Voice Quality Enhancement Suite

VQE Suite Includes: Echo Cancellation, Acoustic Beamformer, Packet Loss Concealment, Automatic Level Control, and Noise Reduction software.

AVAILABILITY

ADT VQE suites are available on all Processors/Platforms including: TI TMS320C5000, TI TMS320C6000, ARM Devices, PC/Windows, and Linux.

DESCRIPTION

The key to integrating a VQE solution is to put the various algorithms together in such a way that maximizes speech quality enhancement. This figure shows one possible con-figuration for enhancing the quality of the speech signal sent from the near end to the far end.

The echo canceller core is the first algorithm to operate on the near end signal. The near end input to the echo canceller includes three components: near end speech (S), additive noise (N), and added echo (E) of the far end speech. The job of the echo canceller core is to reduce the echo component from the signal. The echo canceller core removes most of the echo, but leaves a small amount of residual echo (R). The echo canceller core therefore outputs a composite signal containing near end speech (S), additive noise (N), and residual echo (R).

The output of the echo canceller core is fed into the noise reduction algorithm, which increases the signal-to-noise ratio of the signal by reducing the background noise level. The output of the noise reduction algorithm ideally consists of near end speech (S) plus residual echo (R).

The output of the noise canceller is then fed into the Non-Linear Processor (NLP). The NLP's function is to suppress residual echo based upon estimates of the far end speech signal level and the residual echo signal level. It is advantageous to place the noise reduction before the NLP so the noise does not affect the estimate of the residual signal level. Furthermore, if the noise reduction were to follow the NLP, the NLP would at times suppress both the residual and the background noise. The resulting variations in noise level would cause a reduction in the performance of the noise canceller.

The output of the NLP is fed to the Automatic Level Control (ALC) function. The ALC function introduces the appropriate amount of gain or loss in order to create an output speech signal whose signal level is appropriate for good perceived speech quality.

Noise Reduction

A good noise reduction algorithm should:

reduce background noise under conditions where it exists. This effectively increases the signal to noise ratio of the speech signal.

have a minimum effect on speech signals.

be able to adapt to different conditions, but should also have   programmable parameters that control things such as:
threshold noise level, below which noise reduction is disabled.

maximum noise attenuation or some measure of noise reduction aggressiveness.

A good noise reduction algorithm should not:

degrade the speech quality, especially under conditions where there is little or no noise present.

introduce excessive delay.

not degrade the performance of fax or voiceband data modems   (network EC onl

AUTOMATIC LEVEL CONTROL

A good Automatic Level Control Algorithm should:

provide gain or loss to a speech signal while speech is present.

provide loss under very high signal level conditions.

disable itself in the presence of high speed voiceband modems.

have programmable parameters that control things such as:
 • Maximum gain
 • Maximum attenuation
 • Target level
 • Threshold speech level, below which gain is not applied.

A good Automatic Level Control Algorithm should not:

not cause perceptible artifacts in the speech signal.

not amplify noise when speech is not present.

distort signaling tones such as DTMF tones.

introduce excessive delay

ECHO CANCELLTION

A good echo canceller should:

meet or exceed the requirements set forth in ITU   recommendation G.168-2000 (network canceller) or G.167   (acoustic echo canceller).

have a robust double-talk detector to prevent divergence when   both parties are speaking at the same time.

provide deep echo cancellation (high return loss enhancement).

Include non-linear processing to remove residual echo

Include a comfort noise generator to prevent unwanted   transitions between silence and background noise.

disable cancellation in the presence of high speed voiceband   modems.

A good echo canceller should not:

not diverge in the presence of tones

not degrade the performance of fax or voiceband data modems (network EC only)

introduce excessive delay

Note: Echo Cancellation

There are two (2) types of echo - electrical echo and acoustic echo.

Consequently, there are two (2) types of echo cancellers - network (and line) echo cancellers and acoustic echo cancellers.

These echo cancellers have many similarities and some differences due to the nature of the different types of echoes.

• For more information on the Adaptive Digital Voice Quality Suite or other custom solutions, contact us at 610-825-0182.

Please call and discuss your requirements - 610-825-0182 x120 sales