The Kaldi-based Standalone Pitch Extractor Toolkit

OverView

1. The standalone pitch extractor toolkit is based on Kaldi toolkit (https://github.com/kaldi-asr/kaldi)

2. The implementation of the standalone pitch extractor toolkit is based on the paper--A Pitch Extraction Algorithm Tuned For Automatic Speech Recognition, whose author is Pegah Ghahremani, Daniel Povey and etc, in technology. (http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=6854049)

   Our algorithm is a highly modified version of the getf0 algorithm. It is based on finding lag values that maximize the Normalized Cross Correlation Function. Probably the most important change from getf0 is that rather than making hard decisions about voicing on each frame, we treat all frames as voiced and allow the Viterbi search to naturally interpolate across unviced regions.

3. The toolkit does not depend on any other toolkits, so all the source codes are located in src/. For more information about source codes, see src/README.md

   To build the toolkit: see src/INSTALL. These instructions are valid for UNIX systems including various flavors of Linux; Darwin; and Cygwin (has not been tested on more "exotic" varieties of UNIX).

   To run the example system bilds, see eg/README.md

   If you encounter problems (and you probably will), please do not hesitate to contact the developers (see below). In addition to specific qestions, please let us know if there are specific aspects of the project that you feel could be improved, that you find confusing, etc., and which missing features you most wish it had.

Development pattern for contributors

1. Create a personal fork of the main Kaldi repository in GitHub.
2. Make your changes in a named branch different from master, e.g. you create a branch my-awesome-feature.
3. Generate a pull request through the Web interface of GitHub.
4. As a general rule, please follow Google C++ Style Guide. There are a few exceptions in Kaldi. You can use the Google's cpplint.py to verify that your code is free of basic mistakes.

Authors

Daniel Povey ([email protected]) Pegah Ghahremani ([email protected]) Hang Lyu ([email protected])

Platform specific notes

PowerPC 64bits little-endian (ppc64le)

• The tool is expected to work out of the box in RHEL >= 7 and Ubuntu >= 16.04

Android

(TODO: Test)

• The tool supports cross compiling for Android using Android NDK, clang++
• See this blog post for details.