Certified Interoperability of Loquendo MRCP Server with UniMRCP Connector Bridge.

Loquendo, a leading provider of speech technologies worldwide, and UniMRCP, an open source MRCP project, announce the interoperability of the Loquendo MRCP Server (in its version LSS 7.0) with the Asterisk open source IP telephony platform via the UniMRCP Connector Bridge. The integration is based on MRCP (Media Resource Control Protocol), the widely adopted IETF protocol.

Loquendo MRCP Server supporting MRCP v1 and v2, both for Windows and Linux, is a server solution for speech-enabling large-scale telephony applications, such as contact centers and message reading services. It enables enterprises to significantly reduce costs by interacting with customers via a speech interface - reliable, natural and intuitive to use.

UniMRCP is an open source cross-platform MRCP project, which provides everything required for the implementation and deployment of both an MRCP client and an MRCP server. UniMRCP encapsulates SIP/MRCPv2, RTSP, SDP and RTP/RTCP stacks and provides integrators with an MRCP v1 and v2 user level API.

Loquendo has certified the interoperability of its ASR and TTS, integrated via Loquendo MRCP Server, with the UniMRCP Connector Bridge. Loquendo speech technologies are thus available for enriching and simplifying large-scale telephony deployments on the widely adopted Asterisk Platform, such as IP PBX systems, VoIP gateways, conference servers, etc.

The interoperability with UniMRCP enables integrators to exploit the whole range of Loquendo ASR and TTS functionalities by means of a client-server architecture - flexible, standards-based, supporting multiple operating systems, and so greatly reducing overheads for your customer in terms of hardware investments and maintenance by hosting speech resources on a dedicated server.

“I am always looking to expand the list of MRCP vendors the UniMRCP project is known to work with. The successful interoperability with a speech technology leader such as Loquendo is a very valuable event for the UniMRCP community. Loquendo's voices sound natural, and the recognition results are always accurate,” says Arsen Chaloyan, author of UniMRCP.

“The UniMRCP project is an important initiative which has proved itself to be of invaluable assistance to speech technology providers and integrators of telephony solutions,” says Roberto Pacifici, Product Manager for Loquendo MRCP Server. “This integration makes Loquendo speech technologies, in all 30 languages, available for the Asterisk environment enabling rapid deployment of custom solutions for large-scale, speech-enabled telephony deployments for call centers, self-service applications, auto attendants, and much more.”

See the UniMRCP Connector Bridge press release.

 

About UniMRCP
An open source MRCP Project: www.unimrcp.org
Contact Arsen Chaloyan: arsen.chaloyan@unimrcp.org

 

About Loquendo
Awarded Speech Industry ‘Market Leader’ for the past four consecutive years, Loquendo provides a complete range of speech technologies for server, embedded and desktop solutions – in 30 languages with 72 voices, and constantly growing.

Loquendo TTS, Loquendo ASR and Loquendo Speaker Verification empower people to interact with technology in the most natural way possible – using their voice – creating a next-generation client experience while saving businesses millions each year. Also integrable via the Loquendo MRCP Server and VoxNauta VoiceXML, CCXML & SCXML platform, Loquendo speech technologies power millions of calls every day in the telecommunications and enterprise markets across the globe. Loquendo TTS and ASR are also available as Loquendo Embedded Technologies, deployed in more than 14 million devices worldwide in embedded and mobile environments.

Loquendo is a Telecom Italia company headquartered in Turin, Italy, with offices in the US, UK, Spain, Germany and France, and a global network of partners. For more info, and to hear Loquendo TTS for yourself, go to www.loquendo.com.

I would like to announce the availability of the speaker verification and identification resource for the UniMRCP project.

Speaker verification is a voice authentication methodology that can be used to identify the speaker in order to grant access to sensitive information and transactions. In speaker verification, a recorded utterance is compared to a previously stored voiceprint which is in turn associated with a claimed identity for that user.

Speaker identification is the process of associating an unknown speaker with a member in a population. It does not employ a claim of identity.

This addition allows applications utilizing the UniMRCP client stack to use verifiers and identifiers residing on the MRCPv2 compliant servers. Also, verification and identification engines can be integrated into the UniMRCP server as plugins.

In the meantime, a typical verification scenario, integrated into the UMC client application as well as a simulated, demo verification engine, plugged in the UniMRCP server have already been implemented and available in the trunk. To initiate a sample verification scenario from the UMC client application, do the following

- upgrade to the trunk (r1778)
- upgrade your current configuration (unimrcpclient.xml, unimrcpserver.xml and umcscenarios.xml may need to be upgraded) or just use the default configuration files
- input "run verify" from the UMC console:


Thanks for using UniMRCP.

