Skip to content

Latest commit

 

History

History
20 lines (16 loc) · 888 Bytes

README.md

File metadata and controls

20 lines (16 loc) · 888 Bytes
Raw

iCubrec

This folder contains code to train, test and run an automatic speech recognition (ASR) system. Even though the scripts are quite generic and can be used to trained models for other purposes, our ultimate goal is to provide tools to perform command recognition on the iCub plateform.

The code is split into two subfolders:

  • htk contains the first version of the code which was based on the Hidden Markov Model ToolKit (HTK). As HTK doesn't allow live recognition with deep neural networks (DNNs), it is based on Gaussian mixture models (GMMs) instead.
  • kaldi contains a more recent version of the pipeline based on kaldi. It uses a DNN-based acoustic model and incorporates a voice activity detection (VAD) system to allow hand-free online detection of commands.

License

The code is released under GPLv3 license.