vineeths96 / Spoken-Keyword-Spotting

In this repository, we explore using a hybrid system consisting of a Convolutional Neural Network and a Support Vector Machine for Keyword Spotting task.
MIT License
90 stars 18 forks source link

How to train for new keyword? #18

Open boleamol opened 2 years ago

boleamol commented 2 years ago

Dear @vineeths96

Thank you for awesome recipe.

I have few questions 1) How to train for new keyword? 2) How to tune hyper parameter? 3) How to display position (at what second/minute) of keyword in audio file (e.g. keyword found at 5th second, 9th second in audio file)?

RahulBhalley commented 1 year ago

Hey @boleamol! Did you figure it out? Or is there any better model nowadays?

boleamol commented 1 year ago

@RahulBhalley: Not yet.... If you have any other option, please suggest.

RahulBhalley commented 1 year ago

Hmm, I'm new to this. Just got an idea that maybe we can create a dataset of specific keyword and train it. And the new spoken words can be generated using some TTS software.