A practical wearable fall detection system based on tiny convolutional neural networks

•Fall detection is critical for timely medical intervention in older adults.•Existing algorithms suffer from either poor robustness or high computational cost.•We proposed a tiny convolutional network (TinyCNN) for practical fall detection.•TinyCNN achieved a more balanced performance in both accuracy and practicality.•We further developed a practical wearable prototype system for automatic fall detection.