Speech Transmission Index for Public Address (STIPA) on GitHub

Pavel 2024년 1월 23일
최근 활동 Hans Scharler답글: 2024년 1월 24일

We've released an open-source implementation of STIPA (Speech Transmission Index for Public Address) on GitHub!
What is STIPA?
Speech Transmission Index is a metric designed to assess the quality of speech transmission through a communication channel. It quantifies the intelligibility of speech signals based on amplitude modulations, providing a standardized measure crucial for evaluating public address systems and communication equipment. STIPA is a version of STI using a simplified measurement method and only one test signal.
Quality Representation:
STI values range from 0 to 1, categorizing speech transmission quality from bad to excellent. The raw STI score can be transformed into the likelihood of intelligibility of syllables, words, and sentences being comprehended.
Verification Tests:
To ensure reliability, we've conducted verification tests according to the IEC 60286-16 standard. Download the test signals and run the stipaVerificationTests.m script from our GitHub repository.
Control Measurements:
We've performed comparative measurements in a university auditorium, showcasing the validity of our implementation.
License:
Our STIPA implementation is distributed under the GNU General Public License 3, reflecting our commitment to open-source collaboration.
Hans Scharler
Hans Scharler 2024년 1월 23일
Do you know if this works on MATLAB Online?
Pavel
Pavel 2024년 1월 24일
Yes, the scripts for generating a STIPA test signal and computing the STI using stipa function should work on MATLAB online. Only the stipaVerificationTests have to be slightly modified since MATLAB online does not support folders. So everything has to be in the root directory and the paths in the audioread function have to be modified accordingly.
Hans Scharler
Hans Scharler 2024년 1월 24일
Investigating. But, I am sure MATLAB Online supports folders.
Hans Scharler
Hans Scharler 2024년 1월 23일
Congrats on the release!
Pavel
Pavel 2024년 1월 24일
Thank you!