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

I would like to announce the new release of UniMRCP connector bridge for Asterisk.

The connector bridge is prepackaged with the latest Asterisk-1.6.2.9 and UniMRCP-r1744 (> 1.0.0). However, previous versions of Asterisk and UniMRCP are supported as well. This release contains several enhancements in both res_speech_unimrcp and app_unimrcp modules.


Changes in res_speech_unimrcp include

• Made an enhancement to SpeechLoadGrammar to be able to specify an input grammar as a URI too. (Raymond)
• Fixed compilation of res-speech-unimrcp module for Asterisk 1.4.
• Fixed processing of Set-Input-Timers header field.
• Set an interpreted result based on the <instance> element instead of the <input> one.

Changes in app_unimrcp include

• Changes required for version 1.2 of Asterisk (Issue-64, Igor, Derik)
• Added missing '{' to compile with the released UniMRCP version too. (Issue-65, Igor)
• Bug fix to check if codec descriptor could be obtained (Derik)
• Added support for ABNF grammar (Issue-76, Assanta, Derik)
• Bug fix to speech_channel_destroy (Issue-72, Assanta, Derik)
• Addition of request-timeout configuration parameter (Derik)
• Bug fix to address issue 80  - checking for speech channel state while waiting for audio frames in MRCPRecog (Assanta, Derik)
• Removed incorrect check for resf which was fixed at -1 anyway (Assanta, Derik)
• Added SYNTHSTATUS and RECOGSTATUS variables so that problems can be detected in the dialplan (Assanta, Derik)

The released package can be downloaded from
http://unimrcp.googlecode.com/files/uni-ast-package-0.3.0.tar.gz


For the installation, configuration and usage please refer to the wiki page
http://code.google.com/p/unimrcp/wiki/asteriskUniMRCP


Thanks for using UniMRCP.

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

UniMRCP announces the successful interoperability between its MRCP client and VoiceNavigator, an IVR solution developed by STC. Comprehensive certification tests guarantee excellent and fully compliant operation between products utilizing the UniMRCP client stack and VoiceNavigator.

VoiceNavigator is a cutting-edge IVR solution especially designed for the Russian language. It is based on powerful and reliable speech recognition and synthesis technologies developed by STC, the world leader in speech technologies for Russian.

Recognized as the “best product of the year – 2010” at the Moscow CCWF-2010, STC has got many requests from Asterisk users. In order to meet the customer demand, STC has ensured full interoperability between VoiceNavigator and popular open source softswitches such as Asterisk and FreeSWITCH using  UniMRCP’s unique and comprehensive set of connectivity options.

For more information, you may also check the following:

http://speechpro.ru/media/news/2010-06-21
http://igorg.ru/
http://asterisk.ru/

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

I'm proud to announce the first major release of UniMRCP 1.0.0 r1725.

This release is a result of the development continuously lasting more than two years. The open source initiative works. There have been a number of successful deployments since 0.7.0 release. I cannot recall any major API change or an outstanding issue encountered meanwhile. All these make me think it is the time for 1.0.0 now.

On the other side, this is just another recurrent release, which as usual introduces several enhancements and fixes. The major enhancement is in the support of transparent header fields. While the old API to set/get the header fields by numeric identifiers remains intact, the new methods are added to manipulate with the header fields using string identifiers.

Changes since previous release 0.10.0 r1577 include

• Set the length of the processed line even if it's not properly terminated (Issue-77, Anthony)
• Cancel an MRCP request sent by the application, if there is no MRCPv2 connection established from the client to the server.
• RTCP reports could wrongly indicate 1 lost packet, when there were no RTP packets received.
• Enhanced umc application to demonstrate how to match a request with its response.
• Added an ability to STOP current request from the command line of umc application.
• Added the number of discarded and ignored packets to "Close RTP Receiver" trace. (Anthony)
• Supported transparent header fields. (Vlad)
• Supported white spaces in the header fields, where WSP = SP / HTAB.
• Supported line folding in the header fields (a record spanning multiple lines).
• Used the "const" qualifier where applicable.
• Implemented an option to mask private data in the logs. (Issue-81, Randy )
• Fixed re-introduced message segmentation related issues. (Issue-86, Anthony)
• Modified multipart content generation and parsing to support content-id as well as other arbitrary header fields included in an individual content part. (Anthony)
• Added new methods to task interface to be able to handle start-request and terminate-request events.
• Made open/close methods of the engines/plugins asynchronous. (Anthony)
• Fixed a header field creation from the entire text line. (Issue-88, Vlad)
• Added a new method to codec interface to initialize or fill the specified frame with silence.
• Ensured a media frame read out of the null bridge to be always initialized. It should be filled with the silence at least.
• Initialized the resource location attribute with an empty string, if <resource-location> element is empty.
• Revised several debug traces.
• Added an option to set an informative name to the session being created by the client application. This name is passed along to the other objects created in the scope of the same session and is used for debugging.
• Provided the identifiers of the objects upon creation. Typical objects are engines, agents etc. This allows to further track those objects during their lifetime.
• Added the copyright and the license related information to uni_version.h; made uni_version.h independent from apr_version.h.
• Added Windows resource files for the following applications: umc, unimrcpclient, unimrcpserver. Those resources files include  version, license and copyright related information. (Issue-83, Patrick)
• Added missing dependencies to be able to build the projects from the command line or using the IDE with disabled "Link Library Dependencies" option. (Issue-84, Patrick)
• Upgraded the dependency package to include the recent APR-1.4.2 version.

