Improving Responsiveness of Fall Detection using Spiking Neural Networks
A low-latency fall detection system that leverages the temporal dynamics of Spiking Neural Networks for faster predictions from wearable sensor data. The project introduces Quick Spike Encoding and a time-weighted loss function to prioritize critical motion information and encourage early classification, improving detection responsiveness by over 60% while maintaining high accuracy.
- Spiking Neural Networks
- Neuromorphic Computing
- Fall Detection
- Time-Series Classification
- Wearable Sensors
- Low-Latency Inference
- Spike Encoding
