Codan Radio P25 Training Guide - Manual (Page 105)

P25 RADIO SYSTEMS | TRAINING GUIDE
Chapter 8: IMBE™ and AMBE+2™ Vocoders Page 97
CHAPTER 8: IMBE™ AND AMBE+2™ VOCODERS
P25 radios use the Improved Multi-Band Excitation (IMBE™) vocoder, developed by Digital Voice
Systems, Inc. (DVSI), to convert analog speech into a digital bit stream suitable for transmission
over the P25 Common Air Interface (CAI). At the transmitter, the vocoder consists of an encoder that
converts the analog voice signal from a microphone into a digital bit stream, while at the receiver, the
vocoder consists of the decoder that converts the digital bit stream back into analog voice suitable for
playback through a speaker.
In P25, analog voice is converted into a digital bit stream with a net bit rate of 4.4 kbps for voice
information and a gross bit rate of 7.2 kbps after error control coding (note: after vocoding, 2.4 kbps of
signaling information is added to make 9.6 kbps total). The vocoder uses a frame size of 20 ms.
P25 selected the IMBE™ vocoder in 1992 after a competition with several other proposed vocoders.
All the vocoders were evaluated through an extensive set of Mean Opinion Score (MOS) tests that
compared voice quality for different male and female voices in a range of conditions. These conditions
included simulations of vehicles traveling at various rates of speed. In addition speech was tested with
various background noises, such as sirens, gunshots, and traffi c, that are likely to be encountered by a
public safety radio system. The result of this evaluation was that the IMBE™ vocoder was judged best
by a panel of listeners under almost every test condition. As a result the IMBE™ vocoder was selected
as the standard vocoder for the P25 system.
The IMBE™ vocoder is a model-based speech coder, or vocoder, that does not try to reproduce the
input speech signal on a sample-by-sample basis. Instead, the IMBE™ vocoder constructs a synthetic
speech signal that contains the same perceptual information as the original speech signal. The
IMBE™ vocoder is based on the Multi-Band Excitation (MBE) speech model that was developed at the
Massachusetts Institute of Technology (MIT) from research on high quality, robust speech modeling.
The IMBE™ vocoder models each segment of speech as a frequency dependent combination of voiced
(more periodic) and unvoiced (more noise-like) speech. This ability to mix voiced and unvoiced energy
is a major advantage over traditional speech models that require each segment of speech to be entirely
voiced or unvoiced. This fl exibility gives the IMBE™ vocoder higher voice quality and more robustness
to background noise.