What voice codecs are in use for GSM?

Have you looked into the voice recognition used with Dragon Naturally Speaking. It is able to execute commands used in many of the programs on my computer. The same way that a program is used to recognize the command "page down", or all of the other commands, it could be taught to turn on lights, wipers, turn signals or anything else you wanted. Good luck.

http://www.commsdesign.com/design_corner/showArticle.jhtml?articleID=16501605

http://www.eetasia.com/ART_8800339332_499488_TA_b0da9d9c.HTM

jweida, here are two links that will shed some light on the subject.

The short answer is that the EFR requires more processing power than the FR codec, resulting in a more complex and expensive handset, and it takes longer to encode/decode the traffic -- and time is of a premium on a GSM network. Wireless networks in general have problems with latency (200 -400 ms). The human ear begins to detect problems at around 250 ms. (That's why you sometimes seem to be talking over someone on a wireless call, regardless of the technology.) GSM networks have another issue with time in that they are time-division multiplexed -- each user has a given time slot for traffic. So time is especially critical in GSM.

In practice, FR codecs seem to give the best performance from a user perspective. from a network perspective, Half-rate codecs and adaptive multi-rate (AMR) codecs give better performance because they increase capacity. However, the resulting dissatisfaction from customers can offset these gains. We saw that in the USA when Nextel started using half-rate vocoders. In nearly saturated markets, the results were less than stellar.

Determining which codec is in use can be an issue as most chipset manufacturers create chips that support more than one codec and leave it to the handset manufacturers and carriers to decide which one to use.

I guess the issue for you is to decide which codec to use in your development. My recommendation would be the one with the lowest voice quality, FR. Since you are testing handsfree kits, I assume that voice quality is your primary concern. If you go with the lowest voice quality, actual experience by your customers will be as good or better than your testing.

What a great answer. Thanks for your help!

Yes, voice quality is the primary concern in this case. While from a development standpoint I whole heartedly agree to use the lower performing codec, we also have to compare our "numbers" to other suppliers. I've been told that other suppliers are using the EFR codec during testing, which yields higher numbers, and there's nothing in the specification describing which codec to use. Ahhhh, engineering. At least thanks to you I have some more information to share with our customer.