HMM state-level Forced Aligner

Note: This repository is now combined with https://github.com/jaekookang/p2fa_py3 as of 06/22/2020. It will not be maintained anymore.

---

Introduction

• This is a HMM state-level forced aligner (FA) based on P2FA Yuan & Liberman 2008.
• Based on the trained P2FA model, the current FA appends an additional 'state' tier to show state-level alignment (e.g. s2, s3, s4 in 5-state HMMs excluding initial (s1) and final state (s5)).
• For the changes made, check align.py.
• Example: 908-31957-0001.wav

Prerequisites

• HTK toolkit version 3.4
• Python (tested on Python2.7 and Python3.6)

Run

# Run one file
python align.py example/test.wav example/test.txt example/test.TextGrid

# Run multiple files
. ./test_libri.sh