Real-time emotion recognition with 40-channel EEG, facial analysis & PPG fusion - PyQt6 interface with DEAP dataset, KNN/SVM classifiers

A flow matching framework for EMG generation

Implementation of the core Adaptive Chirplet Transform (ACT) algorithm using THRML (Thermodynamic sampling) to efficiently find the best matching atom in the continuous space

Deterministic ECG codec — Python + Rust, CI parity-gated. Bounded clinical-mode contract: PRD ≤ 2.32% on MIT-BIH (48/48, mean PRD 1.12%); PTB-XL boundary disclosed (max PRD 5.29%). Cardiologist-equivalence and regulatory closure out of scope.

ECG signal processing pipeline for Atrial Fibrillation detection and Heart Rate Variability (HRV) analysis using the MIT-BIH Atrial Fibrillation Database.

EEG/EMG biosignal classification pipeline for prosthetic arm intent recognition, combining Classic ML, CNN, and RNN/BRNN models with shared preprocessing and demo tools.

Operator-based framework for interpretable early-warning detection in coupled EEG/ECG signals using phase embeddings and deterministic instability gates. Supports ablations, synthetic validation, and reproducible biosignal analysis.

Multimodal EEG-EMG deep learning pipeline for upper-limb movement decoding - 84.1% accuracy across 37 subjects (NeBULA dataset)

Python-based ECG analysis project for processing heart signal scans, extracting diagnostic features, and supporting interpretable rhythm assessment through reproducible data science workflows.

Implementation of the research paper “Heart Sounds Classification Using a CNN with 1D-Local Binary and Ternary Patterns”. Includes preprocessing, feature extraction (1D-LBP and 1D-LTP), and convolutional neural network–based classification of heart sound signals.

Sleep stage classification from raw EEG/EOG using a spatial-temporal CNN (Chambon 2018 variant). Trained on PhysioNet SleepEDF-78 with MNE-Python preprocessing, ICA artifact removal, and PyTorch. Achieves ~0.72 Cohen's Kappa on subject-wise held-out test set.

Real-time ECG biosensor using AD8232, Arduino Uno R4, Python signal processing, and Streamlit dashboard with PCB design and validation against Apple Watch

Within-subject decoding of motor imagery / motor movement from pre-cue baseline activity using cross-like spatial PE, spatiotemporal PE, temporal PE, HV spatial PE, and a combined PE features model

Non-commercial research pipeline for pain intensity classification from ECG, EDA/GSR, and EMG biosignals.