python version：https://github.com/IMLHF/PHASEN-PyTorch/blob/master/phasen_torch/sepm.py

This folder contains implementations of objective quality measures (Chapter 11):

MATLAB file	Description                                 Reference

---

comp_snr.m	Overall and segmental SNR                       [1]

comp_wss.m	Weighted-spectral slope metric                  [2]

comp_llr.m	Likelihood-ratio measure                        [3]

comp_is.m	Itakura-Saito measure                           [3]
comp_cep.m	Cepstral distance measure                       [4]
comp_fwseg	Freq. weighted segm. SNR (fwSNRseg)    	        [5],Chap 11
								 
comp_fwseg_variant   Frequency-variant fwSNRseg measure		Chap 11 
								
comp_fwseg_mars	    Frequency variant fwSNRseg measure 		Chap 11
		    based on MARS analysis				 									

comp_pesq.m		PESQ measure (narrowband)   ITU-T P.862             [6]
            PESQ measure (wideband)     ITU-T P.862.2           [7]

composite.m	A composite measure                                 [8]


addnoise_asl.m	Adds noise to the clean signal at specified SNR 
		based on active speech level.                           [9]

---

USAGE

[snr_mean, segsnr_mean]= compSNR(cleanFile.wav, enhdFile.wav);
% where 'snr_mean' is the global overall SNR and 'segsnr_mean' is the segmental SNR.

wss_mean = comp_wss(cleanFile.wav, enhancedFile.wav);

llr_mean= comp_llr(cleanFile.wav, enhancedFile.wav);

is_mean = comp_is(cleanFile.wav, enhancedFile.wav);

cep_mean = comp_cep(cleanFile.wav, enhancedFile.wav);

fwSNRseg = comp_fwseg(cleanFile.wav, enhancedFile.wav);

[SIG,BAK,OVL] = comp_fwseg_variant(cleanFile.wav, enhancedFile.wav);
% where 'SIG' is the predicted rating of speech distortion, 
% 'BAK' is the predicted rating of background noise distortion,
% 'OVL' is the predicted rating of overall quality.

[SIG,BAK,OVL] = comp_fwseg_mars(cleanFile.wav, enhancedFile.wav);

pesq_val = comp_pesq(cleanFile.wav, enhancedFile.wav);
% Only sampling frequencies of 8000 Hz or 16000 Hz are supported.

[Csig,Cbak,Covl] = composite(cleanFile.wav, enhancedFile.wav);
% where 'Csig' is the predicted rating of speech distortion,
% 'Cbak' is the predicted rating of background noise distortion,
% 'Covl' is the predicted rating of overall quality.

addnoise_asl(cleanfile.wav, noisefile.wav, outfile.wav, SNRlevel)

---

REFERENCES:

[1] Hansen, J. and Pellom, B. (1998). An effective quality evaluation protocol for speech enhancement algorithms. Inter. Conf. on Spoken Language Processing, 7(2819), 2822

[2] Klatt, D. (1982). Prediction of perceived phonetic distance from critical band spectra. Proc. IEEE Int. Conf. Acoust. , Speech, Signal Processing, 7, 1278-1281.

[3] Quackenbush, S., Barnwell, T., and Clements, M. (1988). Objective measures of speech quality. NJ: Prentice-Hall, Eaglewood Cliffs.

[4] Kitawaki, N., Nagabuchi, H., and Itoh, K. (1988). Objective quality evaluation for low bit-rate speech coding systems. IEEE J. Select. Areas in Comm., 6(2), 262-273.

[5] Tribolet, J., Noll, P., McDermott, B., and Crochiere, R. E. (1978). A study of complexity and quality of speech waveform coders. Proc. IEEE Int. Conf. Acoust. , Speech, Signal Processing, 586-590.

[6] ITU (2000). Perceptual evaluation of speech quality (PESQ), and objective method for end-to-end speech quality assessment of narrowband telephone networks and speech codecs. ITU-T Recommendation P.862

[7] ITU (2007). Wideband extension to Recommendation P.862 for the assessment of wideband telephone networks and speech codecs. ITU-T Recommendation P.862.2

[8] Hu, Y. and Loizou, P. (2006). Evaluation of objective measures for speech enhancement. Proc. Interspeech

[9] ITU-T (1993). Objective measurement of active speech level. ITU-T Recommendation P. 56

Copyright (c) 2012 by Philipos C. Loizou

Revision: 1.0, Date: 05/14/2012

---