With the development of the GSM
network, subscribers have increasing requirements for high speech quality.
Thus, enhanced full rate (EFR) is adopted.
In comparison to FR, EFR optimizes the
speech codec algorithm and improves the speech quality, thus improving the user
experience.
EFR is implemented on the basis of the
Code Excited Linear Prediction (CELP). FR and HR, however, are implemented on
the basis of Regular Pulse Excited - Linear Predictive Coder (RPELPC). The
differences between CELP and RPE-LPC are as follows: In the CELP model, the excitation
signal is a quantified vector. A group of codevectors that are obtained in
advance through calculation forms a codebook. For each frame of speech signals,
a codevector with the minimum perceptually weighted error is selected from the
codebook to serve as the irregular
excitation source. In the RPE-LPC
model, a regular excitation source such as the sine wave is adopted, and thus
the speech lacks clarity.
The Speech Version parameter specifies
whether to enable EFR.
The EFR function can be enabled
forcibly on the BSC. That is, when the BSC determines that the circuit pool on
the A interface and the MS support the EFR, the BSC can actively selects the EFR
as the speech version to improve the speech quality if the speech versions
carried in the Assignment Request message do not include the EFR. The parameter
Allow
Forced EFR in Cell specifies
whether the EFR function can be forcibly enabled