Hi all,
Has anyone implemented Voice Fingerprint for Carrier Messages on Call Progress Analysis with Media Server?
I'm trying to implement but it doesn't seem to match any audio file configured for the IDs I inserted on cpa.postconnectresult.machine.list for Voice Mail detection during the postconnect phase. On the log files, it always show a very large distance from my sample and NO MATCH as result. I can see "MATCH" during the preconnect phase, but never on the postconnect.
I took the audio samples from the CPA recordings and edited/trimmed so it only contains the audio that MCP should match against.
Is there any configuration that should be done on the OCS side? (like, change the Call Answer Detection algorithm to a specific value, like telephony_preset - I'm currently using "accurate_am_detection")
It also seem to delay a lot detecting silence on the postconnect phase, sometimes timing out or wrongly detecting as cpd.human.
If somebody has implemented this, any thoughs? Should I leave a silence in the begining of the sample audio files, for example, or change the detection algorithim...
Some relevant information:
MCPs are able to read all the files in the cpa.carriermsg.*
OCS is set to call_answer_type_recognition = accurate_am_detection
MCP is set with the following configuration:
cpa.cm_match_percent = 65
cpa.enable_carrier_messages = true
cpa.postconnectresult.machine.list = 0 1 2 3 4 5 6 7 8
cpa.postconnsilduration = 2900
cpa.postconnsilresult = 18
cpa.preconnectresult.busy.list = 9 10 11 12 13 14 15 16 17
cpa.priority_machine_machinegreetingdur = 950
cpa.priority_machine_maxvoicesigdur = 700
cpa.priority_machine_voicepausedur = 500
cpa.voice_level_db = 17.5
cpa.voice_range_db = 25
Thanks in advance for any ideas.
Regards,
HDS
