Torch-Yin

This package implements the Yin pitch estimation algorithm described by A De Cheveigné, et al for the PyTorch deep learning framework. It is based on the excellent NumPy implementation by Patrice Guyot, which has been extended for full vectorization and to support batched computation.

Status

Install with pip

pip install torch-yin

Usage

Here we estimate the fundamental frequency of a simple 1 second sinusoid at 440Hz:

import torch
import torchyin

FS = 48000

y = torch.sin(2 * torch.pi * 440 / FS * torch.arange(FS))

pitch = torchyin.estimate(y, sample_rate=FS)

pitch[0]

tensor(440.3669)

Pitch can also be calculated for batches of signals, shaped [batch, samples]. In this example, we create a batch of signals for the 88 standard piano keys using broadcasting, shaped [88, 48000].

import torch
import torchyin

FS = 48000

f = 2 ** ((torch.arange(88) - 48) / 12) * 440
t = torch.arange(FS)
y = torch.sin(2 * torch.pi * f.unsqueeze(1) / FS * t.unsqueeze(0))

pitch = torchyin.estimate(
    y,
    sample_rate=FS,
    pitch_min=20,
    pitch_max=5000,
)

pitch[:, 0]

tensor([  27.5072,   29.1439,   30.8682,   32.6975,   34.6570,   36.6973,
          38.8979,   41.2017,   43.6364,   46.2428,   48.9796,   51.8919,
          54.9828,   58.2524,   61.6967,   65.3951,   69.2641,   73.3945,
          77.7958,   82.4742,   87.2727,   92.4855,   97.9592,  103.8961,
         110.0917,  116.5049,  123.3933,  130.7902,  138.7283,  146.7890,
         155.3398,  164.9485,  174.5455,  185.3282,  195.9184,  207.7922,
         220.1835,  233.0097,  247.4227,  262.2951,  277.4566,  294.4785,
         311.6883,  328.7671,  350.3650,  369.2308,  393.4426,  413.7931,
         440.3669,  466.0194,  494.8453,  521.7391,  551.7241,  585.3658,
         623.3766,  657.5342,  695.6522,  738.4615,  786.8852,  827.5862,
         872.7272,  941.1765,  979.5918, 1043.4783, 1116.2791, 1170.7317,
        1230.7693, 1333.3334, 1411.7648, 1500.0000, 1548.3871, 1655.1724,
        1777.7778, 1846.1539, 2000.0000, 2086.9565, 2181.8184, 2400.0000,
        2526.3159, 2666.6667, 2823.5295, 3000.0000, 3200.0002, 3428.5715,
        3428.5715, 3692.3079, 4000.0000, 4363.6367])

For more information and detailed parameter descriptions, please check out this blog post, see the module documentation, or run help(torchyin).

Development

Setup

The following script creates a virtual environment using pyenv for the project and installs dependencies.

pyenv install 3.9.10
pyenv virtualenv 3.9.10 torch-yin
pip install -r requirements.txt

You can then run tests, etc. follows:

pytest --cov=torchyin
black .
flake8 .
mypy torchyin

These can also be used with the pre-commiit library to run all checks at commit time.

Deployment

The project uses setup.py for installation and is deployed to PyPI. The source distribution can be built for deployment with the following command:

python setup.py clean --all
rm -r ./dist
python setup.py sdist

The distribution can then be uploaded to PyPI using twine.

twine upload --repository-url=https://upload.pypi.org/legacy/ dist/*

For deployment testing, the following command can be used to upload to the PyPI test repository:

twine upload --repository-url=https://test.pypi.org/legacy/ dist/*

License

Copyright © 2022 Brent M. Spell

Licensed under the MIT License (the "License"). You may not use this package except in compliance with the License. You may obtain a copy of the License at

https://opensource.org/licenses/MIT

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.