Blueworx Voice Response for AIX Version 6.1 supports speech recognition and text-to-speech for VoiceXML applications using speech servers compliant with Media Resource Control Protocol (MRCP) V1.0, such as WebSphere Voice Server 5.1, Nuance Speech Server version 5.1.2 (Recognizer 9.0.13, Vocalizer 5.0.2), Loquendo Speech Server V7, and LumenVox 12.1.
MRCP can now be configured so that in the event of a configured MRCP resource becoming unavailable, a secondary back-up server can take over so that MRCP speech recognition and text-to-speech activity can continue. See How is the RecoService configuration entry used? and How is the TTSService configuration entry used?
Using dynamic engine allocation for speech recognition and text-to-speech is now also supported, so that speech recognition and text-to-speech engines can now be allocated only for the duration of each period of speech recognition or text-to-speech rather than for the duration of a call if required. See Speech Recognition for more information.
The Blueworx Voice Response for AIX MRCP client state table API provides a means by which Blueworx Voice Response state table applications can access MRCP text-to-speech synthesizer resources attached to the Blueworx Voice Response client over an IP network. See The Blueworx Voice Response MRCP state table API for more information.
In addition to the speech barge-in detection method, and the dtmf_only ‘proprietary’ barge-in detection method which stops audio output only after a user has pressed any DTMF key, Blueworx Voice Response also now supports hotword DTMF barge-in when using remote DTMF grammar detection. See Comparing barge-in detection methods for more information.
For SIP calls, it is now also possible to configure Blueworx Voice Response for AIX so that DTMF detection and DTMF grammar compilation are performed remotely by a speech server rather than by Blueworx Voice Response for AIX. See Remote DTMF grammars for more information.
Some proprietary, vendor-specific VoiceXML properties are also supported. Properties that match a given pattern can be passed through from a VoiceXML document then sent to a speech server in an MRCP SET-PARAMS message. See VoiceXML elements and attributes for more information.