Mixing an audio file with a noise file at any Signal-to-Noise Ratio (SNR)

Sato-Kunihiko Last update: Feb 07, 2024

Mixing an audio file with a noise file at any Signal-to-Noise Ratio

The script create_mixed_audio_file.py and create_mixed_audio_file_with_soundfile.py can mix an audio file with a noise file at any Signal-to-Noise Ratio (SNR).

You can listen to the results of mixtures of the clean voice and the noise.

Installation

Python3.7+
macOS

Setup

I recommend to install Pipenv for making set-up easy before proceeding:

If you're on macOS, you can install Pipenv easily with Homebrew:

$ brew install pipenv

After installing Pipenv, you run the following command to build a virtualenv and install packages listed in Pipfile.lock.

$ pipenv install

Then, you can run the following command to activate a virtualenv.

$ pipenv shell

File format

To use this code, the format of source waveforms is as follows.

WAV file
1 channel

Additionally, the sampling rate of a clean file and that of a noise file are supposed to be the same.

Usage

There are two files for creating a mixture.

create_mixed_audio_file.py:
- Uses wave module
- Can read wav file with 16-bit PCM only
create_mixed_audio_file_with_soundfile.py:
- Uses soundfile library
- Can read wav file with various encoding types such as 16-bit PCM, 32-bit PCM, 32-bit float, and 64-bit float.

After activating a virtualenv, you can run the files to mix an audio file with a noise file at any signal-to-noise ratio.

Example of create_mixed_audio_file.py:

python create_mixed_audio_file.py --clean_file ~/workspace/audio-SNR/data/16_bit/source_clean/arctic_a0001.wav --noise_file ~/workspace/audio-SNR/data/16_bit/source_noise/ch01.wav --snr 0 --output_mixed_file ~/workspace/audio-SNR/data/16_bit/output_mixed/0dB.wav

Example of create_mixed_audio_file_with_soundfile.py:

python create_mixed_audio_file_with_soundfile.py --clean_file ~/workspace/audio-SNR/data/64_bit/source_clean/arctic_a0001_64bit.wav --noise_file ~/workspace/audio-SNR/data/64_bit/source_noise/ch01_64bit.wav --snr 0 --output_mixed_file ~/workspace/audio-SNR/data/64_bit/output_mixed/0dB.wav

Dataset

I really appreciate the following public datasets.

Voices - CMU_ARCTIC speech synthesis databases
Noises - DEMAND: a collection of multi-channel recordings of acoustic noise in diverse environments

Detail

There is a detail about create_mixed_audio_file.py on my post (in Japanese).

https://engineering.linecorp.com/ja/blog/voice-waveform-arbitrary-signal-to-noise-ratio-python/

There is a detail about create_mixed_audio_file.py on my post (in Korean).

https://engineering.linecorp.com/ko/blog/voice-waveform-arbitrary-signal-to-noise-ratio-python/

Mixing an audio file with a noise file at any Signal-to-Noise Ratio (SNR)

Mixing an audio file with a noise file at any Signal-to-Noise Ratio

Installation

Setup

File format

Usage

Dataset

Detail

🤖 Template for telegram bot using postgres, pgbouncer, redis, docker, amplitude, prometheus, grafana, CI

Data science and Machine Learning GUI programs/ desktop apps with PySimpleGUI package

ML-capsule is a Project for beginners and experienced data science Enthusiasts who don't have a mentor or guidance and wish to learn Machine learning. Using our repo they can learn ML, DL, and many related technologies with different real-world projects and become Interview ready.

The easiest way to use Machine Learning. Mix and match underlying ML libraries and data set sources. Generate new datasets or modify existing ones with ease.

Text analytics for LLM apps. PostHog for prompts. Extract evaluations, intents and events from text messages. phospho leverages LLM (OpenAI, MistralAI, Ollama, etc.)

a flask profiler which watches endpoint calls and tries to make some analysis.

Analytics services for Django projects

A flexible, easy to use, automation framework allowing users to integrate their capabilities and devices to cut through the repetitive, tedious tasks slowing them down. #nsacyber

Apache Superset is a Data Visualization and Data Exploration Platform

Beginner-friendly collection of Python notebooks for various use cases of machine learning, deep learning, and analytics. For each notebook there is a separate tutorial on the relataly.com blog.