| Author |
|
Pete
Intermediate
Joined: December 05 2006
Posts: 12
|
| Posted: February 20 2007 at 1:26pm | IP Logged
|
|
|
Lanscape support folks,
I'm evaluating a wave-to-text product which expects wave files with a sample rate of 22kHz, 16 bit PCM. Using your singlephone example application, I set the data format to be 22kHz, 16 bit PCM and recorded a phone call. However, when I do a right click/Properties/Summary (windows XP) on the saved wav file it says "Audio Sample Rate 8 kHz". Is there any way I can configure the Media Engine to use something different than 8kHz sample rates when recording wav files?
Pete
|
| Back to Top |
|
| |
support
Administrator
Joined: January 26 2005
Location: United States
Posts: 1666
|
| Posted: February 20 2007 at 3:09pm | IP Logged
|
|
|
Hi Pete,
At the moment, the media engine records all call audio to disk using 8kHz PCM. We have made a note of your request and may offer this as a new feature in a future release.
If you guys are planning to make a purchase, we will be sure to add this capability right away.
I assume your speech recognition engine requires 22kHz sample data for best possible accuracy. If you want to perform speech recognition on full duplex recorded phone line audio (i.e. the speech signals of both call parties), then that will have to wait until we add the new feature we talked about above. Otherwise, you can change the single line soft phone app to write raw received 22kHz IVR PCM samples to a file, write 22kHz locally recorded audio samples to a file (see the speech recognition interface of the media engine), after the call completes mix the raw PCM sample files and then convert the raw mixed sample file to a 22kHz wave file to feed to your speech engine. It sounds like a lot of work but it really isn’t.
If you only want to do speech recognition of the received phone line audio, then you can simply use the receive IVR (Rx IVR) interface APIs and obtain all received 22kHz audio that way. You could modify the single line phone app to write the 22k samples to a wave file. Goto the CodeProject and search for example audio source code to see how to write raw samples into a wave file:
http://www.codeproject.com/audio/#Audio
As an alternative, you could simply write the received raw samples from the Rx IVR phone line interface to a file. Then convert the raw sample file to a 22k PCM wave file. You can use the open source Sox utility to do this:
http://sox.sourceforge.net/
There is a good short document that describes how to do this here:
http://www.ling.upenn.edu/phonetics/sox.html
We only mention the use of "sox" so that you may keep moving forward with your evaluation.
Repost as needed,
Support
|
| Back to Top |
|
| |
|
|
Active Topics 
Member List 
Search 
Help 
Register 
Login
LanScape Support Forum -> LanScape VOIP Media Engine™ - Pre-Sales Technical Support
Topic: Recording phone calls using 22kHz PCM
