Learnable FIR Filter-banks Time-Convolutional Layers

[1] Ahmed Imtiaz Humayun, Md. Tauhiduzzaman Khan, Shabnam Ghaffarzadegan, Zhe Feng and Taufiq Hasan, “An Ensemble of Transfer, Semi-supervised and Supervised Learning Methods for Pathological Heart Sound Classification”, Interspeech 2018, Hyderabad, India.

[2] Ahmed Imtiaz Humayun, Shabnam Ghaffarzadegan, Zhe Feng and Taufiq Hasan, “Learning Front-end Filter-bank Parameters using Convolutional Neural Networks for Abnormal Heart Sound Detection”, Proc. EMBC 2018, Hawaii, USA.

Keras Wrappers for Learnable - FIR built on top of Tensorflow backend

Contents:

• Learnable FIR
• Learnable Linear Phase FIR
• Learnable Zero Phase FIR
• Keras DCT Layer

Requirements:

• Keras
• Tensorflow

Usage:

from custom_layers import Conv1D_linearphase

Use it like any convolutional layer in Keras.

from keras.models import Model
from keras.layers import Input, Dense, Conv1D
from custom_layers import Conv1D_linearphase

x = Input(shape=(2500,1))
x = Conv1D_linearphase(3, 61, padding='valid') # Filterbank of 3 filters, outputs stacked along channel axis
x = Conv1D(8, 5, padding='same', activation='relu')
x = Dense(32)(x)
model = Model(inputs=a, outputs=b)

Cite:

@article{humayun2018learning,
  title={Learning Front-end Filter-bank Parameters using Convolutional Neural Networks for Abnormal Heart Sound Detection},
  author={Humayun, Ahmed Imtiaz and Ghaffarzadegan, Shabnam and Feng, Zhe and Hasan, Taufiq},
  journal={Proc. EMBC 2018},
  year={2018}
}

@inproceedings{Humayun2018,
  author={Ahmed Imtiaz Humayun and Md. Tauhiduzzaman Khan and Shabnam Ghaffarzadegan and Zhe Feng and Taufiq Hasan},
  title={An Ensemble of Transfer, Semi-supervised and Supervised Learning Methods for Pathological Heart Sound Classification},
  year=2018,
  booktitle={Proc. Interspeech 2018},
  pages={127--131},
  doi={10.21437/Interspeech.2018-2413},
  url={http://dx.doi.org/10.21437/Interspeech.2018-2413}
}