About this Directory
This directory is setup with
Google Summer of Code 2016 submission in mind. So here goes, on how the project can be used as such or for reference on how it looks after development.
Instructions or Key points
- Two sub-directories :
amfolder contains the Acoustic Model result parameters.
lmfolder contains the Language Model as well as the Dictionary file required to run the recognition which is explained later in this README.
- These files are not for development. If that's what you are looking for then go here or go to the parent directory of
- To develop Language Model, you may read this or the official CMU Sphinx wiki
- To develop Acoustic Model, you may read this or the official CMU Sphinx wiki
- To get a more clear idea on how to proceed your way through the develop, I recommend reading my blog posts on Medium
- Setting up your development environment is rather straight forward. Even then if you have any hickups, please consider reading the root README.md
How to run the models for recognition
Before running the recognition, make sure you have installed all the sphinx related executables as well as dependencies mentioned in the source README.>
Open up the terminal in this root directory or
cdyour way through upto this.
Type in the command:
$ pocketsphinx_continuous -hmm ./am/ -lm ./lm/ml.arpa -dict ./lm/ml.dic -inmic yes
Note: The output on the terminal is a bit sketchy since Malayalam is not supported. A workaround would be to pipe an export ( the output which is the hypothesis from pocketsphinx ) to a text file.
$ pocketsphinx_continuous -hmm ./am/ -lm ./lm/ml.arpa -dict ./lm/ml.dic -inmic yes | tee hypothesis.txt
Note: This model can also be used for making Android Malayalam Speech Recognition applications. For more details, you can visit here.