Error efficient LOB-based approximate multipliers for error-tolerant applications

Approximate computing (AC) was positioned at the forefront of research in the field of error-tolerant applications. One key facet of AC was the use of approximate arithmetic functions, which offered significant reductions in delay, area, and power consumption at the expense of accuracy. Among these arithmetic functions, multiplication was extensively employed and played a pivotal role in error-tolerance applications. However, as the bit width increased, the design metrics and accuracy of existing multiplication designs tended to reduce. In this paper, novel architectures for leading one-bit-based approximate multipliers (LOBAMs) were proposed, aimed at improving both accuracy and design metrics. This paper focused on 8 × 8 and 16 × 16 approximate multipliers (AMs) designed in 90 nm complementary metal oxide semiconductor technology. The simulation results confirmed that LOBAMs outperformed existing AMs, reducing mean relative error distance, mean error distance, worst-case of error, normalized error distance, and error distance by an average of 74.59%, 80.75%, 41.06%, 84.19%, and 72.3%, respectively. Furthermore, when the proposed LOBAMs were embedded into an image smoothing filter, they demonstrated superior performance in terms of peak signal-to-noise ratio and structural similarity index metric compared to prior AMs. Finally, the proposed LOBAMs were exhibited remarkable advancements in both accuracy and design metrics when compared to existing AMs. This work underscored the potential of LOBAMs to revolutionize AC and contribute to more efficient and accurate error-tolerant systems.