Tiny ML-based Non-Invasive Approach of Cardiac Monitoring

Cardiovascular diseases (CVDs) are a leading cause of mortality worldwide, necessitating the development of effective and continuous cardiac monitoring solutions. Advances in non-invasive monitoring and Tiny Machine Learning (TinyML) have enabled new possibilities for real-time, remote health assessment. This study introduces a novel cardiac monitoring system using TinyML to analyze Electrocardiogram (ECG) and Photoplethysmography (PPG) waveforms collected from wearable sensors. The collected signals undergo rigorous pre-processing, including noise reduction and robust feature extraction, to ensure data integrity. Spectral analysis techniques, such as Mel-Frequency Cepstrum (MFC) and Mel-Frequency Cepstral Coefficients (MFCC), are employed to extract meaningful features from ECG and PPG signals. The convolutional neural network (CNN) model is optimized for deploying on the resource-constrained environment of the Arduino Nano 33 BLE Sense. The model achieves high accuracy in real-time heart state classification, with up to 96.4% accuracy for ECG classification and 100% for PPG classification, while maintaining low computational overhead. These results demonstrate the effectiveness of TinyML usecase for wearable devices in early diagnosis and remote management of CVDs, highlighting the potential for future advancements in Edge AIbased healthcare solutions.