Installation

To install Speech Recognition Engine 2.0, follow these steps:

1. Create the separate directory and copy engine.exe there.

2. Copy RPL 4.018 and SPL 4.0 files to Windows/System or Engine directory.

3*. Turn off the computer.  Insert two sound cards in it and the turn the computer on.  Install both sound cards in control pannel.  This may require solution of some hardware conflicts.  We recommend the usage of two Creative Sound Blaster 16 cards.

4. Insert multi-sensor setup output connectors to the corresponding Line In connectors of sound card(s).  If you use single microphone instead of using the setup, you should use the microphone input.

5. Setup the sound input in Sound Mixer.  To do this, double-click the Bell in the right lower corner of the screen, then choose Options-Properties, choose the sound device (you must do this for every sound card in your PC), check Recording, then turn the recording Line input on and adjust the recording level.

6. Run engine.exe.  Create a new User. 

7. Go to Options dialog pressing "Options..." button.  Choose "Basic Channels" property sheet.  Setup basic channels.  To identify, which type of channel is on which input channel, speak to the setup from some distance, not putting it to your head.  The two channels with changing indicator are microphone inputs (left channel for close microphone and right one for far one), and the rest one are additional channels (left for photo sensor and right for breath sensor).  Choose simple photo channel as a type of foto channel.  Use LP spectrum for both microphones and energy for photo and breath sensors.

8. Now the engine is ready for constructing additional channels.  You can add them on "additional channels" property sheet.  Remember, that to start the recognition, you should:
    a. Install the program properly.
    b. Create some spectral channels
    c. Add some of them to resulting channels list
    d. Create at least one speech detector and train it using Speech detectors property sheet
    e. Create a list of speech commands and train them.

9. Do not forget to reset and setup spectral channels after each new start of Engine program and after adding or modifiing of spectral channels (i.e. every time when you press Options button).  To train channels, just press the "Channels setup..." button, press "Reset all channels", then "Train all channels".  Don't speek for at least one second and press this button again.   The training is complete.

* This step is optional.  The program works with one sound card using 2 (or even 1) audio channels