You may stick with the older versions, if needed; but of course, you are all are encouraged to upgrade.

• Source Distribution

          http://unimrcp.googlecode.com/files/unimrcp-1.0.0.tar.gz
          http://unimrcp.googlecode.com/files/unimrcp-1.0.0.zip

• Windows Installers (32-bit and 64-bit)
  

          http://unimrcp.googlecode.com/files/unimrcp-1.0.0.exe
          http://unimrcp.googlecode.com/files/unimrcp-x64-1.0.0.exe

• Windows SDK Installers (32-bit and 64-bit)
  

          http://unimrcp.googlecode.com/files/unimrcp-sdk-1.0.0.exe
          http://unimrcp.googlecode.com/files/unimrcp-x64-sdk-1.0.0.exe

• Dependency Packages (based on APR-1.4.2, APR-Util-1.3.9,  Sofia-SIP-1.12.10)

          http://unimrcp.googlecode.com/files/unimrcp-deps-1.0.0.tar.gz
          http://unimrcp.googlecode.com/files/unimrcp-deps-1.0.0.zip


Thanks for using UniMRCP.

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

I want to explore in more details one of the basis ideas, which is getting more and more mature. The idea is simple and clear enough to build and/or launch a public MRCP server. The word "public" has  the following meanings in this scope.

1. Multilingual and Nationwide MRCP Server
As you know, I have provided an interface to plug 3-rd party TTS and ASR engines into UniMRCP server. I may only wonder how many plugins have already been written and exist nowadays. To recall, there have been contacts from all the continents from East to West and South to North. If we could unite all these, it would be indeed a unique solution, which speaks and recognizes so many languages.

2. Globally Accessible MRCP Server
The mentioned above solution can also be globally accessible from everywhere in the world provided as a software as a service (SaaS). While MRCP servers are usually deployed with MRCP clients located on the same LAN and this is still a typical approach, there are many use cases, where a publicly available MRCP server can be quite useful. Note that there are a few MRCP aware devices today and there will be more tomorrow.


Although we have almost all the required components now and an ability to turn this undertaking into a truly success, there is still a certain way to go.
I'm looking forward to a robust and mutually beneficial partnership with TTS and ASR vendors which might be interested in this idea. Feel free to express your thoughts either publicly or contact me off-list.

Thanks for using UniMRCP.

--
Arsen Chaloyan
The author of UniMRCP
arsen.chaloyan@unimrcp.org
http://www.unimrcp.org

UniMRCP turns 2 during the days.

Consistency, stability and overall reliability are what I'm usually trying to achieve not only in this project, but in an everyday life. Time has proved how much I was right creating this project. UniMRCP has been evolved probably a bit slowly, but consistently and reliably, welcoming everyone interested in the project and bypassing possible troubles we all meet from time to time in our life.

Overall integrity and stability of the software, current level of support, a number of developed solutions and readiness for commercial deployments make management of the project more easier and predictable now.

What to expect from the project in the nearest future.
I'll roll out 1.0.0 release in a month, most probably by the end of May. No new major features will be added in the meantime. It'd be better to stop for a moment and polish what we have for now. But don't worry, new features will follow up thereafter in Q2.

What I expect from all of you.
First of all, I need your help in all what the development of an open source project incurs. Commercial support options are becoming more and more actual. This is probably the most viable and mutually beneficial option I have seen so far. Also, I'm looking forward to strengthen the relationships basically established with companies located all over the world. There have been contacts mostly from US, a few from Europe and I'm glad to notice increasing activity from Russia recently.

