ArrythML: An Autoencoder-Based TinyML Approach for On-Device Arrhythmia Detection on Resource-Constrained Embedded Systems
Researchers have developed ArrythML, a TinyML approach for on-device arrhythmia detection using autoencoder models. These INT8 quantized models are designed for resource-constrained embedded systems, processing over 95,000 ECG segments on an ESP32-S3 microcontroller. The best-performing model achieved an 84% recall and 79% F1-score with a 180 KB size and 9 ms inference latency, demonstrating the potential for low-power, privacy-preserving wearable systems. AI
IMPACT Enables low-power, privacy-preserving wearable devices for real-time health monitoring.