AN2094 Freescale Semiconductor / Motorola, AN2094 Datasheet - Page 2

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

G.729 Recommendation for Speech Compression
This application note is written for SC140 programmers, system engineers, tool developers, and project managers.
1
Speech compression technology is widely used in digital communication systems such as wireless systems, VoIP,
and video conference technology. Speech compression reduces data redundancy and thus eases bandwidth
requirements. The compression technique described in the ITU-T G.729 Recommendation is commonly employed
in speech transmission systems because of the quality of the reconstructed speech signal.
1.1 Assessing Speech Quality
The Mean Opinion Score (MOS) is a commonly used test to assess speech quality. In this test, listeners rate a coded
phrase based on a fixed scale [2, 3, and 4]. A MOS of four or higher is considered ‘toll’ quality, which means that
the reconstructed speech is indistinguishable from the original speech. Tests have shown that encoding systems
based on G.729 at 8 kbits/s provide toll-quality speech for most operating conditions, as shown in Table 1
7]
.
1.2 Technical Overview of ITU-T G.729
The International Telecommunications Union–Telecommunications Standardization Sector (ITU-T) G.729
Recommendation defines an algorithm for coding speech signals at 8 kbit/s using Conjugate-Structure Algebraic-
Code-Excited Linear-Prediction (CS-ACELP). In this system, an analog voice signal is passed through a
300 Hz – 3400 Hz bandpass filter and sampled at 8 kHz to yield digital data that is converted to a 16-bit linear
PCM speech signal. An encoder analyzes the speech signal to extract the parameters of the CELP model. These
parameters are encoded and transmitted in a bitstream. The decoder for this system uses the received parameters to
retrieve the synthesis filter coefficients. The speech is then reconstructed by filtering the excitation codebook as
shown in Figure 1. The vocoder operates on 10 ms frames with 5 ms look-ahead for linear-prediction (LP)
analysis. The overall algorithmic delay is 15 ms.
2
.
•
•
G.729 Recommendation for Speech Compression
Stack optimization for improved multi-tasking
Peak performance of 1200 DSP-MIPS at 300 MHz
Clean
Encoder Background Noise
Channel Errors
Note:
ITU-T G.729 Implementation on the StarCore™ SC140/SC1400 Cores, Rev. 1
Modulated noise reference unit
Test Conditions
Table 1. G.729 MOS Results
10 dB MNRU
30 dB MNRU
1.0% bit error
0.1% bit error
1
4.125
3.975
MOS
3.65
3.95
3.3
Freescale Semiconductor
[5, 6, and