Surface electromyography (sEMG) signal classification is challenged by high computational cost and energy consumption, which limits its deployment on resource-constrained wearable devices. To address these issues, we propose Hyper-Dimensional Spiking Neural Networks (DSN), a hybrid framework that combines Spiking Neural Networks (SNN) with Hyper-dimen-sional Computing (HDC). The SNN module adopts randomized initialized weights with lightweight one-step training for efficient feature extraction, while the HDC module leverages high-dimensional distributed representations for robust classification. Based on theoretical derivation, we've proved its noise robustness, efficiency and hardware error robustness. Experimental results show that our method achieves up to 12.03 × lower training energy consumption compared with LSTM and 3.46 × lower compared with GRU, while maintaining over 90% accuracy using only 20% of the training data. Moreover, the framework exhibits strong robustness to noise and hardware faults, making it highly suitable for real-time, energy-efficient sEMG monitoring on edge healthcare platforms.