A simple approach to Speech Recognition - step.0

 

I’ve spent several months in developing a speech recognition system. At the beginning of this project journey, I know nothing about the theorems of how to make machines understand human talk.

First of all, I found some primary sample module diagram both with some papers on the topic to take my first step.

Here are the summaries of the serial steps to a speech recognition system.
1. pick out which part of the signal collected is valid speech;
2. extract some characteristics from the selected;
3. analyze the characteristics using templates pre-defined;
4. decide according to the comparison of current speech and templates.

And these steps are often called as:
1. end-point detection (I think terminal point is better :-);
2. feature extraction;
3. distance measurement;
4. decision strategy.

To be continue…