Researchers Propose Quantum-Assisted Method for Wi-Fi Human Activity Recognition

A research preprint published on arXiv introduces Q-MET, a novel quantum-assisted memory-efficient training framework designed for Wi-Fi-based human activity recognition (HAR). Traditional deep learning solutions in this area require high memory and computational power due to the simultaneous optimization of millions of parameters.

Q-MET leverages a hybrid quantum-classical neural network to indirectly generate HAR model parameters, significantly reducing the trainable parameter footprint by 90% to 95% compared to standard backpropagation methods. This efficiency is achieved while maintaining or exceeding conventional classification accuracy.

To facilitate deployment on resource-constrained hardware, the framework also incorporates structured pruning during training. Experimental results demonstrate a model sparsity of 75% to 85% with less than a 2% drop in accuracy.