Aculab's echo cancellation meets the requirements of service providers, system integrators and developers, bringing superior voice quality to speech based applications . It is designed to expand and enhance the capabilities of TiNG and Prosody media processing resources. This ensures that Aculab continues to provide developers and integrators with the ultimate combination of speech processing capabilities, accessible through one board, one slot and one generic API.

What is echo cancellation?
Echo is created when a signal encounters an impedance mismatch in the telephone network, such as that typically caused by a two to four wire (hybrid) conversion in an analogue phone. The echo is exacerbated by distance and by certain kinds of network equipment. To eliminate echo, a cancellation device at the near end uses a complex adaptive digital filter on the transmitted signal that models the returning echo and cancels it by subtracting it from the returned signal (see diagram below).

Why do we need echo cancellation?
Echo can lead to undesirable detection effects. Words echoed from an outgoing prompt, for example, may be incorrectly recognised as having been said by the caller. In such an instance excessive echo can reduce the effectiveness of automatic speech recognition systems. A similar problem affects DTMF signals, especially when used within a multi-party conference, as the many separate impedance mismatches will produce multiple echoes, which reinforce each other, increasing both echo amplitude and dispersion.

In addition an echo canceller facilitates 'barge-in', where outgoing voice prompts may be interrupted by DTMF keypad entry or spoken responses. Otherwise the DTMF detector or speech recogniser would need to be disabled during prompt output in order to avoid false recognition.

Echo cancellation and TiNG
The firmware for echo cancellation is incorporated into the TiNG architecture. TiNG allows a developer to choose multiple functions, within the capacities of each DSP, to create the most cost effective communications solution. In most applications, TiNG can provide 24 channels of echo cancellation per DSP.

There are various options when using an echo canceller within TiNG. Echo cancellation can be used in stand-alone mode or it can be used in parallel with conferencing. Using echo cancellation with conferencing is important, for example, if the policy of the conference is to press '#' to exit the call. The echo canceller prevents the '#' tone being heard through the conference causing all other participants to be released.

A number of Prosody algorithms can be used to produce host based automatic speech recognition or DSP based tone recognition applications. Aculab's echo cancellation firmware is ideally suited to use in speech recognition systems. Using echo cancellation in parallel with simultaneous record, playback and DTMF detection on every channel allows channel feed information to be communicated to the host based speech recognition software. This also allows prompts to be played back on the same channel without generating false recognition results.

Cancellation is controlled by one simple API call, which initiates the echo canceller training sequence that adapts rapidly upon an optimal configuration to cancel echoes based upon current line conditions.

The echo cancellation algorithm fully conforms to G.168 specifications and even exceeds it in several tests. For example, convergence is significantly faster with the canceller fully adapting within 500ms and having improved attenuation in most instances. Aculab's echo cancellation firmware supports an echo path delay coverage of 25ms. Industry studies and reports indicate that this is more than adequate for almost all known applications within PSTN networks.

The echo cancellation firmware is available for Prosody PCI and Prosody cPCI boards and is selectable by integrators and developers to complement other functions at no extra cost. The software may be downloaded at any time via the Aculab installation tool (AIT) from software downloads.

Related links:

Product pages Prosody X PCI Prosody X cPCI Prosody cPCI Media processing resources Product portfolio Protocols and approvals Software downloads     Case study Viztel Varetis Media processing resources Conferencing Fax     Third party integrations Prosody integration with SkunkWorks Microsoft Speech Server

Print this page

Email to a colleague

Search our site