Introduction to Deep Learning for Audio and Speech Applications

Gabriele Bunkheila, MathWorks

Overview

Are you an audio or speech processing engineer working on product development or DSP algorithms and looking to integrate AI capabilities within your projects? In this session you will learn the basics of deep learning for audio applications by walking through a detailed example of speech classification, entirely based on MATLAB code. We will cover creating and accessing labeled data, using time-frequency transformations, extracting features, designing and training deep neural network architectures, and testing prototypes on real-time audio. We will also discuss interoperability with other popular deep learning tools, including exploiting available pre-trained networks.

Highlights

Acquiring, segmenting and labeling audio recordings and ingesting existing datasets
Extracting standard speech and audio features and using 2D time-frequency representations
Designing and analyzing deep networks and exchanging models with other popular frameworks (e.g. via ONNX)
Accelerating computations using GPUs and prototyping trained models on real-world signals

About the Presenter

Gabriele Bunkheila is a senior product manager at MathWorks for audio and DSP applications. After joining MathWorks in 2008, he worked as a signal processing application engineer for several years, supporting MATLAB and Simulink users across industries from algorithm design to real-time implementations. Before MathWorks, he held a number of research and development positions, and he was a lecturer of sound theory and technologies at the national film school of Rome. He has a master’s degree in physics and a Ph.D. in communications engineering.