LumenVox Speech Recognition Certified with UniMRCP Open Source Project

Interoperability provides accurate and affordable speech recognition and MRCP stack for voice platforms, media servers, and IP–PBXes

San Diego, CA (APRIL 27, 2009) — Speech recognition provider LumenVox and open-source MRCP project UniMRCP announced today the successful conclusion of interoperability testing between their products.

"The integration guarantees that anyone who uses UniMRCP libraries for their MRCP stack will also be able to use the LumenVox Speech Engine," said LumenVox CTO Jim Blake.

"UniMRCP provides a free method of adding MRCP support to any voice platform, gateways, or media server," said Arsen Chaloyan, the author and maintainer of UniMRCP. "Anyone wishing to add support for LumenVox speech recognition can do so easily and affordably."

MRCP, the media resource control protocol, is a standard method for different pieces of software to communicate. It is most commonly used for telephony systems such as voice platforms to control automatic speech recognition and text to speech applications. Nearly every modern ASR and TTS supports some form of MRCP, as do many voice platforms.

UniMRCP is an open source cross–platform project that provides the necessary libraries to build an MRCP client or server. Its typical users would be programmers who are working on a voice platform and wish to add support for MRCP.

"We are always looking to expand the number of platforms with which we are supported," Blake said. "Full integration with UniMRCP makes that even easier."

The LumenVox Speech Engine is an accurate and affordable speech recognition solution that includes support for both MRCPv1 and MRCPv2.

The two versions of the protocol are similar, except version one uses RTSP as its signaling protocol while version two uses SIP. LumenVox and UniMRCP support both versions, and both were tested as part of the interoperability certification.

UniMRCP encapsulates SIP/MRCPv2, RTSP, SDP and RTP stacks and provides an MRCP version independent user level interface, meaning platform providers can integrate the UniMRCP stack into their software and easily support both MRCPv1 and MRCPv2.

While UniMRCP is mainly a programming library to be used by developers adding MRCP support into an existing product, and not a standalone application, a simple standalone MRCP client is available as part of the downloads at the UniMRCP Web page. UniMRCP has its legacy in a previous open source project called OpenMRCP.

LumenVox provides a full–featured suite of tools along with its Speech Engine, including a graphical Speech Tuner for improving the performance of applications that use speech recognition, and a grammar editor for testing and validating the SRGS grammars used in speech recognition.

Both UniMRCP and LumenVox support Windows and Linux.

For futher information, contact us.