Overview
Are you a signal processing engineer working on DSP algorithms, product development, or signal measurements? Are you trying to use more machine learning or deep learning in your projects?
In this session you will learn the fundamental ideas around the application of deep learning to audio, speech, and acoustics.
We start by discussing the use of established pre-trained deep learning models to solve a few complex but standard problems. We then show how to design, train, and deploy a complete speech command recognition system from scratch using MATLAB, starting from a reasonably large dataset and ending up with a real-time prototype.
Highlights
- Using application-specific pre-trained networks (e.g. YAMNet, CREPE) with a single line of MATLAB code
- Extracting signal features with established algorithms (e.g. Mel Spectrograms, MFCC) and pre-trained deep networks (e.g. VGGish, OpenL3)
- Designing and training deep networks (including with CNN, LSTM, and GRU layers) using GPUs
- Generating embedded C++ for feature extraction and network prediction
- Creating real-time deep learning prototypes with RaspberryPi boards
About the Presenter
Gabriele Bunkheila is a senior product manager at MathWorks, where he coordinates the strategy of MATLAB toolboxes for audio and DSP. 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.
