Scholarlink

Analysis of ECG recordings is mandatory for detection of sparse and intermittent symptoms. Experiments on MIT-BIH Arrhythmia database are performed by using a well-defined data division of trained data and test data being used as a benchmark in the literature. Inter-patient heart-beat classification by a hybrid of Extreme Learning Machine and a decision rule (ELM+DR) is proposed in this paper. The main objective of this work is to identify the diseases in heart based on AAMI classes. Proposed (ELM+DR) achieved Sensitivities of 97.37%, 82.04%, 70.32%, 45.62%, PPV% of 97.24%, 83.26%, 74.7%, 32.24% for Normal, S, V and F classes respectively and a TCA of 94.65%. By using the post-processing Rule step, the sensitivity of AAMI S class has improved from 0.11% to 82.04%. PPV is improved from 1.23% to 83.26%. And Total classification accuracy (TCA) is improved from 91.97% to 94.65%. The viability and superiority of the proposed approach is confirmed in terms of (TCA).

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

ECG Heart Beat Classification by Hybrid of ELM and Decision Rule Using Feature Selection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jagadeeswara Rao Annam,
  • Bala Krishna Tilakachuri,
  • Raja Sekhar Annam

摘要

Analysis of ECG recordings is mandatory for detection of sparse and intermittent symptoms. Experiments on MIT-BIH Arrhythmia database are performed by using a well-defined data division of trained data and test data being used as a benchmark in the literature. Inter-patient heart-beat classification by a hybrid of Extreme Learning Machine and a decision rule (ELM+DR) is proposed in this paper. The main objective of this work is to identify the diseases in heart based on AAMI classes. Proposed (ELM+DR) achieved Sensitivities of 97.37%, 82.04%, 70.32%, 45.62%, PPV% of 97.24%, 83.26%, 74.7%, 32.24% for Normal, S, V and F classes respectively and a TCA of 94.65%. By using the post-processing Rule step, the sensitivity of AAMI S class has improved from 0.11% to 82.04%. PPV is improved from 1.23% to 83.26%. And Total classification accuracy (TCA) is improved from 91.97% to 94.65%. The viability and superiority of the proposed approach is confirmed in terms of (TCA).