The wearable indoor pedestrian navigation method based on foot binding inertial measurement units and zero velocity update(ZUPT)method as the core principle has a good application prospect.At present,indoor pedestrian...The wearable indoor pedestrian navigation method based on foot binding inertial measurement units and zero velocity update(ZUPT)method as the core principle has a good application prospect.At present,indoor pedestrian navigation methods mainly study the navigation of pedestrians during walking movements.However,when performing special tasks such as rescue and medical search,there are usually multiple motion modes such as running,going upstairs,and going downstairs,which can affect the dynamic performance of wearable indoor pedestrian navigation methods.For indoor pedestrian navigation with some similarities,based on a multi-node wearable inertial sensor network,this paper proposes a multi-level hierarchical motion modes recognition method based on an optimized long short-term memory network using a sparrow search algorithm;In response to the problems of error divergence during long-term navigation in wearable indoor pedestrian navigation methods and the inability to rely on active navigation information from the outside during special tasks,this paper proposes an indoor pedestrian navigation method based on multiple constraints.The experimental results show that in the indoor environment of approximately 2600 m2,with a total distance of over 1038.6 m,the overall recognition rate of the proposed multi-motion mode recognition methods reaches 99%,and the navigation error and RMSE value are less than 4m.展开更多
为了提升运动想象脑电(MI-EEG)信号的分类精度,提出多尺度滑窗注意力时序卷积网络(MSWATCN),充分挖掘MI-EEG信号的时空信息.结合多尺度双流分组卷积、滑动窗口多头注意力机制和窗口化时间卷积模块,实现对MI-EEG信号复杂时空特性的精准解...为了提升运动想象脑电(MI-EEG)信号的分类精度,提出多尺度滑窗注意力时序卷积网络(MSWATCN),充分挖掘MI-EEG信号的时空信息.结合多尺度双流分组卷积、滑动窗口多头注意力机制和窗口化时间卷积模块,实现对MI-EEG信号复杂时空特性的精准解码.利用多尺度卷积模块提取信号的底层时空特征,通过滑动窗口注意力机制聚焦局部关键特征,突出对分类任务重要的信息.窗口化时间卷积模块通过建模时间序列中的长期依赖关系,增强模型处理时序信息的能力.实验结果表明,MSWATCN在BCI Competition IV 2a和2b数据集上的分类准确率和一致性优于对比网络和基准模型.展开更多
基金partially supported by the National Natural Science Foundation of China(Grant No.62103285)the National Defense Basic Research Program(JCKY2020605C009)the Fundamental Research Funds for the Central Universities(Grant No.QZPY202310)
文摘The wearable indoor pedestrian navigation method based on foot binding inertial measurement units and zero velocity update(ZUPT)method as the core principle has a good application prospect.At present,indoor pedestrian navigation methods mainly study the navigation of pedestrians during walking movements.However,when performing special tasks such as rescue and medical search,there are usually multiple motion modes such as running,going upstairs,and going downstairs,which can affect the dynamic performance of wearable indoor pedestrian navigation methods.For indoor pedestrian navigation with some similarities,based on a multi-node wearable inertial sensor network,this paper proposes a multi-level hierarchical motion modes recognition method based on an optimized long short-term memory network using a sparrow search algorithm;In response to the problems of error divergence during long-term navigation in wearable indoor pedestrian navigation methods and the inability to rely on active navigation information from the outside during special tasks,this paper proposes an indoor pedestrian navigation method based on multiple constraints.The experimental results show that in the indoor environment of approximately 2600 m2,with a total distance of over 1038.6 m,the overall recognition rate of the proposed multi-motion mode recognition methods reaches 99%,and the navigation error and RMSE value are less than 4m.
文摘为了提升运动想象脑电(MI-EEG)信号的分类精度,提出多尺度滑窗注意力时序卷积网络(MSWATCN),充分挖掘MI-EEG信号的时空信息.结合多尺度双流分组卷积、滑动窗口多头注意力机制和窗口化时间卷积模块,实现对MI-EEG信号复杂时空特性的精准解码.利用多尺度卷积模块提取信号的底层时空特征,通过滑动窗口注意力机制聚焦局部关键特征,突出对分类任务重要的信息.窗口化时间卷积模块通过建模时间序列中的长期依赖关系,增强模型处理时序信息的能力.实验结果表明,MSWATCN在BCI Competition IV 2a和2b数据集上的分类准确率和一致性优于对比网络和基准模型.
