QUESTION Voicemail Audio Attachments in High Fidelity

[redstonemason]

Embarking on a project to solve hard-of-hearing problems. This is big. An aging population and I am soon going to be part of the demographic.

Step one. Anyone have a solution to my voicemail attachments arriving at the destination in MP3 format. That is my first problem.

I need lossless. It makes a huge difference to the overall experience, especially after I have recommended wired over wifi and higher bandwidth and more GB per month and of course the most important tool - noise cancellation solved with lightweight headphones/boom microphone between the $35 to $200 range.

In "VoiceMail / Setting / Email Config" I have "/usr/local/sbin/sendmailmp3"

Works okay but need to be able to send much bigger and better voicemail attachments.

BTW, I run postfix together with AWS-SES ESMTP email relay to ensure 99.99% deliverability.

 

[dallas]

I think you might be trying to unscramble an egg.
The mp3 quality depends on the voice quality of the incoming call and the codec in use, particularly if it's not alaw or ulaw.

 

[KNERD]

What he is saying is you no control on the incoming fidelity on any call so you are basically wasting your time. However, you still may be able to increase volume and change the format with some post processing like with sox.

 

[dicko]

Given the nature of age related hearing loss, without an individualized, quantified audiometric profile it will be very difficult, for some compressing the dynamic range works better than expanding it , barring 'training sessions' you best leave it 'flat' but sox has lots of options to shape equalizations and dynamics if you cared to have users go through the 'is a better or worse than b . . .' rigmarole

Another possible route is STT | tee email/SMS | TTS using a voice that is well enunciated. Kinda like an automatic

Telecommunications Relay Services (TRS)

Telecommunications relay service (TRS) allow a person who is deaf, hard of hearing, deaf-blind, or who has a speech disability to communicate over the telecommunications network in a manner functionality equivalent to a hearing person who does not have a speech disability using voice...

www.fcc.gov

 

[billsimon]

Step one. Anyone have a solution to my voicemail attachments arriving at the destination in MP3 format. That is my first problem.
...
In "VoiceMail / Setting / Email Config" I have "/usr/local/sbin/sendmailmp3"

You have a custom script that converts the wav to mp3 and then emails it. So: don't do that. By default Asterisk sends you wav files. Start by removing that custom mailer.

 

[redstonemason]

Don't really disagree with anyone so far. Thanks for the professional opinions.

Being a bit of a FLAC nut when it comes to my music, I constantly complain about the fidelity of calls especially from smartphone users (which is just about everyone).

Wish cell service was still AMPS

 

[redstonemason]

Turns out the "/usr/local/sbin/sendmailmp3" script seems to be standard issue that I have not touched.

So I modified it to NOT convert the native wav file to mp3. Makes attachment about 10 times larger and really has no impact on fidelity of spoken voice.

BTW, the Google Speech to Text system (which is not free) is not all that impressive when I run the wav files thru it. Transcriptions are not very accurate. Maybe it's my Canadian accent EH?

 

[KNERD]

Turns out the "/usr/local/sbin/sendmailmp3" script seems to be standard issue that I have not touched.

So I modified it to NOT convert the native wav file to mp3. Makes attachment about 10 times larger and really has no impact on fidelity of spoken voice.

BTW, the Google Speech to Text system (which is not free) is not all that impressive when I run the wav files thru it. Transcriptions are not very accurate. Maybe it's my Canadian accent EH?

I think the IBM one still has a free level.

Free IBM Voicemail Transcription with Incredible PBX 2020 – Nerd Vittles

nerdvittles.com

 

[kenn10]

Yep. IBM does and I use it. I don't know about Canuck accents but I can tell y'all it does a mighty fine job on Southern accents.

 

[dicko]

Turns out the "/usr/local/sbin/sendmailmp3" script seems to be standard issue that I have not touched.

So I modified it to NOT convert the native wav file to mp3. Makes attachment about 10 times larger and really has no impact on fidelity of spoken voice.

BTW, the Google Speech to Text system (which is not free) is not all that impressive when I run the wav files thru it. Transcriptions are not very accurate. Maybe it's my Canadian accent EH?

You speak of fidelity, technically mp3 being 'lossy' will have less 'fidelity' than the native gsm(on cell phones) , that you don't notice that is that the 'energy' is limited to between 400 and 4000 Hz, but the dynamic range remains the same for both codecs (gsm is the result of lots of research, as was g711 back in the '60s). It's all about 'psycho-acoustics', whereby as the body ages, what is heard is more and more 'lossy' , correcting that used to cost $1000's but now only $100's in the US, second guessing such changes unilaterally is probably futile.