Published Paper Details:

Edge intelligence enables embedded systems to perform real-time decision making without continuous dependence on cloud infrastructure. This study presents a lightweight inference framework for the ARM Cortex-M3-based LPC1768 microcontroller using analog sensor signals acquired through the on-chip analog-to-digital converter (ADC). In contrast to conventional fixed-threshold approaches, the proposed method applies a parameterized linear decision model that is computationally inexpensive and suitable for resource-constrained hardware. The framework integrates ADC-based signal acquisition, sample averaging for noise mitigation, and deterministic inference logic to support low-latency operation. Experimental observations indicate that the proposed implementation can provide responsive on-device classification with modest memory requirements, making it suitable for low-power sensing applications. The work highlights a practical pathway toward deployable TinyML-inspired inference on legacy microcontroller platforms.