Communicator Demo / TCL & Communicator / Demonstrations / Software / Home

This page describes how to download and install a recognition server based on our prototype system that integrates with the DARPA Communicator client/server architecture. This server decodes audio data and displays the resulting hypothesis in a text window. A demo program is included that makes it easy to get started. There are five easy steps to Communicator bliss: Step 1: Install the DARPA Communicator system. Step 2: Install the ISIP Prototype System. Step 3: Install the recognition demo server. Step 4: Run the demo. " Step 5: Modify the hub program file. Step 1: Install the DARPA Communicator system In order to run this server, one must have previously installed the DARPA Communicator system and be somewhat familiar with its operation. Step 2: Install the prototype system Download the latest version of the ISIP prototype system in which we provide an integrated API for the DARPA Communicator system. The system can be easily installed following the steps below: tar xzvf isip_proto_v5.9.tar.gz cd isip_proto ./configure --prefix=. source ISIP_ENV.sh make make install In addition to installing the recognition binaries, this step builds a library containing the API that will be linked into the applications described below. Step 3: Install the recognition server demo Once the DARPA Communicator and ISIP prototype systems have been successfully installed, we can download the recognition server demo (Solaris x-86 and linux) or recognition server demo (Solaris sparc) and proceed with building the server program: tar xzvf rec_demo_v2_0.tar.gz ln -s rec_demo_v2_0 rec_demo cd rec_demo ./configure [--prefix=Communicator install directory] source GC_ENV.sh make Note that the --prefix option is used to specify the directory in which you installed Communicator. If you have set the Communicator environment variable, GC_ROOT, you do not need to use the --prefix option. The default directory for the Communicator installation is /usr/local/communicator. Before compiling the recognition server, make sure you have sourced the ISIP environment (source ISIP_ENV.sh) to establish the proper run-time environment for the prototype system. Step 4: Run the application demo Steps 1 to 3 above must have been successfully completed for the next step to work. The recognition demo program can now be invoked as follows: ./recognizer.csh Before running the recognition server, make sure you have sourced the Communicator environment in step 3 (source GC_ENV.sh) to establish the necessary environment. Now you are finally ready to run an audio server that will send data to the recognizer in batch mode: ./demo.csh The audio demo server reads data from a file and sends it to the recognition server. The recognition server decodes the audio data and displays the hypothesis at the end of the utterance in a text window. Step 5: Modify the hub program file The demo_pgm.text can be easily modified to operate in your local environment: ;; ------------------------------------------------- ;; SERVERS ;; ------------------------------------------------- SERVER: recognize_serv ;; HOST: 130.18.6.111 HOST: localhost PORT: 12346 OPERATIONS: reinitialize recognize_speech ;; ------------------------------------------------- ;; CONDITIONS (default order) ;; ------------------------------------------------- RULE: :binary_port & :binary_host & :call_id --> recognizer_speech IN: :binary_port :binary_host :call_id OUT: none! The tag recognize_speech handles incoming audio broker requests, decodes the received audio data and outputs the hypothesis. Currently, the recognizer expects the broker to send GAL_INT_16 or GAL_BINARY data and expects the audio server to send the following control messages of type GAL_STRING: "new_utt": audio is about to be sent "done": no more samples will be sent (end of utterance) Note: A demonstration system based on a real-time Resource Management system is also included in the final_model/ directory of the demo release. The acoustic models for this system were trained on 16kHz sampled data. Hence, it is important that the Audio server sends compatible sampled data. Success! Kick back, smoke a victory cigar, and recite George Peppard s famous line from the A-Team: "I love it when a plan comes together."