Signal Processing Techniques for Deep Learning on Sensor Data
Avinash Nehemiah, Product Manager for Deep Learning, Computer Vision and Automated Driving, MathWorks

Our world is filled with data in the form of signals. This overwhelming variety of data makes it necessary for us to process the signals in ways that can extract and enhance the meaningful information. Further, developing models that can automatically identify the essential patterns in signals for the purposes of enabling analytics is not a straightforward task. In this session, you will see how MATLAB makes it easy for you to build analytics on signals without going through the manual feature extraction process.

Specifically, we will go over how you can use sharp time-frequency techniques to enhance the information present in the signals and subsequently use deep Convolutional Neural Networks to enable you to build predictive models which can be used for tasks like signal classification. Using real 1-D signals, we will go over the following topics:

• Quick primer on Time-Frequency representations for Signals
• Leverage pre-trained CNN models for obtaining insights
• Leverage advanced signal processing techniques for pre-processing
• Explore other Deep Learning techniques.