damellis / ESP

The Example-based Sensor Predictions (ESP) system applies machine learning to real-time sensor data.
BSD 3-Clause "New" or "Revised" License
224 stars 52 forks source link

Basic audio keyword spotting w/ DTW. #308

Open damellis opened 8 years ago

damellis commented 8 years ago

With MFCC features?

nebgnahz commented 8 years ago

This paper "Discriminative Keyword Spotting" has some discussion on this subject. http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/34559.pdf

In particular, it suggests:

keyword spotting has first been addressed with models based on Dynamic Time Warping (DTW) (Bridle 1973; Higgins and Wohlford 1985)

damellis commented 8 years ago

Also see: http://www.ipcsit.com/vol31/028-ICIII2012-C30005.pdf

damellis commented 8 years ago

I doubt I can make this happen for the workshop. @nebgnahz feel free to take a shot at it.