In addition, I have a few items on the list to discuss with you. These items include a major and a few minor, but still important to me tasks, which I'm trying to sort out now probably for Q3 and so on. I'll follow up with separate posts next week.

Please stay tuned and thanks for using UniMRCP.
--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

Advanced Voice Solutions Support for the Telephony Industry
Acapela Group Announces Full Support for MRCPv2

 

Acapela Speech Synthesis Powered by UniMRCP Open Source Project

Acapela Group, a leading Voice expert, and UniMRCP, an open source MRCP project, are announcing the release of Acapela MRCP add-on 2.000, an extension of Acapela TTS for Windows and Linux servers, providing customers with a greater choice of high-end speech solutions using MRCPv2.

Telephony providers can now benefit from Acapela’s’s high quality and pleasant voices while relying on MRCPv2 for optimized integration of speech into their services and applications.

MRCP (Media Resource Control Protocol) allows telephony applications to communicate with speech resources, controlling media processing resources over the network using a distributed, client/server architecture. The main media processing resources specified by the MRCP standard are the Speech Synthesizer (TTS), Speech Recognizer (ASR), Speech Recorder (SR), and the Speaker Verifier (SV).

UniMRCP is an open source cross-platform MRCP project, which provides everything required for the implementation and deployment of both an MRCP client and an MRCP server. UniMRCP encapsulates SIP/MRCPv2, RTSP, SDP and RTP/RTCP stacks and provides integrators with an MRCP v1 and v2 user level API.

”We are very happy to announce the compliance with MRCPv2 of our Acapela TTS Server solution realized in cooperation with UniMRCP Open Source Project, which has been an outstanding partner in this process. Acapela Group, as a voice expert dedicated to innovation and finding the best possible voice solutions, fully supports market standards to ease and optimize text to speech use in telephony applications. The full interoperability of Acapela’s speech engine with MRCPv2 is part of the company’s commitment to support the industry and help optimize the easy test and deployment process”–comments Lars-Erik Larsson, CEO of Acapela Group.

“I am proud to announce the availability of Acapela TTS voices, based on the UniMRCP server software library. The number of languages and set of high quality voices Acapela supports are quite impressive. Through the integration with the UniMRCP server, Acapela MRCP add-on 2.000, makes its TTS engine available to the users of any MRCPv2-aware platform. This solution might be an ideal choice to be used in conjunction with well-known open source, as well as commercial products for the deployment of IVRs, Call Centers, IPPBXs,” said Arsen Chaloyan, author of UniMRCP.

The Acapela MRCPv2 add-on is available on both Acapela TTS for Windows and Linux servers. It enhances Acapela’s capability to fully respond to the telephony market’s requirements and reinforces the company’s position as an unequalled partner in the domain of computer-based voice processing. Acapela’s expertise and commitment to quality and service covers all aspects of voice, from the ease of integration and deployment to the delivery of custom voices, providing unique voice personas for companies or brands and- guaranteeing highest customer satisfaction and affinity.

About UniMRCP:
Open Source MRCP Project,  www.unimrcp.org
Contact Arsen Chaloyan,  arsen.chaloyan@unimrcp.org

About Acapela Group
Acapela Group, the leading European voice expert, invents text to speech solutions to give your content a voice in up to 25 languages. Our speech solutions allow you to turn any written text into natural speech files, using any of our 50 High Quality standard voices or your own synthezised voice talent. Acapela can answer all text to speech needs and provide perfect vocalization whether for voice integration and development, online & on demand use, audio files production, or ready to speak products for personal accessibility use. http://www.acapela-group.com

Check out http://www.acapela.tv - the sparkling laboratory of Acapela Group, for a fresh and exciting take on how speech synthesis can now be used: acapela.tv is a showcase and test playground that shares the possibilities of speech synthesis with Internet users, allowing them to discover innovative and attractive ways to make smart use of speech applications online.

Contact Caroline Houel, caroline.houel@acapela-group.com

UniMRCP 0.10.0 r1577 has been released and now is available for download.

• Source Distribution 

          http://unimrcp.googlecode.com/files/unimrcp-0.10.0.tar.gz
          http://unimrcp.googlecode.com/files/unimrcp-0.10.0.zip

• Windows Installers (32-bit and 64-bit)
  

          http://unimrcp.googlecode.com/files/unimrcp-0.10.0.exe
          http://unimrcp.googlecode.com/files/unimrcp-x64-0.10.0.exe

• Windows SDK Installers (32-bit and 64-bit) 

          http://unimrcp.googlecode.com/files/unimrcp-sdk-0.10.0.exe
          http://unimrcp.googlecode.com/files/unimrcp-x64-sdk-0.10.0.exe

• Dependency Packages (based on APR-1.3.8, APR-Util-1.3.9, Sofia-SIP-1.12.10) 

          http://unimrcp.googlecode.com/files/unimrcp-deps-0.9.0.tar.gz
          http://unimrcp.googlecode.com/files/unimrcp-deps-0.9.0.zip


This release introduces revamped configuration format. The new format allows to use profiles in more flexible way. Also, it addresses XML conformance issues. Corresponding XML schemas are available as well. Please upgrade your local configuration files as the old format is no longer valid.

