Thursday, September 9, 2010

HD Voice (aka wideband codecs)

Recently there have been a lot buzz around the concept of HD Voice specially in telco forums.  This post will try to answer the most frequently asked questions around the concept.

What is HD Voice?
High Definition (HD) Voice is a marketing term referring to products and technologies supporting the transmission of audio signals in conversational services with higher quality (that means making the speech signal played in the receiver side as similar as possible to the original speech captured in the originator side).
As you can imagine, higher quality implies higher bandwidth consumption.
The technical term for HD Voice would be the use of wideband codecs as it would be described later.

Why is it important?
Because voice quality impacts in customer satisfaction being a competitive advantage over competence, and because according to some studies by skype  voice quality and average call duration are proportional [1].

Is it something new?
No. Skype popularized this already existing technology in 2003.

What does imply from a technical point of view?
The higher speech quality perceived by customers comes from the fact of capturing and encoding the audio at higher than usual rates.  Traditionally voice signals are captured at 8Khz (that means 8000 samples per second are taken by the capture device ) and in case of wideband codecs the rates are equal or higher than 16KHz. 
Making use of Nyquist theorem of information theory, sampling at 16KHz allows capturing the band of audio frequencies from 0 to 8Khz (doubling the usual 4Khz band when  sampling at 8KHz) [2].
Wideband codecs take advantage of human audio perception to optimize encoding of this captured audio differently at different frequencies.
Even higher rates codecs are usually named as superwideband codecs.

What codecs are the most common today?
The most common wideband codec in the telco environments is AMR-WB.  In case of Internet the most promising and royalty-free solution is being designed in the IETF right now combining existing SILK and CELT wideband codecs [3].  Apple is using AAC-ELD in their Facetime aplication.
You can find a good summary of popular audio codecs here [4].