摘要
针对传统意图识别方法的识别单一性和实时性差的问题,提出基于GASVM-HMM算法的飞行员操控意图识别方法。20名被试者在A320飞行模拟器上进行测试,采集飞行过程中飞行员与显示屏和控制装置的交互动作数据,并建立操控意图数据集。方法将GA与SVM算法结合进行优化,提高识别的精度,并将GASVM层的输出转化为概率作为HMM层的输入值,进一步提高整体意图识别模型的准确性。与传统的算法进行对比后发现,GASVM-HMM算法的准确率较高,达到92.92%。最后进行实时验证,证明了算法的有效性。
To address the singleness and poor real-time performance of traditional intention recognition methods,this paper proposed a pilot manipulation intention recognition method based on the GASVM-HMM algorithms.20 subjects were tested on the A320 flight simulator,and the interactive action data between the pilot and the display screens or the control device were collected during the flight.The manipulation intention dataset was established as the input of the model.The model was optimized by combining GA and SVM algorithms.Then,the output by the GASVM layer was converted into the probability as the input of the HMM layer,which further improved the accuracy of intention recognition.After comparing with traditional algorithms,it is found that the overall accuracy of the GASVM-HMM algorithm is higher,reaching 92.92%.Finally,real-time verification was carried out to prove the effectiveness of the algorithm.
作者
江佳运
孙有朝
晏传奇
JIANG Jia-yun;SUN You-chao;YAN Chuan-qi(Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China)
出处
《航空计算技术》
2023年第1期43-47,共5页
Aeronautical Computing Technique
基金
国家自然科学基金与民航联合研究基金重点支持项目资助(U2033202)
国家自然科学基金项目资助(52172387)
南京航空航天大学科研与实践创新计划项目资助(xcxjh20210701)。