The other major addition is the implementation of request timers. It aims to address a potential vulnerability of the UniMRCP client stack in case there is no response back from an MRCP server. RTSP, MRCPv2 request timeouts as well as several SIP timeouts are configurable now.

Yet another interesting addition is an utility tool called svnrev, which allows to generate uni_revision.h header file containing SVN revision number and build stamp. Although this tool has basically been integrated into VS and GNU build, it's experimental and probably informative only yet.

Changes since previous release 0.9.0 r1383 include

• Enhanced MRCP, RTSP parsers to be able to process messages with headers, which don't entirely fit the read buffer. The issue encountered on processing of a huge number of vendor specific parameters included in the response to GET-PARAMS request. (Issue-57, Vlad)
• Disabled simultaneous termination of tasks. Now tasks are terminated one by one. (Issue-56, Derik)
• Fixed apt_pair_parse() to copy name/value strings from the stream.
• Fixed possible stack overflow on the construction of log entries. Log entries are limited with 4096 bytes.
• Added a sample scenario to umc application to show how SET-PARAMS and GET-PARAMS methods should be used.
• Fixed RTSP client stack crash while processing the response to final TEARDOWN request. (Issue-60, Chris)
• Provided an option to adjust size of buffers used to receive/transmit MRCPv2 messages. (Issue-57, Vlad)
• Added the main page for doxygen generated documentation. (Vali)
• Implemented an option to trace current thread identifier in the logs. (Anthony)
• Checked playout delay to be codec frame aligned. (Shy)
• Fixed remote address in the trace for "Close RTP Transmitter".
• Revised C/C++ header guards not to violate naming convention of the C standard. (Issue-62, Markus)
• Added SVN keyword Id to all the source files to be automatically expanded on each commit with the revision number and stamp of the file.
• Fixed access violation while reading server config. (Issue-66, Vali)
• Enhanced detection of Flite configuration. (Issue-67, nombrandue)
• Added logger.xml configuration file to load the settings of logger from.
• Added missing mpf_codec_open() and mpf_codec_close() calls, which might be required for integration of new codecs. (Shy)
  
• Added multipart content generator and parser.
• Enhanced asrclient library built on top of libunimrcpclient to provide streaming interface as an alternative to file input.
• Introduced generic timer queue and poller task.
• Removed no longer referenced net client and server tasks.
  
• Fixed UniMRCP server crash in case there is no MPF engine configured for the MRCPv1 profile. (Issue-74, Anthony)
• Increased XML_FILE_BUFFER_LENGTH to ensure configuration files fit the buffer.
• Revised task and inter-task communication related debug traces.
  

Thanks for using UniMRCP.

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

I would like to announce the second release of UniMRCP connector bridge for Asterisk.

The connector bridge is prepackaged with Asterisk-1.6.1.14 and UniMRCP-r1456 (> 0.9.0). These are the recommended versions to use, though other versions are also supported.

This release contains a few enhancements in speech resource module (res_speech_unimrcp), which implements the Generic Speech Recognition API of Asterisk.
* Added internal configuration file (/etc/asterisk/res-speech-unimrcp.conf) with further reference to UniMRCP client configuration.
* Supported various recognition grammars including builtin grammar, text/uri-list, multipart/mixed and everything else.
* Added an option to preload grammars for a channel
* Added an option to set properties (MRCP header fields) for a channel

It also introduces a new and great application module (app_unimrcp) gracefully contributed and actively maintained by J.W.F. Thirion <derik@molo.co.za> Molo Afrika Speech Technologies (Pty) Ltd.

Asterisk dialplan applications are:
* MRCPSynth - text to speech,
* MRCPRecog - speech recognition


The released package can be downloaded from
http://unimrcp.googlecode.com/files/uni-ast-package-0.2.0.tar.gz

For the installation, configuration and usage please refer to the wiki page
http://code.google.com/p/unimrcp/wiki/asteriskUniMRCP

Feel free to participate in the discussions with the questions and issues you may have
http://groups.google.com/group/unimrcp


Thanks for using UniMRCP,

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org