keras-audio

Implementation and examples for audio signal processing with using Keras.

This repository is supplementary for my presentation: "Hand-On Audio for Deep Learning (Practice)"

Sample audio

In data/audio_files, 10 audio samples from Freesound.

Example codes

There are some example codes for treating audio data. It basically uses librosa for I/O, sounddevice and IPython.display.Audio for playback. In this repo, all example codes will use them by default, except for basic codes for other libraries.

You can choose another options like as below:

for I/O:

• wave (Python3)
• scipy.io
• pydub

for playback:

• pyaudio
• scikit-sound
• pygame.mixer

01. Basic

Example codes for basic load/save and play to listen.

• load/save wave
• playback
• plot waveform

02. Preprocessing

Example codes for feature extraction, converting, and feature visualization.

• STFT (spectrogram)
• magnitude & phase
• mel-spectrogram
• MFCC (mel-frequency cepstral coefficient)
• inverse STFT
• RMS normalization

03. Augmentation