Digital Signal Processing Reference
In-Depth Information
Figure 12-7.
Hardware OEM-module vicCORE2195G3
The algorithms were implemented with a strict code data separation.
Since the software reference is an experimental system and contains by far
more optional algorithms than used in a given application, a technology was
developed for the easy porting of all the control structures and knowledge
bases of the recognizer. A number of automatically generated C-headers,
which contain the whole initialized bunch of control structures and know-
ledge bases, are included automatically in the compilation of the DSP code.
In this way, a fast and efficient porting is achieved, without loss in accuracy
and without the need of type conversions from one platform to another.
6.
SUMMARY
A new speech recognition technology was presented in this chapter. It
was shown, how this technology was applied in a typical car application and
which steps were taken to provide the necessary robustness under adverse
acoustic conditions.
It should be lined out, that technology is not the only key component for
the success of a product. Since modern technical products have implemented
more and more complex functions, information sometimes overload users.
Reading of extensive manuals is needed, and this often leads to the effect
that the user only knows the features that are necessary for a basic device
operation.
The design of the suitable user interface more and more decides on the
success of a product. Speech interfaces allow a much more powerful and
efficient user interface than available in conventional devices. The
replacement of buttons by speech commands is only a small part of the
possible advantages of speech interfaces.
In the future, collaboration between speech interface designers, device
vendors and customers should be much closer to explore fully the new
