Pure Integer Hyperdimensional Computing (HDC) for Low-Power Embedded Systems

Hyperdimensional Computing (HDC) models cognitive association using high-dimensional random vectors. A primary implementation of HDC is Fourier Holographic Reduced Representation (FHRR), which employs continuous phase angles represented as floating-point complex numbers. In this paper, we propose a pure integer reformulation of FHRR designed for low-power microcontrollers and digital signal processors. Phase angles are quantized as 8-bit integers mapped over a modular ring Z_256. Vector binding is implemented as modular addition, and bundling is solved through majority coordinate voting. Furthermore, we use the CORDIC (Coordinate Rotation Digital Computer) algorithm and integer lookup tables to resolve trigonometric functions without floating-point units.