Hi Everyone,
This is my week 9 Progress Report.
Tasks Done:
-
Restructuring the Crossword game…and working on it’s GUI.
-
Updating the documentation…with examples…and detailed description and explanation of game-codes.
-
I went through the contents of a Speech Recognition course…to fully understand the principle of acoustic model and language model generation…though this might not be directly linked with my project work…but had to do it…to improve the work on auto-enhancement of the phonetic dictionary. Right now…what I am doing is…enhancing the dictionary by using the results generated from the hypothesis files after decoding…and further improving the results by splitting the output into phonemes present already in the dictionary previously…
This has produced erroneous results sometimes…sometimes good results. As of now…the speech recognition accuracy has improved tremendously using the AGC microphone…and the previous data in the dictionary itself suffices…but hope to improve the auto-enhanching part in the future…using all the acoustic model generation algos…in my workflow.
Issues Faced:
- Sorry to say that the breakout board which I had ordered… isn’t arriving before Friday this week…so wouldn’t be able to give any status on my hardware work before that.
Work to be done:
-
Resolve the remaining part of my project.
-
Updating the documentation with all the hardware issues and troubleshooting mechanisms…which were being discussed in the mails.
-
The bigger picture of the project was also giving a background for development of future voice based digital assistants with the pocketbeagle…so will explain the ways to utilise the speech processing API efficiently for the purpose…and if possible…make some necessary improvements on it.
Regards,
Anirban