If you are using a C# interface to the LumenVox software, callback functions using stdcall are no longer valid. Instead, the callback function passed into the Engine must be cdeclspec. This only applies to C# developments; users making use of the native C and C++ APIs are not affected.
In addition to specifying frequencies for dial tone disconnects, you may also now specify the length of the tones. Five new parameters have been added to LV_Platform.ini to offer finer control over how the Call Handler detects dial tones before disconnecting a call. See [DIAL_TONE_ENDS_CALL] for complete details.
Fixed a bug that was causing problems when using custom pronunciations with concept/phrase grammars.
The Call Handler will now correctly identify when Dialogic hardware has been changed, fixing some inappropriate behavior on some systems.
The Voice Activity Detection has been completely revamped. The Platform should now be more accurate at filtering out background noise, barging in correctly, etc. These changes, however, have caused us to alter several VAD parameters.
The following parameters have been removed: NoiseFloor, P2AThld, BargeInLevel, and EnableFreqVad.
Three new parameters have been added: NoiseSensitivity, VolumeSensitivity, and Sensitivity.
Please read the new [VADParams] section to ensure you are aware of the new values. These settings should require less configuration than the old settings, but in particular VolumeSensitivity is useful for configurations that have problems with echo cancellation.
All of the existing .ini files have been merged into the new LV_Platform.ini file
Increased the timeout value when dialing outbound calls using the Visual Basic LVCA. This value was previously too short, causing calls to be dropped prematurely.
When starting calls, the Call Handler was displaying an erroneous error message. This has been fixed.
The flash hook was not correctly working, but this has been fixed.
Fixed a bug that was causing LVCA_ClearAudioQueue to not work from the RPC interface.
Using the Platform Designer, you may now set the language for a project. Note that this only applies when using concept-phrase pair grammars. SRGS grammars loaded using the custom action interface will be decoded using the acoustic model specified in the grammar header. See Working With Languages for more details.
Added a new confidence for strings of digits that weights the overall confidence score for the set of digits down toward the lowest single score in the set. This score is called ADJUSTED_CONF_SCORE and is available via LVCA_GetCustomStringData.
Set the default P2A Threshold to 70 and set the default Scale to 50. You can control these values in the Stream.ini file.
Error messages that appear in the Call Handler screen will now be more verbose and generally be more useful.
Changed the default settings for prompt normalization to 40.
"Pure" silence in prompts can be now set to be a more "natural" silence to help with Dialogic's echo cancellation.
Changed the way recording audio to a file works so that recordings will now discard a very small amount of audio from the end to help eliminate button press sounds.
You can now use remote procedure calls to load custom action DLLs in a separate codespace from the Platform Call Handler. This makes applications using the Platform more robust (the DLL cannot affect memory in the Call Engine) and gives developers greater flexibility in debugging DLLs.
The Call Handler will now start faster as it is able to remember Dialogic hardware information from previous sessions, saving it from having to rediscover that information from scratch at each start-up.
You can now tell the Call Engine to end a call automatically when it detects a dial tone by using our new dial tone detection settings.
Added new call properties that allow you to control the end of speech, barge-in, and noise floor by using LVCA_SetCallProperty.
Added the property LVCA_KEY_PLATFORM_DIRECTORY that allows you get the location of the Platform's installation directory. You can get this string using LVCA_GetCustomStringData.
Added function LVCA_IsGrammarValid to the Platform's CAPI functionality. It allows developers to check whether their grammars are valid.
Improvements to the built-in date grammar have yielded an approximately 3% gain in accuracy tests.
Added new DTMF properties to SetCallProperty.
The "Settings" menu now has a button called "Voice Activity Settings" that allows direct access to edit the Stream.ini file which controls settings for voice activity. The Barge-In energy threshold option has been removed from the settings menu in favor of using the .ini file.
The Call Engine now looks at semantic confidence scores instead of raw text confidence scores.
Removed "Shutdown" menu option. To shutdown services, use "Immediate Shutdown."
Removed the Microphone Call Handler.
Added DTMF license functionality. Unchecking Use Speech Rec in the Line Settings menu will cause the engine to use a DTMF license instead of a speech recognition license.
Added support for branding of the title bar in the Call Handler. This is done through an .ini file. See The Platform.ini File for more information.
Fixed a bug that was causing the Line enable checkbox in the Line Settings menu to not be recognized.
Fixed a bug where the box containing the project to run in the Line Settings menu was not being cleared from the previous entry if no project had been set for the line selected.
Fixed a bug that may have caused applications to try closing invalid handles.
Complete Help Topic List | Speech Platform Product Information