audioFlux

A library for audio and music analysis, feature extraction.

Table of Contents

• Overview
  • Description
  • Functionality
    • Transform
    • Feature
    • MIR
• Quickstart
  • Mel & MFCC
  • CWT & Synchrosqueezing
  • Other examples
• Installation
  • Python Package Intsall
  • iOS build
  • Android build
  • Building from source
• Benchmark
  • Server performance
  • Mobile performance
• Documentation
• Contributing
• Citing
• License

Overview

Description

In audio domain, feature extraction is particularly important for Audio Classification, Speech enhancement, Audio/Music Separation,music-information-retrieval(MIR), ASR and other audio task.

In the above tasks, mel spectrogram and mfcc features are commonly used in traditional machine-learning based on statistics and deep-learning based on neural network.

audioFlux provides systematic, comprehensive and multi-dimensional feature extraction and combination, and combines various deep learning network models to conduct research and development learning in different fields.

Can be used for deep learning, pattern recognition, signal processing, bioinformatics, statistics, finance, etc.

Functionality

audioFlux is based on the design of data flow. It decouples each algorithm module structurally, and it is convenient, fast and efficient to extract features from large batches.The following are the main feature architecture diagrams, specific and detailed description view the documentation.

The main functions of audioFlux include transform, feature and mir modules.

1. Transform

In the time–frequency representation, main transform algorithm:

• BFT   -   Based Fourier Transform, similar short-time Fourier transform.
• NSGT -   Non-Stationary Gabor Transform.
• CWT   -   Continuous Wavelet Transform.
• PWT   -   Pseudo Wavelet Transform.

The above transform supports all the following frequency scale types:

• Linear - Short-time Fourier transform spectrogram.
• Linspace - Linspace-scale spectrogram.
• Mel - Mel-scale spectrogram.
• Bark - Bark-scale spectrogram.
• Erb - Erb-scale spectrogram.
• Octave - Octave-scale spectrogram.
• Log - Logarithmic-scale spectrogram.

The following transform are not supports multiple frequency scale types, only used as independent transform:

• CQT -   Constant-Q Transform.
• VQT -   Variable-Q Transform.
• ST   -   S-Transform/Stockwell Transform.
• FST -   Fast S-Transform.
• DWT -   Discrete Wavelet Transform.
• WPT -   Wave Packet Transform.
• SWT -   Stationary Wavelet Transform.

Detailed transform function, description, and use view the documentation.

The synchrosqueezing or reassignment is a technique for sharpening a time-frequency representation, contains the following algorithms:

• reassign - reassign transform for STFT.
• synsq - reassign data use CWT data.
• wsst - reassign transform for CWT.

2. Feature

The feature module contains the following algorithms:

• spectral - Spectrum feature, supports all spectrum types.
• xxcc - Cepstrum coefficients, supports all spectrum types.
• deconv - Deconvolution for spectrum, supports all spectrum types.
• chroma - Chroma feature, only supports CQT spectrum, Linear/Octave spectrum based on BFT.

3. MIR

The mir module contains the following algorithms:

• pitch - YIN, STFT, etc algorithm.
• onset - Spectrum flux, novelty, etc algorithm.
• hpss - Median filtering, NMF algorithm.

Quickstart

To install the audioFlux package, Python >=3.6, using the released python package:

pip install audioflux

Mel & MFCC

Mel spectrogram and Mel-frequency cepstral coefficients

import numpy as np
import audioflux as af

import matplotlib.pyplot as plt
from audioflux.display import fill_spec

# Get a 220Hz's audio file path
sample_path = af.utils.sample_path('220')

# Read audio data and sample rate
audio_arr, sr = af.read(sample_path)

# Extract mel spectrogram
spec_arr, mel_fre_band_arr = af.mel_spectrogram(audio_arr, num=128, radix2_exp=12, samplate=sr)
spec_arr = np.abs(spec_arr)

# Extract mfcc
mfcc_arr, _ = af.mfcc(audio_arr, cc_num=13, mel_num=128, radix2_exp=12, samplate=sr)

# Display
audio_len = audio_arr.shape[0]
# calculate x/y-coords
x_coords = np.linspace(0, audio_len / sr, spec_arr.shape[1] + 1)
y_coords = np.insert(mel_fre_band_arr, 0, 0)
fig, ax = plt.subplots()
img = fill_spec(spec_arr, axes=ax,
                x_coords=x_coords, y_coords=y_coords,
                x_axis='time', y_axis='log',
                title='Mel Spectrogram')
fig.colorbar(img, ax=ax)

fig, ax = plt.subplots()
img = fill_spec(mfcc_arr, axes=ax,
                x_coords=x_coords, x_axis='time',
                title='MFCC')
fig.colorbar(img, ax=ax)

plt.show()

