Deep Learning for Audio Signal Processing, with Python and Pytorch Tutorial - TEASER- AES FALL 2021

Deep Learning for Audio Signal Processing, with Python and Pytorch Examples Tutorial - TEASER- AES FALL 2021

In this tutorial, we will show some basic building blocks of deep learning, particularly for audio, from the perspective of signal processing. The idea is to show some similarities between familiar signal processing structures and deep learning architectures. For that, we use examples in Python and Pytorch.

We start with best practices for deep learning, then exploring convolutional neural networks as filter banks (analysis and synthesis) and autoencoders as a filter bank-based audio coder, and finally, we discuss recurrent neural networks as IIR (infinite impulse response) filters. This is done using audio examples and Python Pytorch program examples.

Content:

Best Practices for machine learning in audio
Specific properties of audio signals and typical features
Convolutional layers as filter banks
Autoencoders as Filter bank with optimization
Variational autoencoder as audio coder with quantization
Recurrent Neural Networks as Infinite Impulse Response filters

Detailed Content:
Introduction
- Machine Learning for Audio
- Applications for Data-Driven Techniques in Audio
- Data Science Venn Diagram
- Classification of Learning Tasks
a) Supervised Learning
b) Unsupervised Learning
c) Reinforcement Learning
- The Four Pilars of Machine Learning
Structured Development of Data-Driven Methods
- Problem Definition
- Data Preparation (Pre-Processing of Audio Signals)
a) Evaluation of Signal Quality – Capturing
-Continuous Signal
- Sampling and Quantization
b) Evaluation of Signal Quality – Analysis Techniques
- Segmentation of Signals
Connection between Audio Signal Processing and Deep Learning
- Terminology Signal Processing ---- Deep Learning
- Filtering in Audio Signal Processing
- The Filtering Operation
- Sampling Rate Changes
- PyTorch 1-Dimensional Convolution "Conv1d"
- Converting an Audio Signal as Pytorch Input
- Generating an Example Audio Signal
- Example: Audio Predictor
- A Convolutional Layer as Predictor
a) Input and Target
b) Setting up the Network
c) Run the Optimization or Training
d) Testing
e) Plotting the PyTorch Signals
f) Reading the weights
- Example: An Audio Recognizer or Classifier
a) Pre-Processing
b) The Network
c) The Network Output
d) The "Deep" Recognizer Network in Pytorch
e) The Training Set
f) The Target
g) Training
- The Autoencoder
a) The Convolutional Autoencoder
- A Filter Bank with critical Sampling
- The Convolutional Autoencoder Network
- Testing
b) The Denoising Autoencoder
c) The Variational Autoencoder (VAE)
- Similarity to Audio Coding
- Training the modified Variational Autoencoder
- Testing with a training sample
- Testing decoder with only noise in latent space
- Recurrent Neural Networks (RNN)
a) An IIR Filter
- IIR Filter Difference Equation
b) RNN Equation
c) Comparing IIR vs. RNN
d) Damped Oscillator
e) An RNN as IIR Filter
- RNN Input Format
- Training the RNN as Damped Oscillator
- Testing on the Training Signal
Summary and Conclusions

https://audioengineeringmonth2021.sched.com/event/mKSZ/deep-learning-for-audio-signal-processing-with-python-and-pytorch-examples-tutorial?iframe=no

#python #pytorch #pyTorch #PyTorch #deeplearning #machinelearming #signalprocessing #dsp #musictechnology #audioprogramming #googlecolab #jupyternotebooks