Patient-Specific CNN-RNN for Lung-Sound Detection With 4× Smaller Memory

Log-quant cuts model memory 4× with minimal accuracy loss; compute fits mobile SoCs, enabling real-time, wearable lung-sound screening.