Not a trivial problem

Question

linux

Good day to all!

there is a customized complex for speech recognition (with a well-read dictionary and morphemes for the Russian language)
There are scripts for performing actions regarding voice commands received after recognition.
Speech synthesizer set up.
Scripts were written to monitor the server via SMS using a modified 3G modem ZTE MF180S It is

necessary to teach the server to communicate using voice commands!
Research in this area:
In the system, the modem is represented by devices:
/ dev / ttyUSB0 - unknown purpose / dev /
ttyUSB1 - control port / dev
/ ttyUSB2 - port for sending AT commands : "+99897XXXXXXX",145,,,,0

the script catches this command, parses the string by checking the number of the incoming call with the numbers in the database and, depending on this, sends a command to pick up the phone (ATA) or to end the connection !!!

and now the most interesting thing is that after the modem picked up the phone, you need to transfer a voice file in WAV 8bit / 8000Hz format. They

tried to write to the port (and which port to use for this?) in a stream and blocks of 320 bytes, as a result, silence in the handset!
maybe when you initialize the call for voice transmission, you need to use some other AT commands ???

UPD. I am familiar with Asterisk, please do not offer it as a solution.
To everyone who helps, I promise a plus to Karma, the problem is very important.

Answer 1

Our craftsmen nevertheless solved this problem !!!
On a new whistle from Huawei, after picking up the phone with the ATA
command The command AT ^ DDSETEX = 2 is written to the control port,
after that PCM 8000Hz 16bit Mono sound is written to the interface port in blocks of 320 bytes every 20 milliseconds if written to the port from a file, then position the pointer in the file it is necessary with an offset of 44 bytes (this is the header of the WAV file)
(The problem was solved thanks to the COM port sniffer, and a terrible desire to solve the problem with my own method)

Answer 2

en.wikipedia.org/wiki/Voice_modem_command_set
You can still try this. Once upon a time with nokia through gprs worked.

Answer 3

Judging by the forums on the Internet, it seems that for ZTE modems, voice is transmitted through a separate USB port (/dev/ttyUSB0?), which requires drivers, and moreover, to unlock this function, you need to reflash it.

Answer 4

google carefully about asterisk and chan_datacard or chan_dongle
there is almost everything you need, and receiving a call and “speaking” wav files, etc. etc.
understand the materiel about voice functions - you probably have a firmware with a voice locked only under windows. will have to reflash.
you read about asterisk what you find on Habré - there were a couple of very sensible articles. (and no one canceled Google, pay attention to the article "... a private setup experience ... from scratch")
look at the beginning of the book "Asterisk - the art of telephony" - to understand the dialplan.
write the ne part of the dialplan that you want and that does what you want.
this will allow you not to suffer with the invention of the bicycle.
oh yes, I think your voice recognition system can also be screwed to *.
PS. the screw "labuda" is called dashboard.