Scholarlink

With the increasing need for IoT-based solutions in healthcare, there is a critical need for energy-efficient, low-power processors capable of executing machine learning (ML) models for real-time health monitoring at the edge level. This paper presents the first-of-its-kind analysis of the Indigenous RISC-V Shakti processor, specifically evaluating its potential as an edge-based health monitoring solution through machine learning-driven diabetic prediction. The Shakti processor is assessed using the Shakti SDK framework. A performance comparison is done with the PYNQ-Z2’s Zynq processor regarding execution efficiency and latency of various machine learning models under different optimization techniques, including quantization and dimensionality reduction via Linear Discriminant Analysis. The Zynq processor is evaluated using Python language in the PYNQ framework. The Shakti’s use of LDA and quantization optimization techniques led to significant improvements in the performance of ANN and SVM models, reducing the processing time by nearly 30% compared to the ZYNQ SoC. This research underscores the Shakti processor’s viability for low-cost, real-time diabetic prediction at the edge level, advancing scalable healthcare solutions and contributing to the development of self-reliable, indigenous IoT devices.

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

Machine Learning-Based Diabetic Prediction on Indigenous Shakti Processor for Edge Health Monitoring

  • 章节

  • 发表日期：

    2026年

查看全文
  • P. C. Muhammed Raees,
  • M. R. Akshayraj,
  • G. Lakshminarayanan,
  • Varun P. Gopi,
  • G. R. Gangadharan

摘要

With the increasing need for IoT-based solutions in healthcare, there is a critical need for energy-efficient, low-power processors capable of executing machine learning (ML) models for real-time health monitoring at the edge level. This paper presents the first-of-its-kind analysis of the Indigenous RISC-V Shakti processor, specifically evaluating its potential as an edge-based health monitoring solution through machine learning-driven diabetic prediction. The Shakti processor is assessed using the Shakti SDK framework. A performance comparison is done with the PYNQ-Z2’s Zynq processor regarding execution efficiency and latency of various machine learning models under different optimization techniques, including quantization and dimensionality reduction via Linear Discriminant Analysis. The Zynq processor is evaluated using Python language in the PYNQ framework. The Shakti’s use of LDA and quantization optimization techniques led to significant improvements in the performance of ANN and SVM models, reducing the processing time by nearly 30% compared to the ZYNQ SoC. This research underscores the Shakti processor’s viability for low-cost, real-time diabetic prediction at the edge level, advancing scalable healthcare solutions and contributing to the development of self-reliable, indigenous IoT devices.