schmiph2 / pysepm

Python implementation of performance metrics in Loizou's Speech Enhancement book

pysepm - Python Speech Enhancement Performance Measures (Quality and Intelligibility)

Python implementation of objective quality and intelligibilty measures mentioned in Philipos C. Loizou's great Speech Enhancement Book. The Python implementations are checked with the MATLAB implementations attached to the book (see Link)

Install with pip

Install pysepm:

pip3 install https://github.com/schmiph2/pysepm/archive/master.zip

Examples

Please find a Jupyter Notebook with examples for all implemented measures in the examples folder.

Implemented Measures

Speech Quality Measures

• Segmental Signal-to-Noise Ratio (SNRseg)
• Frequency-weighted Segmental SNR (fwSNRseg)
• Log-likelihood Ratio (LLR)
• Weighted Spectral Slope (WSS)
• Perceptual Evaluation of Speech Quality (PESQ), (python-pesq implementation by ludlows)
• Composite Objective Speech Quality (composite)
• Cepstrum Distance Objective Speech Quality Measure (CD)

Speech Intelligibility Measures

• Short-time objective intelligibility (STOI), (pystoi implementation by mpariente)
• Coherence and speech intelligibility index (CSII)
• Normalized-covariance measure (NCM)

Dereverberation Measures (TODO)

• Bark spectral distortion (BSD)
• Scale-invariant signal to distortion ratio (SI-SDR)