Skip to main content

Python Wrapper for PESQ Score (narrow band and wide band)

Project description

python-pesq

PESQ (Perceptual Evaluation of Speech Quality) Wrapper for Python Users

This code is designed for numpy array specially.

Requirements

C compiler
numpy
cython

Build and Install

$ 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

Install with pip

# 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

Usage for narrow-band and wide-band Modes

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, 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'))

Correctness

The correctness is verified by running samples in audio folder.

PESQ computed by this code in wide band mode is 1.0832337141036987

PESQ computed by this code in narrow band mode is 1.6072081327438354

Note

Sampling rate (fs|rate) - No default. Must select either 8000Hz or 16000Hz.

Note there is narrow band (nb) mode only when sampling rate is 8000Hz.

The original C soure code is modified.

Who is using python-pesq

Please click here to see these repositories, whose owners include Facebook Research, SpeechBrain, NVIDIA .etc.

Buy me a Coffee

Buy me a Coffee if my work helps you in some ways.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pesq-0.0.3.tar.gz (35.9 kB view details)

Uploaded Source

File details

Details for the file pesq-0.0.3.tar.gz.

File metadata

  • Download URL: pesq-0.0.3.tar.gz
  • Upload date:
  • Size: 35.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.4.1 importlib_metadata/3.10.0 pkginfo/1.7.0 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.59.0 CPython/3.6.13

File hashes

Hashes for pesq-0.0.3.tar.gz
Algorithm Hash digest
SHA256 93d3a6cc1de88f3c8b57f0c1bb3cbec80e7b25423cf8451dab816853bf86fde7
MD5 f6de0bb311226171791dc12b95ed0a70
BLAKE2b-256 7ccffb0b7fd0c2b51004d5357d7ffb5a964d1934012663ab91b7559e309902fc

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page