Scholarlink

Abstract <p>This study proposes a power-efficient analog design methodology for implementing the learning vector quantization (LVQ) algorithm using 45 nm CMOS technology in the Cadence design environment. The architecture is built around a highly optimized Euclidean distance circuit (EDC), supported by current mirrors (CM) for accurate current replication and a loser-take-all (LTA) circuit for final classification. Designed to operate at an ultralow supply voltage of 0.3 V, the system significantly reduces power consumption while maintaining high computational performance. The proposed design is validated through extensive Cadence-based simulations in the context of brain tumor classification, achieving a classification accuracy of 97%. Results demonstrate the effectiveness of the approach in delivering accurate and energy-efficient analog computation. Comparative analysis further confirms its superiority over existing techniques in terms of both power efficiency and classification reliability, making it well-suited for low-power, edge-based biomedical applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Low-Power Analog Circuit for Learning Vector Quantization-Based Brain Tumor Classification Using CMOS 45 nm Technology

  • 期刊论文

  • 发表日期：

    2025年10月1日

查看全文
  • A. Arunkumar Gudivada,
  • Sayedu Khasim Noorbasha,
  • Sd. K. Muskan

摘要

Abstract

This study proposes a power-efficient analog design methodology for implementing the learning vector quantization (LVQ) algorithm using 45 nm CMOS technology in the Cadence design environment. The architecture is built around a highly optimized Euclidean distance circuit (EDC), supported by current mirrors (CM) for accurate current replication and a loser-take-all (LTA) circuit for final classification. Designed to operate at an ultralow supply voltage of 0.3 V, the system significantly reduces power consumption while maintaining high computational performance. The proposed design is validated through extensive Cadence-based simulations in the context of brain tumor classification, achieving a classification accuracy of 97%. Results demonstrate the effectiveness of the approach in delivering accurate and energy-efficient analog computation. Comparative analysis further confirms its superiority over existing techniques in terms of both power efficiency and classification reliability, making it well-suited for low-power, edge-based biomedical applications.