Unify classical signal processing and modern AI under one paradigm
Leading machine learning architectures--CNNs, Transformers, Graph Neural Networks--are fundamentally signal-processing chains, yet most CS and AI curricula omit formal signal theory. Signal Processing for Modern Informatics: A Bridge between Theory and Modern Artificial Intelligence closes that gap. It re-examines discrete-time signal processing from sampling through filtering, then systematically places each concept within current AI practice and deep model design.
The book provides an integrated treatment tying Hilbert transforms, STFT, wavelets, Graph Fourier Transform, CNN filters, and attention heads together under a unified signal-processing framework. It compares model-based versus data-driven methods in depth, with guidance on constructing hybrid models. Coverage extends to empirical mode decomposition, singular spectrum analysis, and feature extraction for audio and image data.
Readers will also find:
- End-of-chapter code snippets executed on publicly available datasets, enabling immediate replication and experimentation with each technique
- Detailed treatment of graph signal processing and filtering methods applied to robotics and AI system design
- Coverage of sampling, quantization, and linear time-invariant systems grounded in their direct relevance to deep learning pipelines
- Practical examples demonstrating the Fourier Transform and FFT applied to real-world signal analysis and frequency-domain applications
- Analysis of how convolutional neural network architectures and transformer attention mechanisms relate directly to classical signal structures
Signal Processing for Modern Informatics serves professors, graduate students, and senior undergraduates in DSP and machine learning courses, as well as researchers and industry professionals seeking a unified conceptual framework. It is also suited for professional continuing-education programs in deep learning for audio, images, and graph signal processing.
Vincenzo Dentamaro, PhD, is an Assistant Professor at the University of Bari and co-author of more than 50 peer-reviewed publications. His industrial collaborations include IBM, where he holds one patent, and the Italian AI start-up Nextome, which he co-founded. He is a co-founder of Geodesia.ai, an AI research lab based in San Francisco whose first product, G-1, provides real-time, model-agnostic validation and auditable evidence for enterprise deployments of large language models. He is a Member of IEEE, with research interests spanning machine learning for LLMs safety and auditability, healthcare signals, pattern recognition, and indoor positioning.
