At the forefront of Artificial Intelligence

Home » Reviews » Hardware » Speech Technologies

Sensory Voice Extreme Toolkit

The Voice Extreme Toolkit from Sensory Inc. is prototyping kit that contains a development board and software required for building interactive speech applications.

The kit (pictured right) contains:

Voice Extreme Development Board
Voice Extreme Module
Power Supply
Serial cable for program download
Quick start guide
CD-ROM with software

Getting started with the kit is very simple, the software installs and will detect the COM port you've attached the development board. The Quick Start Guide was a little confusing at times, but nothing that couldn't be overcome with a little thinking!

Voice Extreme IDE

The review kit shipped with version 2.2.2 of the IDE. The IDE will definitely suffice any speech development that will be done for the VE Module. The VE Module is programmed in VE-C which is very similar to ANSI-C bar a few additions and omissions. For example, the first "Hello World!" program I created in the VE IDE looked like:

extern SPEECH VPhelloworld;
main ()
{
  Talk(0, &VPhelloworld); // Say "hello world"
}

Where VPhelloworld has been declared in the project file as a audio file that says "Hello World!".

My experience with VE-C and the VE IDE so far has been very positive. The SDK has a range of helper functions that enable development to be completed quicker once you've understood their purpose. The IDE will compile and link the projects and download them to the development board. This process, though, does have some room for improvement. Firstly, I've had the compiler crash on me several times when the open/closing bracket count wasn't equal. Secondly, the download procedure requires you to press a button on the development board itself before downloading will begin. It is quite easy to forget to do this, despite the prompt.

This aside, the development environment itself is well-designed, and allows any programmer to start developing speech applications quickly.

Voice Extreme Development Board & Module

The development board comes with 3 buttons and 3 LEDs (red, yellow, green) as well as a microphone and speaker, ensuring that you have all the necessary hardware to test your speech application. Furthermore, the VE Module can be detached from the board easily, allowing you to connect it to other electrical equipment for further testing. Finally, the development board comes with a small prototyping area for custom electronics to be wired in. As such, there is nothing bad I can say about the development board.

The VE Module itself, is similarly impressive. While the VE Module does not support dynamic speech synthesis, it relies on audio clips that it can splice together to make commands and prompts. This works better, since most embedded speech applications will warrant a clear, natural narrator for the prompts. The VE Module's speech recognition is very impressive, it nearly always correctly recognized what I said, from a distance of about 1 metre, on a small button microphone that ships with the kit.

One element that did disappoint me was the lack of speaker-independent recognition. The VE Module does in fact support it, but requires a 'Sensory linquist' to create the necessary weights file (at additional cost). This said, the kit doesn't come completely devoid of any SI weight file, but comes with 6 very well hidden (not mentioned in documentation) weight files that contain the digits from zero to nine, commands useful for mobile robots ("go forward", "stop"), commands useful for an answering machine ("call", "erase"), and another set useful for a security system ("arm", "disarm").

Similarly, while the VE Module also supports music, the manual states that Sensory should be contacted if users want to create their own music.

Talking of the manual, it is also worth noting that the manual does an excellent job. While the Quick Start guide was a little clumsy, the programmer's manual was well written, with all necessary functions documented. I would have liked to have seen the online help file better indexed, and perhaps context-sensitive help for the SDK functions.

Conclusion

Overall, the Sensory Voice Extreme Toolkit is a very nice package. The development board and support software are well designed and interfaced. Additionally, while not perfect, the documentation is comprehensive and well written. The only major problem lies with the carrot-on-a-stick nature of the speaker-independent recognition and music features of the VE Module. Nevertheless, programmers looking at prototyping speech applications for embedded platforms shouldn't need to look further than Sensory's VE Kit.

Cover 8.9

Price: £99.50

Liked: Good, well-featured IDE, decent voice recognition

Disliked: Serial port only, limited voice-independent libraries supplied

Website: http://www.active-robots.com/

Submitted: 08/06/2003

Article Toolbar

BibTeX entry