A novel word graph expansion algorithm for voice command understanding is presented as the basis of a more accurate and robust voice command system. Tests show that the algorithm has much better performance than the generally-adopted N-best algorithm while maintaining high computational efficiency. Also the error tolerance method used to improve the robustness of the voice command understanding module further decreases the understanding error rate (UER) to 16. 6% with the computational efficiency almost unchanged compared to the case without error tolerance. A novel word graph expansion algorithm for voice command understanding is presented as the basis of a more accurate and robust voice command system. Tests show that the algorithm has much better performance than the generally-adopted N-best algorithm while maintaining high computational efficiency. the error tolerance method used to improve the robustness of the voice command understanding module further reduces the understanding error rate (UER) to 16. 6% with the computational efficiency almost unchanged compared to the case without error tolerance.