So, this is not really news… for the past 12 months or so, the scammers have been busy. They’ve been calling us long before we moved to the NBN, and of course we’ve just hung up the moment they started their spiel. The dead giveaway is the seconds of silence at the start of the call. Dead silence.
Of course, it’s not just the NBN, we’ve had “Amazon Prime”, “Visa”, “Telstra” and others call. Far and above all others has been NBN-related scams.
The latest on the NBN front is they claim your connection has been “compromised” by “other users”, in a British accent.
This is the call I received this morning. You can hear other callers in the back-ground. This is not a professional call-centre, this is a back-yard operation!
The home number recently moved from the PSTN to a VoIP service, so this actually gives me a lot of scope for dealing with this. For now, it’s a manual process: when they call, put them on hold. If I put someone on hold on this number, you better be a Deborah Harry fan!
Long term, I’ll probably look at seeing if I can sample the first 2 seconds of call audio, and if silent, direct the call to a voicemail service or IVR menu. In the meantime, it’s a manual process.
Thankfully we get caller ID now, something Telstra used to charge for.
There’s three big considerations with music on hold:
- Licensing: You need to do the research into how music is licensed in your country. If you want to be safe, go look for something that is “public domain” or one of the “Creative Commons” family of licenses. In Australia, you probably want to have a look at this page if you want to use a piece of commercial music (like “Hangin’ On The Telephone”).
- Appropriateness: is the caller likely to get offended by your choice of hold music? (Then again, maybe that’s your goal?)
- Suitability for your chosen audio CODEC: Some audio CODECs, particularly the lower-bitrate ones, do an unsurprisingly terrible job, with music.
Regarding point (3) always test your music choice! Try different CODEC settings, and ensure it sounds “good” with ALL of them. Asterisk actually supports transcoding, but will choose the format that takes the least effort. RIFF Wave files (
.wav) can be used too, but they must be mono files.
I slapped a CD-quality 44.1kHz stereo version in there, and wondered why it got ignored: that’s why — it wasn’t mono and Asterisk won’t down-mix.
Signed 16-bit linear is a pretty safe bet: effort of going to that to PCMA/PCMU (G.711a/G.711u) isn’t a big deal, but to anything else, you’re at the mercy of the CODEC implementation. Using G.722, things sounded fine, but I found even with Speex settings cranked right up (quality=10 complexity=10 enhancement=true), my selection of audio sounded terrible in Ultra-wideband Speex mode. I wound up with the following in my MoH directory:
vk4msl-gap# ls -l /usr/local/share/moh/ total 8280 -rw-r--r-- 1 root wheel 527836 Aug 29 17:02 moh.sln -rw-r--r-- 1 root wheel 1055670 Aug 29 17:02 moh.sln16 -rw-r--r-- 1 root wheel 2111342 Aug 29 17:01 moh.sln32 -rw-r--r-- 1 root wheel 104793 Sep 5 12:17 moh.spx -rw-r--r-- 1 root wheel 177879 Sep 5 12:34 moh.spx16 -rw-r--r-- 1 root wheel 184617 Sep 5 12:16 moh.spx32
.sln* is for 16-bit signed linear, the 16 and 32 suffixes refer to the sample rate, so 16kHz (wideband) and 32kHz (ultra-wideband). These should otherwise be “raw” files (no headers). Use
sox <input> -r <rate> -b 16 -e signed-integer -c 1 <output>.slnto convert.
.spx* is Speex: Here again, I’ve got 8kHz, 16kHz and 32kHz versions. These were encoded using the following command:
speexenc --quality 10 --comp 10 moh.wav moh.spx
There are various other CODEC selections, but right now, I’ve just focussed on signed linear and Speex since the latter is what needs careful attention paid. I tested between my laptop running Twinkle and the ATA on my network, and when I placed the call on hold from my laptop it sounded fine there, so I figure it’ll be “good enough”.
“Visa Security Department”
So, had “Visa” call me this morning… this too, is another scam. Anonymous caller. Bear in mind I do not actually have a credit card. Never have had one, never will.
They didn’t stick around, seems their system just drops the call if it hears a noise which isn’t a DTMF tone.
Interestingly, both this call, and the previous one were G.711u (µ-law PCM). Australia normally uses A-law PCM. America uses µ-law encoding. What’s the difference? Both are logarithmic encoding schemes. µ-law encoding has a wider dynamic range, however A-law has less distortion for quieter signals.
Almost the same structure as before. Audio CODEC was G.729 this time.