Menu

Pocketsphinx Recognition fails completely for one particular word for all speakers in several sentences

Help
Q3Varnam
2018-05-02
2018-05-04
  • Q3Varnam

    Q3Varnam - 2018-05-02

    I have the words in dictionary, phones as well as the transcription in unicode text. It works well. However I added a new set of sentences, around 30, which had a new word, the same word present in all those 30 sentences in different usages. I have recordings of 3 speakers who have spoken these sentences.

    The dictionary,phone list, transcription, fileids - all have been correctly updated. No extra spaces or any such first level errors

    Sphinxtrain run hasn't produced any errors for those transcription or files, I have checked the logs inside 20.ci_hmm folder, the said audio files have been processed without error.

    But when the test is run with the same files used for training - no match occurs for this particular word alone - 100% failure.

    Any suggestion as to where to look further to debug this problem

     
  • Q3Varnam

    Q3Varnam - 2018-05-02

    I have fixed this problem by increasing the senones from 200 to 2000 in the config file

     

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.