Quantum-Enhanced Intent Recognition for Speech-Based Human-Drone Interaction

The interaction between humans and drones through the Human-Drone Interaction (HDI) by voice commands is often affected by the environment and the computing resources of edge-based drone controllers. Although classical deep learning models have gone a long way, they often fail to comply with the non-linear complexity of the intent/command recognition of low signal-to-noise ratios (SNR). A quantum-enhanced framework of sturdy intent recognition is proposed in this paper. We use a hybrid algorithm where Variational Quantum Circuits (VQC) and Quantum Kernel-based Support Vector Machines (QSVM) are used to project acoustic characteristics into high-dimensional Hilbert spaces. Through quantum entanglement and superposition, our method proves a better capability of identifying subtle semantic intentions of noisy speech cues than classical baselines. There is evidence in the literature of experimental results that quantum feature mapping can give a considerable improvement in classification accuracy, especially with heavy propeller-noise conditions, and that a more cognitive and resilient HDI can be made in the future.