CWT & Synchrosqueezing

Continuous Wavelet Transform spectrogram and its corresponding synchrosqueezing reassignment spectrogram

import numpy as np
import audioflux as af
from audioflux.type import SpectralFilterBankScaleType, WaveletContinueType
from audioflux.utils import note_to_hz

import matplotlib.pyplot as plt
from audioflux.display import fill_spec

# Get a 220Hz's audio file path
sample_path = af.utils.sample_path('220')

# Read audio data and sample rate
audio_arr, sr = af.read(sample_path)
audio_arr = audio_arr[:4096]

cwt_obj = af.CWT(num=84, radix2_exp=12, samplate=sr, low_fre=note_to_hz('C1'),
                 bin_per_octave=12, wavelet_type=WaveletContinueType.MORSE,
                 scale_type=SpectralFilterBankScaleType.OCTAVE)

cwt_spec_arr = cwt_obj.cwt(audio_arr)

synsq_obj = af.Synsq(num=cwt_obj.num,
                     radix2_exp=cwt_obj.radix2_exp,
                     samplate=cwt_obj.samplate)

synsq_arr = synsq_obj.synsq(cwt_spec_arr,
                            filter_bank_type=cwt_obj.scale_type,
                            fre_arr=cwt_obj.get_fre_band_arr())

# Show CWT
fig, ax = plt.subplots(figsize=(7, 4))
img = fill_spec(np.abs(cwt_spec_arr), axes=ax,
                x_coords=cwt_obj.x_coords(),
                y_coords=cwt_obj.y_coords(),
                x_axis='time', y_axis='log',
                title='CWT')
fig.colorbar(img, ax=ax)
# Show Synsq
fig, ax = plt.subplots(figsize=(7, 4))
img = fill_spec(np.abs(synsq_arr), axes=ax,
                x_coords=cwt_obj.x_coords(),
                y_coords=cwt_obj.y_coords(),
                x_axis='time', y_axis='log',
                title='Synsq')
fig.colorbar(img, ax=ax)

plt.show()

Other examples

• CQT & Chroma
• Different Wavelet Type
• Spectral Features
• Pitch Estimate
• Onset Detection
• Harmonic Percussive Source Separation

More example scripts are provided in the Documentation section.

Installation

The library is cross-platform and currently supports Linux, macOS, Windows, iOS and Android systems.

Python Package Intsall

Using PyPI:

$ pip install audioflux 

Using Anaconda:

$ conda install -c tanky25 -c conda-forge audioflux

iOS build

To compile iOS on a Mac, Xcode Command Line Tools must exist in the system:

• Install the full Xcode package
• install Xcode Command Line Tools when triggered by a command or run xcode-select command:

$ xcode-select --install 

Enter the audioFlux project scripts directory and switch to the current directory, run the following script to build and compile:

$ ./build_iOS.sh

Build and compile successfully, the project build compilation results are in the build folder

Android build

The current system development environment needs to be installed android NDK, ndk version>=16,after installation, set the environment variable ndk path.

For example, ndk installation path is ~/Android/android-ndk-r16b:

$ export NDK_ROOT=~/Android/android-ndk-r16b
$ export PATH=$NDK_ROOT:$PATH

Android audioFlux build uses fftw library to accelerate performance, compile the single-floating point version for android platform. fftw lib successful compilation, copy to audioFlux project scripts/android/fftw3 directory.

Enter the audioFlux project scripts directory and switch to the current directory, run the following script to build and compile:

$ ./build_android.sh

Build and compile successfully, the project build compilation results are in the build folder

Building from source

For Linux, macOS, Windows systems. Read installation instructions:

• docs/installing.md

Benchmark

Server performance

server hardware:

- CPU: AMD Ryzen Threadripper 3970X 32-Core Processor
- Memory: 128GB

Each sample data is 128ms(sampling rate: 32000, data length: 4096).

The total time spent on extracting features for 1000 sample data.

Package	audioFlux	librosa	pyAudioAnalysis	python_speech_features
Mel	0.777s	2.967s	--	--
MFCC	0.797s	2.963s	0.805s	2.150s
CQT	5.743s	21.477s	--	--
Chroma	0.155s	2.174s	1.287s	--

Mobile performance

For 128ms audio data per frame(sampling rate: 32000, data length: 4096).

The time spent on extracting features for 1 frame data.

Mobile	iPhone 13 Pro	iPhone X	Honor V40	OPPO Reno4 SE 5G
Mel	0.249ms	0.359ms	0.313ms	0.891ms
MFCC	0.249ms	0.361ms	0.315ms	1.116ms
CQT	0.350ms	0.609ms	0.786ms	1.779ms
Chroma	0.354ms	0.615ms	0.803ms	1.775ms

Documentation

Documentation of the package can be found online:

https://audioflux.top

Contributing

We are more than happy to collaborate and receive your contributions to audioFlux. If you want to contribute, please fork the latest git repository and create a feature branch. Submitted requests should pass all continuous integration tests.

You are also more than welcome to suggest any improvements, including proposals for need help, find a bug, have a feature request, ask a general question, new algorithms. Open an issue

Citing

If you want to cite audioFlux in a scholarly work, there are two ways to do it.

• If you are using the library for your work, for the sake of reproducibility, please cite the version you used as indexed at Zenodo:

  

License

audioFlux project is available MIT License.