log：
/Users/wangshaoce/test_project/pysepm/test.py:6: WavFileWarning: Chunk (non-data) not understood, skipping it.
fs, clean_speech = scipy.io.wavfile.read('1_speech_16000_Hz.wav')
/Users/wangshaoce/test_project/pysepm/test.py:7: WavFileWarning: Chunk (non-data) not understood, skipping it.
fs, noisy_speech = scipy.io.wavfile.read('1_noisySpeech_16000_Hz.wav')
this is my code：
import pysepm
import scipy.io.wavfile
import wavio
import sys
sys.path.append("../")
fs, clean_speech = scipy.io.wavfile.read('1_speech_16000_Hz.wav')
fs, noisy_speech = scipy.io.wavfile.read('1_noisySpeech_16000_Hz.wav')

fs, enhanced_speech = scipy.io.wavfile.read('1_processed_16000_Hz.wav')

pysepm.pesq(clean_speech, noisy_speech, fs)

pysepm.pesq(clean_speech, enhanced_speech, fs)

Google didn't solve my problem
I look forward to your reply
best wishes

Hi,

thank you for the library.
I am working on rewriting of some of measures to C#. I would like to ask, if the BSD measures already works or not. I have seen in TODO list but I found the code in repository.

Thx
pauquail

I am looking at the DNS-2020 test dataset:

https://github.com/microsoft/DNS-Challenge/tree/interspeech2020/master/datasets/test_set/synthetic/no_reverb

And I want to see the results of as many metrics as possible:

def get_evaluation(clean_speech, noisy_speech, sr = 16000):
    
    Y0 = pysepm.fwSNRseg(clean_speech, noisy_speech, sr)
    Y1 = pysepm.SNRseg(clean_speech, noisy_speech, sr)
    Y2 = pysepm.llr(clean_speech, noisy_speech, sr)
    Y3 = pysepm.wss(clean_speech, noisy_speech, sr)
    Y4 = pysepm.cepstrum_distance(clean_speech, noisy_speech, sr)
    Y5 = pysepm.stoi(clean_speech, noisy_speech, sr)
    Y6, Y7, Y8 = pysepm.csii(clean_speech, noisy_speech, sr)
    _, Y9 = pysepm.pesq(clean_speech, noisy_speech, sr)
    Y10, Y11, Y12 = pysepm.composite(clean_speech, noisy_speech, sr)
    Y13 = pysepm.ncm(clean_speech, noisy_speech, sr)
    
    return [Y0,  Y1,  Y2,  Y3,  Y4,
            Y5,  Y6,  Y7,  Y8,  Y9,
            Y10, Y11, Y12, Y13]

But my output looks like this:

[nan,
  8.621018369535282,
  0.1498153037618391,
  nan,
  1.9050031356095016,
  0.9807254867015377,
  nan,
  nan,
  nan,
  2.3495934009552,
  nan,
  nan,
  nan,
  nan]

What may be the issue?

Hello, I am studying about voice improvement.

If you look at your notbook, you are comparing clean and noisy, clean and enhanced.

But I've always compared noisy (noise + voice) to enhanced (denoised). (Of course, only pesq indicators were analyzed.)

I wonder if your comparison is correct. I think I made a mistake just after conducting the research.

Here is a part of my code.

And this is the error.

Here is the data detail.

Hi, great package.
I just have one question, I guess the composite function outputs SIG, BAK and OVL measurements, but in what order?

Thanks for this convenient python implementation for quality measures.
However,I am confused about the composite terms. Can you point out where I can get more information about the linear coefficients of CSIG,CBAK and COVL?Because the coefficients do not correspond to those reported in this paper https://ieeexplore.ieee.org/document/4389058 (table 7).

In the test example jupyter notebook, for some metrics the enhanced result is smaller than noisy one's, while some's are larger; e.g, PESQ should be the larger the better, but it's not the case in the demo.

and the scale/range of the result isn't mentioned in readme, I know some lie in 0 to 5, but not familiar with others.

I install pysepm just like what you say in readme. However, when I import pysepm, python tells me that not find module. I am worry about it, could you help me?

May I ask that "dB" is the unit of Frequency-weighted Segmental SNR?
Thanks

hi, can pesq be used as a loss? i.e. if pesq is differentiable and pass gradient flow?
for example, is si-snr is implemented by pytorch, it can be used as a loss, while it can evaluate the performance.

Thank you for making a useful library.
During the experiment, I found an error in the llr calculation of composite funtion.
I attached samples.
https://drive.google.com/drive/folders/1FqArlwPXnLHQ53Kn99oP24IS7XoaBCNl?usp=sharing

238.wav is well calculated by CSIG, CBAK, COVL
However, CSIG and COVL are not properly calculated(outputs are 1.0) for the 101.wav
As a result of the analysis, the llr_mean of the composite function had an 'inf' value.

So I made some corrections by comparing with matlab composite function.

I modified line181 in qualityMeasures.py as below.
clean_speech_framed=extract_overlapped_windows(clean_speech+eps,winlength,winlength-skiprate,hannWin)
processed_speech_framed=extract_overlapped_windows(processed_speech+eps,winlength,winlength-skiprate,hannWin)

Still the problem has not been solved the issue and lines 151-154 have been modified as follows.
if np.abs(E[i]) < eps:
rcoeff[i]=(R[i+1] - sum_term) / (E[i])
else:
rcoeff[i]=(R[i+1] - sum_term) / (E[i])

Then, We have a value that is almost similar to the result obtained through matlab, but it still does not exactly match (to the second decimal place).

Since my modifications have not been verified, I think further modifications will be necessary.