Gesture recognition is certainly gaining popularity across various domains. In scenarios where dexterity with hands or feet is challenging, such recognition methods proves particularly advantageous. The objective of this study is to compare the efficacy and suitability of an existing control scheme employing battery powered tags on lower limbs on different new positions for controlling a medical treatment bed. Following an evaluation of the new options, the best-resulting position is examined more closely and then compared with the previous published version. Thereby the preexisting neural network of the previous version is used to achieve findings of the model accuracy. The system takes gyroscope and acceleration data as input variables transmitted by a wireless tag. By comparing the positioning of the tag at the ankle with the placement in the trouser pocket, an accuracy of 97.6% which constitutes a significant improvement of the previous model with an overall accuracy of 89.1%. Given the highly favorable initial results, the potential exists to improve the performance by recording larger amounts of data, giving a greater variety of information.