Acoustic Keyword Spotting Steps to Follow

[Yasin Shihab]

Hello,

I want to develop acoustic keyword spotting for a new language(non-english).
My goal is to find out, if some word(s) exist in some sound(.wav) file(s).
Like, "word_1" exist in "sound_file_2" , "word_2" exist in "sound_file_1" etc.
I want to build that system using pocket-sphinx and I'm new to CMU Sphinx.

If anyone could help me giving steps to follow(step by step approach).

Thanks in advance.

[Nickolay V. Shmyrev]

See

https://www.quora.com/How-can-I-get-started-with-the-CMUSphinx-setup-for-building-a-new-languages-speech-recognition

[Yasin Shihab]

Nickolay V. Shmyrev thanks for that informative link. I read this post also
https://sourceforge.net/p/cmusphinx/discussion/help/thread/1f90526e/

Here you suggested Arun venkatesh to follow https://github.com/castorini/honk for
small set of keyword spotting. In my case it will be 150 - 200 words or a little more. From stored sound files my system will be able to find out those files that are relveant to search word & containing utterance of that word at least once.

My question: Is CMU SPhinx matured enough for such stuff ? Or, I have to move for other NN solutions or else...

Thanks again.

[Nickolay V. Shmyrev]

In my case it will be 150 - 200 words or a little more.

Honk will not work well for 200 words

My question: Is CMU SPhinx matured enough for such stuff ? Or, I have to move for other NN solutions or else...

You can use Kalid

[Yasin Shihab]

Thanks Nickolay V. Shmyrev .