readme
1. quickly understand speech SDK easiest method call, you can look directly at the sample code (if necessary Examples ), speech synthesis (text into voice) See ttsdemo , speech transcription (Speech to Text) See iatdemo ;
2. systematic understanding of the voice SDK See DOC "Development Manual" and "list of parameters that can be set" folder, if you want to understand the development of the recognition grammar, please refer to "Syntax Development Guide";
3. If your development environment vs2005 or VS2008 , prompt "Unable to run msc.dll ", please Project Properties -> Debug -> working directory Add msc.dll file path.
If you are using vs2010 or vs2012 , you can skip this step;
4. Open correspond to your environment sln solution, you can see a few examples of projects listed in the Solution Explorer, when you need to debug a project, right-click the project name, select "Set as StartUp Project "; if you do not like a solution that contains multiple projects, you can go to the Examples folder, enter the need to debug the project folder, open your development environment corresponding project file ( VS2005 and vs2008 correspond vcproj file, VS2010 and vs2012 correspondence vcxproj file), if you do not understand the role of various documents, it is recommended not to make any changes;
The project itself contains WIN32 and x64 configuration for both platforms, you can directly select the Configuration Manager also cited the corresponding dynamic library code based on the platform of choice, of course, the premise is you vs the corresponding installation platform compiler;
6.bin folder is a debugging path, the role of which file is as follows:
msc.dll is WIN32 DLL platform program required;
msc_x64.dll is x64 dynamic library platform program required;
msc.cfg contains configuration information, debug, without changes to the test, the product can be released, "the fifth row of the inverse of the Output = 1 " was changed to " the Output = 0 " , " the Output = 1 " will run when the output to the log msc .log , debug problems can occur when you send us help analyze, " the output = 0 turns off log output" for the user to save disk space;
asr_keywords_utf8.txt is asrdemo word list of commands used in encoding is UTF8 , when modify its contents, please do not use the system comes with Notepad program, we recommend using Notepad ++ or UltraEdit text editor such as software;
gm_continuous_digit.abnf is asrdemo_with_abnf used in the syntax file, coded as GB2312 , when modify its contents, please refer to DOC "Grammar Development Guide" folder;
wav folder which has a pre-prepared recording papers were used asrdemo , asrdemo_with_abnf , iatdemo , iatdemo_with_multi-Result ;
After running through the sample program, but also to generate a md5 file is the name of the folder in which to save the file contents of the performance information and error messages recorded in the program is running, will be uploaded to the server, it will help us improve the quality of service Please do not delete.
7. New iatdemo_with_user_vocabulary (upload user vocabulary Demo ), after uploading keyword, when the sessions had no effect, the clouds take some time to activate the keyword table, you can use normal later.
Big recognition error rate Analysis
Check grammar file ABNF , found that the English called "continuous numeric", it is speculated that the syntax is only suited to a particular environment. Since the SDK is only one syntax file, so I use the "Voice Dictation" do not use "voice recognition" in the next process , after a series of changes in procedures successfully identified the text.