ProcessAudio

Libreria python para hacer data augmentation en audios y/o extraer caracteristicas a audios

---

Installation

pip install ProcessAudio

---

Description

A ProcessAudio object should be created and use its attributes.

This library have tree main functions:

• Features: Extract features from audio
• AudioAugmentation: Augment audio in different ways
• AllDataAugmentation: Augment audio in different ways and extract features
• Util: Read audio and denoise audio
• Split: Split audio in n seconds or at desired cut points
• Graph: Graph spectrogram or log_mel_spectrogram for an audio

---

Back To Top ↥

Features methods

• set_data(data_audio:str="<path_audio_file>): Set data to extract features
• get_croma(): Extract croma features
• get_mfcc(): Extract mfcc features
• get_rmse(): Extract rmse features
• get_centroide_espectral(): Extract spectral centroid features
• get_rolloff(): Extract spectral rolloff features
• get_cruce_por_cero(): Extract zero crossing rate features
• get_ancho_banda_espectral(): Extract spectral bandwidth features
• get_tonnetz(): Extract tonnetz features
• build_basic(): Extract a basic features in a list

---

Back To Top ↥

AudioAugmentation methods

• loudness(): Apply loudness to audio file creating a new data
• add_mask(): Apply mask to audio file creating a new data
• pitch(): Apply pitch to audio file creating a new data
• get_original(): Get original audio file
• add_crop(): Apply crop to audio file creating a new data
• add_noise(): Apply noise to audio file creating a new data
• add_noise2(): Apply noise to audio file creating a new data
• shift(): Apply shift to audio file creating a new data
• stretch(): Apply stretch to audio file creating a new data
• speed(): Apply speed to audio file creating a new data
• normalizer(): Apply normalizer to audio file creating a new data
• polarizer(): Apply polarizer to audio file creating a new data
• write_audio_file(): Write audio file
• plot_time_series(): Plot time series

---

Back To Top ↥

AllDataAugmentation methods

• build_all(extract_features: bool): Augment audio and extract features if extract_features is True

---

Back To Top ↥

Util methods

• read_audio(file_path: str, force_convert_wav: bool): Read Read audio, if the format isn't wav, the method convert that before to read
• audio_convert_wav(audio_path: str, output_path: str): Convert audio to wav format
• denoise_audio(data: np.array, sr: int): remove the noise of audio data

---

Back To Top ↥

Split methods

• split(self, start: int, end: int, save: bool): Split audio in start to end, if you need seconds, start and end have to multiples of 1000
• split_by_seconds(self, seconds: int, save: bool): Cut audio en segments of parameter and save each one

---

Back To Top ↥

Graph methods

• spectrogram(data: np.array, sr: int, output_path: str, title: str): Create the spectrogram for audio data
• log_mel_spectrogram(data: np.array, sr: int, output_path: str, title: str): Create log mel spectrogram for audio data

---

Back To Top ↥

Usage

Example Features

import os
from ProcessAudio.Features import Features

filepath = os.path.dirname(os.path.abspath(__file__)) + os.sep
path_file = filepath + "demo" + os.sep + "dat_92.wav"

features = Features()
features.set_data(path_file)
DATA = features.build_basic() # Extract all features
print(DATA)
print(len(DATA))  

---

Back To Top ↥

Example AudioAugmentation

import os
from ProcessAudio.AudioAugmentation import AudioAugmentation

filepath = os.path.dirname(os.path.abspath(__file__)) + os.sep
path_file = filepath + "demo" + os.sep + "dat_92.wav"
folder_save = filepath + "new_audios" + os.sep

aumentation = AudioAugmentation(audio_file=path_file, save=folder_save)
audio_con_ruido = aumentation.add_noise(factor_ruido=0.05)
audio_normalizer = aumentation.normalizer()
audio_loudness = aumentation.loudness()

---

Back To Top ↥

Example AllDataAugmentation

import os
from ProcessAudio.AllDataAugmentation import AllDataAugmentation

filepath = os.path.dirname(os.path.abspath(__file__)) + os.sep
path_file = filepath + "demo" + os.sep + "dat_92.wav"
folder_save = filepath + "new_audios" + os.sep

aumentation = AllDataAugmentation(path_file, path_save=folder_save, label=["cero", "uno"])
data, label = aumentation.build_all(extract_features=True)
print(len(data), len(label))
print(len(data[0]), label[0])

---

Back To Top ↥

Citing

If you want to cite ProcessAudio in an academic paper, there are two ways to do it.

• APA:

  WISROVI, W.S.R.V. (2022). Python library to augment audio data and/or extract audio features (Version 0.22.11) [Computer Software]. https://github.com/wisrovi/ProcessAudio

• BibTex:

  @software{WISROVI_Instrument_Classifier_2022, author = {WISROVI, William Steve Rodríguez Villamizar}, month = {10}, title = {{Python library to augment audio data and/or extract audio features}}, URL = {https://github.com/wisrovi/ProcessAudio}, version = {0.22.11}, year = {2022} }

---

Back To Top ↥

License

GPLv3 License

---

Back To Top ↥

Support: