2026 ASEE Annual Conference & Exposition

Signals and Systems is a foundational course in electrical and computer engineering, yet students often perceive it as abstract and disconnected from modern engineering practice. While traditional instruction emphasizes mathematical tools such as convolution, Fourier transforms, and LTI system analysis, students frequently struggle to apply these concepts to emerging technologies like Artificial Intelligence (AI). This Work-in-Progress (WIP) paper describes the development and initial implementation of three AI-enhanced learning modules designed to bridge this gap by framing neural network architectures as a natural extension of classical signal theory.

The proposed modules include: (1) visualizing linear neural networks and encoder-decoder structures from an LTI system perspective; (2) implementing circulant matrices and convolutional layers using Python and MATLAB; and (3) a comparative ``competition" experiment between classical Butterworth filters and denoising autoencoders. At this stage, the theoretical framework and module content have been established, and pilot data collection is currently underway.

The proposed evaluation strategy utilizes a multi-method approach to measure pedagogical impact. Conceptual mastery will be assessed through pre- and post-tests aligned with the Signals and Systems Concept Inventory (SSCI), while student engagement and perceived career relevance will be measured via a 5-point Likert scale survey. This paper details the design of the modules, the mathematical isomorphisms used to connect LTI systems to neural layers, and the preliminary qualitative observations from the initial pilot group. Final results and a comprehensive statistical analysis of the student performance data will be presented in a future full paper.