Speech Trainer Engine

       

How to start?

Go to -> Start -> Programs Files -> Speech -> Trainer.exe OR open Trainer.exe from the installation directory. Once open the system will look something like below.

1. "Write Training Sentence Here."
User will type the training sentence here with a "." at the end. System will be trained for the words in the sentence.

2. "UserID"
User will type the UserID in this text area. There is no uniqueness in the ID, hence if ID already exists then speech data may be corrupted or overwritten.

3."         "
This box is nothing but a status panel that shows the current processing and any user instruction.

4."         "
This box shows the current training sentence.

5."         "
This box shows a list of words for which user has provided features. It means words that are to be trained for the user.

6."         "
This box shows a list of words trained by the user.

 

How to use?

1. Select a UserID Entering the UserID can be done in 2 ways. First one, Click on the "User Name" button and supply the UserID in the popped box that looks like below. Second one, Enter the UserID in the Text Box (labeled as "UserID") and press Enter key.

2. Set Number of Samples Number of samples are nothing but the number of times user will utter the training sentence. Its default value is 5, to change click on the "Configuration" button and enter any integer value (more than 5) in the popped box that looks like below. Please note: minimum of 5 samples are required.

3. Select Training Sentence Training sentence is to be selected by the user to train the system for the words in the sentence. Enter the sentence in the Text Box labeled as "Write Training Sentence Here." and press Enter key or button labeled as "Enter Sentence".

Note: Training sentence must end with a "." & User has to take proper care while entering training sentence else recognition will be erroneous.

4. Uttering Training Sentence Once the Training sentence is selected user needs to utter the sentence for minimum of 5 times or as configured. Press "RECORD" button and start uttering and Press "STOP" button to stop uttering for each time.
Once features are taken for the selected sentence, all the words will be added in Box 5.

Note: Maintain a small pause between two words else word separation will be faulty leading to erroneous training. This part only takes the features for the sentence and doesn't train actually. Final training is still pending and will be discussed in the following sections.

5. Word Training After features for few words are taken user can train the system. Just click on "Train Pending Words" button.

Note: Higher the number of words in the pending list higher will be the recognition performance. Hence, it is suggested to extract features for large number of sentences.

Miscellaneous

This section deals with functionalities that are left in upper sections.

1. Cancel Sample
If user is not satisfied for his last utterances, he may cancel those utterances by clicking on the button labeled as "Cancel sample".

2. Refresh Status By clicking on this button Status Panel can be refreshed.

3. Refresh Trainer By clicking this button new session can be started.


Animesh Kumar
animesh@vectorstar.net

Contact Phone: +91 9835141930