AN2094 Freescale Semiconductor / Motorola, AN2094 Datasheet - Page 3

• Manufacturer Part Number: AN2094
• Description: ITU-T G.729 Implementation on StarCore SC140
• Manufacturer: Freescale Semiconductor / Motorola

1.2.1 Encoding
The G.729 encoding scheme is based on a code-excited linear-prediction model. In this model, the locally decoded
signal is compared with the original signal. The filter parameters are then selected to minimize the mean-square
weighted error between the original and reconstructed signal. The encoding principle is shown in Figure 2.
The input samples are passed through a 140 Hz high-pass filter, a tenth-order LP analysis is performed on the
samples, and the resulting LP parameters are quantized in the line spectral pair (LSP) domain with 18 bits. The
input frame is divided into two 5 ms subframes to optimize tracking of the pitch and gain parameters and reduce
the complexity of the codebook searches. Interpolated LP coefficients are applied to the first subframe, and
quantized and unquantized LP filter coefficients are applied to the second subframe. The excitation in each
subframe is represented by both an adaptive-codebook contribution, which simulates the pitch structure of the
voiced sounds, and a fixed codebook contribution, which simulates unvoiced sounds. The adaptive and fixed
codebook parameters are transmitted every subframe.
The adaptive codebook component represents the periodicity in the excitation signal using a fractional pitch lag
with 1/3 sample resolution. The adaptive codebook is searched using a two-step procedure. An open-loop pitch lag
is estimated per frame based on a perceptually weighted speech signal. The adaptive codebook index and gain are
found by a closed-loop search around the open-loop pitch lag. The signal to be matched, referred to as the target
signal, is computed by filtering the LP residual through the weighted synthesis filter. The adaptive codebook index
is encoded with 8 bits in the first subframe and differentially encoded with 5 bits in the second subframe. The target
signal is updated by removing the adaptive codebook contribution, and this new target is used in the fixed
codebook search. The fixed codebook is a an algebraic codebook with 17 bits. The gains of the adaptive and fixed
codebooks are vector-quantized with 7 bits using a conjugate structure codebook.
Freescale Semiconductor
Excitation
Codebook
ITU-T G.729 Implementation on the StarCore™ SC140/SC1400 Cores, Rev. 1
Long-term
Synthesis
Filter
Figure 1. CELP Synthesis Model
Parameter Decoding
Received Bitstream
Short-term
Synthesis
Filter
G.729 Recommendation for Speech Compression
Filter
Post
Speech
Output
3