PESQ (Perceptual Evaluation of Speech Quality) Wrapper for Python Users
This code is designed for numpy array specially.
C compiler
numpy
cython
$ git clone https://github.com/ludlows/python-pesq.git
$ cd python-pesq
$ pip install . # for python 2
$ pip3 install . # for python 3
$ cd ..
$ rm -rf python-pesq # remove the code folder since it exists in the python package folder
# PyPi Repository
$ pip install pesq
# The Latest Version
$ pip install https://github.com/ludlows/python-pesq/archive/master.zip
# or
$ pip3 install https://github.com/ludlows/python-pesq/archive/master.zip
Please note that the sampling rate (frequency) should be 16000 or 8000 (Hz).
And using 8000Hz is supported for narrow band only.
The code supports error-handling behaviors now.
def pesq(fs, ref, deg, mode='wb', on_error=PesqError.RAISE_EXCEPTION):
"""
Args:
ref: numpy 1D array, reference audio signal
deg: numpy 1D array, degraded audio signal
fs: integer, sampling rate
mode: 'wb' (wide-band) or 'nb' (narrow-band)
on_error: error-handling behavior, it could be PesqError.RETURN_VALUES or PesqError.RAISE_EXCEPTION by default
Returns:
pesq_score: float, P.862.2 Prediction (MOS-LQO)
"""
Once you select PesqError.RETURN_VALUES
, the pesq
function will return -1 when an error occurs.
Once you select PesqError.RAISE_EXCEPTION
, the pesq
function will raise an exception when an error occurs.
It supports the following errors now: InvalidSampleRateError
, OutOfMemoryError
,BufferTooShortError
,NoUtterancesError
,PesqError
(other unknown errors).
from scipy.io import wavfile
from pesq import pesq
rate, ref = wavfile.read("./audio/speech.wav")
rate, deg = wavfile.read("./audio/speech_bab_0dB.wav")
print(pesq(rate, ref, deg, 'wb'))
print(pesq(rate, ref, deg, 'nb'))
def pesq_batch(fs, ref, deg, mode='wb', n_processor=None, on_error=PesqError.RAISE_EXCEPTION):
"""
Running `pesq` using multiple processors
Args:
on_error:
ref: numpy 1D (n_sample,) or 2D array (n_file, n_sample), reference audio signal
deg: numpy 1D (n_sample,) or 2D array (n_file, n_sample), degraded audio signal
fs: integer, sampling rate
mode: 'wb' (wide-band) or 'nb' (narrow-band)
n_processor: None (without multiprocessing) or number of processors
on_error:
Returns:
pesq_score: list of pesq scores, P.862.2 Prediction (MOS-LQO)
"""
this function uses multiprocessing
features to boost time efficiency.
When the ref
is an 1-D numpy array and deg
is a 2-D numpy array, the result of pesq_batch
is identical to the value of [pesq(fs, ref, deg[i,:],**kwargs) for i in range(deg.shape[0])]
.
When the ref
is a 2-D numpy array and deg
is a 2-D numpy array, the result of pesq_batch
is identical to the value of [pesq(fs, ref[i,:], deg[i,:],**kwargs) for i in range(deg.shape[0])]
.
The correctness is verified by running samples in audio folder.
PESQ computed by this code in wideband mode is 1.0832337141036987
PESQ computed by this code in narrowband mode is 1.6072081327438354
Sampling rate (fs|rate) - No default. Must select either 8000Hz or 16000Hz.
Note there is narrowband (nb) mode only when sampling rate is 8000Hz.
The original C source code is modified.
Please click here to see these repositories, whose owners include Facebook Research
, SpeechBrain
, NVIDIA
.etc.
This work was funded by the Natural Sciences and Engineering Research Council of Canada.
This work was also funded by the Concordia University, Montreal, Canada.