Marways7/ECG_IdentificationX
🫀 基于深度学习的ECG心电身份识别系统 | Deep Learning ECG Biometric Identification System (98.44% Accuracy)
What's novel
🫀 基于深度学习的ECG心电身份识别系统 | Deep Learning ECG Biometric Identification System (98.44% Accuracy)
Code Analysis
10 files read · 3 roundsA complete ECG biometric identification pipeline that preprocesses raw ECG signals (Pan-Tompkins R-peak detection, wavelet denoising), extracts multi-domain features (HRV, MFCC, morphological), and classifies subjects using a multi-branch deep learning fusion model (1D-CNN + TDNN + MFCC-MLP with att
Strengths
Genuinely deep implementation of the full signal processing pipeline (Pan-Tompkins, HRV with DFA/SampEn, MFCC) and a well-structured multi-architecture deep learning system with proper training practices (early stopping, gradient clipping, stratified splits). Clean modular architecture with clear separation between preprocessing, feature extraction, modeling, and application layers.
Weaknesses
Zero test coverage for a project of this complexity, and a likely bug in prepare_data() where the second train_test_split references variables not properly created from the first split. The O(N²) entropy implementations and the somewhat ineffective outlier removal logic in ImprovedRPeakDetector suggest areas where edge cases weren't fully considered.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
19
Contributors
1
Files
35
Active weeks
2
Repository
Language
Python
Stars
2
Forks
0
License
MIT