A Context-Aware Huffman Coding for Resource-Constrained Systems using NAND Flash Memory

In resource-constrained systems with limited computing and memory resources, efficient and reliable data processing is critical, especially when NAND flash memory is used for temporary data storage. Despite its high capacity and shock resistance, NAND flash memory exhibits retention errors and out-of-place write behavior, which complicate reliable usage in systems with minimal volatile memory and constrained compute capabilities. Furthermore, since data transmission often consumes significantly more energy/time than local computation, resource-constrained systems benefit from efficient data compression/encoding before transmission or storage. This article proposes a context-aware Huffman coding tailored for such environments. By exploiting context similarity across data segments, the proposed method minimizes the number of Huffman tree buildings, thereby reducing encoding overhead and improving overall system efficiency. The design requires only lightweight data structures and minimal memory, making it suitable for low-end controllers with NAND flash memory. Experimental results show that our method reduces Huffman tree buildings by 48%–49%, improves encoding time by 14%–29%, and lowers power consumption by 7%–27% compared to a traditional Huffman coding. Additionally, we introduce a compact crash recovery mechanism that ensures consistent system state after unexpected resets.