期刊名称：Journal of Automation, Mobile Robotics & Intelligent Systems (JAMRIS)
印刷版ISSN：1897-8649
电子版ISSN：2080-2145
出版年度：2019
卷号：13
期号：2
页码：28-39
DOI：10.14313/JAMRIS/2-2019/16
出版社：Industrial Research Inst. for Automation and Measurements, Warsaw
摘要：We describe simple to build mechanomyography sensors,with one or two channels, based on electret microphones.We evaluate their applicaon as a source of informaon about the operators hand sness, which can beused for changing a robots gripper sness during teleoperaon. We explain a data acuision procedure forfurther employment of a machine-learning. Finally, wepresent the results of three experiments and various machinelearning algorithms. upport vector classicaon,random forests, and neural-network architectures (fullyconnectedarcial neural networks, recurrent, convolu-onal were compared in two experiments. In rst andsecond, two probes were used with a single parcipant,with probes displaced during learning and tesng to evaluatethe inuence of probe placement on classicaon.In the third experiment, a dataset was collected using twoprobes and seven parcipants. s a result of the singleprobetests, we achieved a (binary classicaon accuracyof . uring the mul-probe tests, large crossparcipant dierences in classicaon accuracy were noted,even when normaliing per-parcipant.
