Implementation of the classical and extended Short Term Objective Intelligibility measures
Intelligibility measure which is highly correlated with the intelligibility of degraded speech signals, e.g., due to additive noise, single/multi-channel noise reduction, binary masking and vocoded speech as in CI simulations. The STOI-measure is intrusive, i.e., a function of the clean and degraded speech signals. STOI may be a good alternative to the speech intelligibility index (SII) or the speech transmission index (STI), when you are interested in the effect of nonlinear processing to noisy speech, e.g., noise reduction, binary masking algorithms, on speech intelligibility.
Description taken from Cees Taal's website
pip install pystoi or
pip3 install pystoi
import soundfile as sf from pystoi import stoi clean, fs = sf.read('path/to/clean/audio') denoised, fs = sf.read('path/to/denoised/audio') # Clean and den should have the same length, and be 1D d = stoi(clean, denoised, fs, extended=False)
All the Matlab code in this repo is taken from or adapted from the code available here (STOI – Short-Time Objective Intelligibility Measure – ) written by Cees Taal.
Thanks to Cees Taal who open-sourced his Matlab implementation and enabled thorough testing of this python code.
If you want to run the tests, you will need Matlab,
matlab.engine (install instructions here) and
matlab_wrapper (install with
pip install matlab_wrapper).
The tests can only be ran under Python 2.7 as
matlab_wrapper are only compatible with Python2.7
Tests are passing at relative and absolute tolerance of
1e-3, which is enough for the considered application (all the variability is coming from the resampling method when signals are not natively sampled at 10kHz).
Very big thanks to @gauss256 who translated all the matlab scripts to Octave, and wrote all the tests for it!
Any contribution are welcome~, specially to improve the execution speed of the code~ (thank you Przemek Pobrotyn for a 4x speed-up!) :
tests/.~ This can be considered a solved issue thanks to @gauss256 !