Acoustic Echo Cancellation - Polycom SoundStructure C16 Design Manual

Hide thumbs Also See for SoundStructure C16:
Table of Contents

Advertisement

then be transmitted to the remote site along with the local talker's audio signal and the reduced
signal-to-noise ratio will contribute to lowered intelligibility of the remote audio and increased listener fatigue
for the remote talkers.
It is best to eliminate or at least reduce the ambient noise through architectural means such as changing
HVAC ductwork, moving microphones away from noise sources, and removing or dampening noise
sources. If these approaches are not adequate or possible, an additional option is to process the
microphone signal with advanced signal processing techniques that reduce the level of background noise
while maintaining the quality of the local talker's voice.
Techniques for reducing the background noise picked up by the microphone range from simple noise gates
to advanced digital adaptive filters. Noise gate techniques will reduce the noise when the local talkers are
not talking by suppressing the signal that is below a given threshold, but the noise will still be present when
the local talkers begin to speak again. The gating of the background noise will sound unnatural at the remote
site as the local talker speaks and then stops speaking.
More sophisticated techniques such as adaptive filter techniques are used quite successfully in audio
conferencing applications. While not all adaptive noise reduction techniques (commonly referred to as noise
cancellation) have the same performance, the objectives are the same - to first identify the characteristics
of the noise (broadband such as HVAC noise, or narrowband such as a whine from a mechanical source)
and then remove that noise signal from the microphone audio signal without any additional information
about the noise. These techniques work best with noise that has stationary statistics -for instance, the noise
signals may be random, but the style of randomness is fixed such as the noise from a fan source. As these
techniques typically take several seconds to identify the characteristics of noise, these techniques do not
work well with impulsive noises such as clicks from pen tapping or paper rustling on a microphone. These
systems are typically designed to work with speech signals and are not usually suitable for use with music.
As not all implementations are the same, there can be a large variation in the amount of residual noise or
spectral artifacts that are introduced into the processed signal. These artifacts can sound like chirps or
worse and may be perceived to be worse in quality than the original noise. These artifacts may be minimized
by lowering the amount of noise cancellation provided - typically it can be adjusted from 0 to 15 dB or more.
With current techniques 5 -10 dB of noise cancellation can be achieved without significant distortion of the
underlying local talker's signal (depending on the manufacturer).
Ceiling microphones benefit the most from noise cancellation techniques as these microphones are closest
to the ceiling noise sources of HVAC and projectors. The noise cancellation can make an otherwise useless
room usable. However, if ceiling microphones are swaying due to the air flow from nearby HVAC ducts,
noise cancellation may not be able to completely remove that noise.

Acoustic Echo Cancellation

In audio conferencing applications, acoustic echoes occur because an open-air acoustic path exists
between the local loudspeaker and the local microphone. As shown in the following figure, speech
originating in either the local or remote room is transmitted over a communications network to the other
room where it is amplified and reproduced by the local loudspeaker. The output of the loudspeaker (or
multiple loudspeakers) will fill the local room and, from many paths of reflections, reach the microphones in
the local room. In the absence of an acoustic echo canceller, this acoustically-echoed version of the remote
talker's audio is transmitted back over the network to the originating room and is reproduced by the
loudspeaker where it is perceived as an acoustic echo.
Acoustic echoes degrade the quality of speech communications because echoes of one's speech are
subjectively annoying to the person speaking. In fact, if the elapsed time between when a word is spoken
and when its echo is heard is more than 300 msec, the echo will actually cause most talkers to stutter. In
Polycom, Inc.
723

Advertisement

Table of Contents
loading

This manual is also suitable for:

Soundstructure c8Soundstructure c12Soundstructure sr12

Table of Contents