5.5.1 Stereo Signal Low-Band Encoding

26.2903GPPAudio codec processing functionsExtended Adaptive Multi-Rate - Wideband (AMR-WB+) codecRelease 17Transcoding functionsTS

5.5.1.1 Principle

The stereo Low band encoder receives the signals and for encoding. The Low band encoder is based on fidelity optimized encoding of the low band side signal. The Lo side signal is obtained by computing the difference

,

The encoding of the side signal is performed following a similar approach to that of the core encoder, except that the ACELP mode is not used. For each input signal block, the encoder decides in closed loop which encoding models to use. A signal to noise ratio fidelity criterion is used.

All 4 encoding models are based on encoding in the frequency domain a redundancy reduced side signal. In order to account for transients, there is an encoding model that uses pre-echo reduction. Encoding in the frequency domain uses the same split multi-rate lattice VQ.

Within each super-frame, the different encoding modes are:

mode	Duration (length + overlap)	Encoding
0	40 + 8	RR + Pre-echo + AVEQ
1	40 + 8	RR + AVEQ
2	80 + 16	RR + AVEQ
3	160 + 32	RR + AVEQ

The timing chart as well as the possible mode combinations is similar to that of the core encoder and is described in the following figure:

Selection of the encoding mode is done by closed loop search identical to the one used for the core encoder.

5.5.1.2 Signal Windowing

The two signals and are windowed prior to redundancy removal and frequency transformation. This is necessary in order to apply overlap-add to minimize the framing effects due to quantization. The window shape is adaptive depending on the previous coding mode and is similar to that described in section 5.3.5.4. The windowed signals are denoted by and .

5.5.1.3 Pre-echo mode

In order to encode transients more efficiently, a pre-echo mode is used. It is often the case in transients that the energy envelope of the mono signal is highly correlated with that of the side signal. The energy envelope of the mono signal is derived and normalized, it is then used to compensate for the energy envelope of the side signal.

5.5.1.4 Redundancy reduction

For all encoding modes, a balance factor is used in order to remove the portion of the side signal that is correlated with the mono signal. The balance factor is given by

the balance factor is quantized by a uniform scalar quantizer with 7 bits.