	Sign In Sign-Up

My efforts are dedicated to my "Source of Inspiration..."

anim1058.gif (13911 bytes)
[ BASICS ] [ Architecture ] [ Text To Speech ] [ Speech Recognition ] [ Microphones ] [ Test Cases ] [ For OEM's ] [ Future ] [ SAPI4.0 ] [ Wow ]

Test Cases for Speech Aware Applications

Adding speech recognition and text-to-speech adds a few more things that could go wrong with an application. Here's a list of tests to try with your speech aware application.

Setup

	Does it install over existing speech API installs? - Test that an install over existing speech API installs works. An upgrade over an old version of the API should replace the old API components. An upgrade over a newer version of the API should overwrite nothing.
	Does it install over existing speech engine installs? - Test that an install over existing engine installs works. An upgrade over an old version of the engine should replace the old engine components. An upgrade over a newer version of the engine should overwrite nothing.
	Does it work with other speech aware applications? - Don't break other speech aware applications by overwriting their components or registry entries.
	Uninstall - Make sure that uninstalling your software doesn't prevent other speech aware applications from working.

Sound Cards

	Different sound cards and driver versions - Make sure to test on different sound cards and driver versions. The variation in the sound cards' mixers and quality might cause problems with your application.
	Test for sound card - Identify if the user doesn't have a sound card, or the user's sound card can't handle SR/TTS, and inform the user.
	More than one sound card - The user might have more than one sound card. Make sure they can chose the right one.
	Half and full duplex sound cards - Test on half-duplex and full-duplex sound cards. Your application may behave differently in half-duplex than full-duplex.
	Voice-Fax-Modem cards - Voice modem cards provide wave-out and wave-in devices for the phone. Make sure your application doesn't accidentally chose these as a primary listening and playback device.
	USB microphones and speakers - USB speakers are already on the market. USB microphones will appear soon. They may behave differently than sound cards.

Speech Engines

	Does your product work with different engines? - It's likely that users will install other products using speech recognition and text-to-speech. Users may select different engines to be used with your application, so you should test with them. Make sure that subtle functionality differences don't interfere with your application.
	Does your application require a specific engine? - If your application requires a specific engine to function, make sure that it will select that engine, even if others are installed on the system.
	Get the right type engine - Engines come in different categories; Speech recognition can be dictation and/or command & control. Text-to-speech can be synthesized and/or recorded words. Make sure your application doesn't accidentally select the wrong type of engine. Ex: If you application uses command & control, try installing a dictation engine and make sure the application doesn't select it or allow the user to select it.
	Get the right type language - Engines supporting different languages might be installed on the PC. Make sure the application chooses the proper language.
	Phone vs. PC models - Many engines have special models for phone. Make sure your application doesn't accidentally use these if they're installed.
	No engine installed - The application should fail gracefully if no engine is installed on the system.
	Select favorite engine - The user may wish to select their favorite speech recognition or text-to-speech engine. Make sure this feature works.

Usability

	Can the user get the SR/TTS working? - Make sure the user can get the speech recognition and/or text-to-speech installed and working with minimal effort.
	Can the user get their microphone working? - Do extensive testing to make sure the user can plug in their microphone and use the microphone correctly.
	Does the user understand how to talk to the PC? - Make sure that users can figure out how to talk to their PCs.

Miscellaneous

	Proper machine speed - Some engines require Pentium-speed or better to function. Inform the user if their PC is too slow for speech recognition and/or text-to-speech. Some older Pentiums with faulty math coprocessors will have their math coprocessor disabled by Windows NT. If the speech engine relies on floating point, it may be too slow on these machines.
	Gracefully fail if the API/engine is not working - Another application's install may break your install of the API and/or engine. If this happens, CoCreateInstance() and other calls will fail. Inform the user of the failure and recommend they reinstall to get speech recognition and/or text-to-speech working again.

[ Home ] [ Up ] [ BASICS ] [ Architecture ] [ Text To Speech ] [ Speech Recognition ] [ Microphones ] [ Test Cases ] [ For OEM's ] [ Future ] [ SAPI4.0 ] [ Wow ]

Please sign my guest book:

Send mail to askazad@hotmail.com with questions or comments about this web site.
Copyright � 2001 Engineered Station
Last modified: July 06, 2001

This site is been visited by surfers