[Freeswitch-dev] Speaker Identification implementation

[Christopher Rienzo]

I suspect this will be very similar in implementation to switch_core_asr, so
we could either reuse that interface, or implement a new Speaker
Identification and Verification (SIV) interface.  Hop on #freeswitch-dev to
talk this over and figure out the best approach.

If you implement a new interface, you probably need at least the following
functions:

switch_core_siv_open
switch_core_siv_close
switch_core_siv_pause
switch_core_siv_resume
switch_core_siv_feed
switch_core_siv_get_results
switch_core_siv_start_input_timers

And then some new functions in switch_ivr_async.c:

switch_ivr_detect_speaker
switch_ivr_pause_detect_speaker
switch_ivr_resume_detect_speaker
switch_ivr_stop_detect_speaker
switch_ivr_detect_speaker_start_input_timers

And a new application in mod_dptools.c

detect_speaker



On Thu, Aug 18, 2011 at 8:35 AM, Viktor Jamrich
<viktor_jamrich at centrum.sk>wrote:

> That's it ! I was talking about speaker identification by comparing the
> speaker's voiceprint with
> pre-trained voiceprint saved in a voiceprint repository. Sure, I've
> heard about Nuance is planning to acquire
> Loquendo, another speech processing shark ;)
>
> I was thinking about creating a new abstract interface called e.g.
> switch_sid_interface (SID = speaker identification) which
> could be binded to e.g. a proprietary interface module or (the better
> solution) to mod_unimrcp since MRCPv2 protocol supports
> speaker identification/verification. Using this sid interface w/
> mod_unimrcp FreeSWITCH would be able to use MRCP
> servers for speaker identification/verification. Currently Nuance and
> Loquendo has few MRCP servers supporting this service and
> of course our company is planning to deploy a MRCPv2 server soon.
> Example application is e.g. instead of looking for your customer
> ID or telling your personal ID data to the operator who needs to have
> your identity confirmed before he/she tells you some confidental info,
> the company (whether call center or whatever) would just use speaker
> identification service via MRCPv2 and from 3 to 5 seconds of voice
> he/she will know, whether the person is really the one he/she is
> claiming he/she is or not.
>
> What do you think about it ??
>
> Have a nice day ;)
>
> Sincerely, Viktor Jamrich.
>
> _______________________________________________
> Join us at ClueCon 2011, Aug 9-11, Chicago
> http://www.cluecon.com 877-7-4ACLUE
>
> FreeSWITCH-dev mailing list
> FreeSWITCH-dev at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-dev
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-dev
> http://www.freeswitch.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-dev/attachments/20110818/91b1ca9a/attachment.html