Data-Centric AI for Signal Processing Applications

For some applications like autonomous driving, speech recognition, or machine translation, the adoption of AI can count on large datasets and abundant research. In those domains, investments most often focus on improving system performance through the design of ever more complex machine learning and deep learning models. On the other hand, in most industrial signal processing applications, data tend to be scarce and noisy, tailored models very rare, and traditional AI expertise hard to find.

This talk focuses on how data-centric workflows driven by domain-specific expertise can be used to significantly improve model performance and enable the adoption of AI in real-world applications. Learn more about signal data and specific recipes related to improving data and label quality, reducing variance and dimensionality, and selecting optimized feature-space representations and signal transformations. Explore popular simulation-based methods for data synthesis and augmentation and present the latest options for selecting suitable AI models to use as starting point.