Thank you, Shmyrev.
I am looking for something more specific. In my understanding, recognition task is to find the most likely path, based on the language model and acoustic model. I am trying to understand the interplay of these two models in the decoding process. During my search, I came across terms like stack decoding, A* search etc. I was able to find the thesis "Efficient Algorithms for Speech Recognition" by Mosur K. Ravishankar which explains Sphinx2 baseline system. I was hoping that there would be a published paper or thesis which explains the theory behind the algorithms used in pocketsphinx as well.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Could you please refer to the decoder algorithm used in pocket sphinx ?
Last edit: sarvesh 2013-05-28
http://en.wikipedia.org/wiki/Viterbi_algorithm
Thank you, Shmyrev.
I am looking for something more specific. In my understanding, recognition task is to find the most likely path, based on the language model and acoustic model. I am trying to understand the interplay of these two models in the decoding process. During my search, I came across terms like stack decoding, A* search etc. I was able to find the thesis "Efficient Algorithms for Speech Recognition" by Mosur K. Ravishankar which explains Sphinx2 baseline system. I was hoping that there would be a published paper or thesis which explains the theory behind the algorithms used in pocketsphinx as well.
Pocketsphinx algorithms are the same as in sphinx2, sphinx3 or even sphinx4, it's just a more modern implementation of the same idea