardword

[@thiesenf]

You should definitly check out www.websdr.org and put your SDR's there...