On the Design of a Two-Dimensional Sensor Calibration Processor Using a Variable Polynomial Computation for Enhancing Sensor Non-Linearity

This paper presents a hardware architecture for two-dimensional sensor calibration based on a segmented polynomial computation framework, which divides the input space into multiple regions and applies a locally optimized calibration function to each. To determine optimal segment boundaries, the system incorporates an Adaptive Segmentation Module (ASM) that automatically analyzes sensor characteristics and selects boundary configurations that minimize global calibration error. At the algorithmic level, the proposed method integrates a Simplified Progressive Polynomial Calibration (SPPC) technique, which addresses the exponential growth in polynomial order observed in conventional Progressive Polynomial Calibration (PPC) methods as the number of calibration points increases. SPPC mitigates this complexity through a linear-order growth strategy while preserving high calibration accuracy within each segment. The system is implemented on a RISC-V-based System-on-Chip (SoC) equipped with a dedicated Sensor Calibration Module (SCM) optimized for 8-bit signed fixed-point arithmetic. The proposed approach was validated through MATLAB simulations and RTL-level hardware implementation. Experimental results demonstrated up to an 80% reduction in execution time compared to conventional PPC and a 2.35% improvement in calibration accuracy through segmented calibration. Notably, the ASM-based segmentation achieved up to a 5.87% reduction in calibration error compared to non-segmented methods, even under varying input conditions, effectively reducing the need for manual tuning. These results confirm that the proposed architecture provides a high-performance, hardware-efficient, and adaptive solution for real-time calibration of sensors affected by nonlinearities and cross-sensitivity, making it well-suited for a wide range of embedded sensing applications.