Echo Cancellation for ASR Applications

Automatic Speech Recognition (ASR) is rapidly becoming a critical component of the New NetworkTM. The explosion in mobile communications and vastly improved technology finally make ASR a compelling feature of today's enhanced services offerings. Its ability to deliver what you want, where and when you want it, makes ASR an ideal complement to a variety of value-added services. In applications ranging from traditional IVR and auto-attendants to speech-enabled unified communications and messaging systems, an accurate, easy-to-use ASR system can help maintain customer loyalty, increase talk time revenues, and reduce subscriber churn.

However, users demand systems with highly accurate recognition capabilities and this requires careful attention to echo cancellation. Today's advanced ASR applications allow users to "barge in" during a prompt, providing a more responsive, efficient interactive session. Unfortunately, even small amounts of echo from the outgoing prompt can destroy recognition accuracy. Acceptable performance can usually be achieved using the onboard echo cancellation provided on DSP-based telephony boards. However, echo problems discovered during system installation can be expensive to resolve making enhanced long-tail onboard echo cancellation a welcome option.

Echo cancellation problems in ASR systems are not mere annoyances; they can destroy the accuracy, usability and value of an ASR system, as well as directly reduce customer satisfaction. That is why developers should give additional thought to ensuring that echo cancellation performance is effective under all anticipated circumstances.

This white paper will identify special considerations for echo cancellation in ASR systems, examine how echo is created, and describe approaches to eliminating it. This paper ends with a discussion on the need for echo in ASR systems.

View Entire Paper | Previous Page | White Papers Search

If you found this page useful, bookmark and share it on: