Surface Electrical Impedance Myography Measurements for Recognition of Numbers in American Sign Language
Surface Electrical Impedance Myograghy (sEIM) is one of the newest methods to study muscle contraction state. Essentially sEIM consists of the application of a low intensity sinusoidal electrical current, injected into the body at a set of frequencies in which the impedance of muscle is calculated based on the measured resulting voltage. This impedance is estimated to presents relevant data for hand gesture recognition. In this paper, the change of surface electrical impedance in the forearm muscle was collected during performing gestures representing the American Sign Language numbers. The measurements were carried out on the flexor muscle on the right forearm (dominant hand) on a group of 10 healthy subjects of the same age approximately. A four-pole method was considered with an inter-electrodes distance of 6 cm and a current injected amplitude equal to 0.1 mA in the frequency range between 1 kHz and 1 MHz. In order to reach high accuracy, features inspired by the medical use of sEIM for cancer and neuromuscular diseases diagnosis are extracted from the collected measurements in the first step followed by the implementation of an extreme learning machine (ELM) classifier. A five fold cross validation was applied in the training phase of the model and a prediction accuracy of 70,17% was noted during the testing phase.
Paper
Full text
Surface Electrical Impedance Myography Measurements for Recognition of Numbers in American Sign Language
Semantic Scholar · Engineering · 2020
Abstract
Surface Electrical Impedance Myograghy (sEIM) is one of the newest methods to study muscle contraction state. Essentially sEIM consists of the application of a low intensity sinusoidal electrical current, injected into the body at a set of frequencies in which the impedance of muscle is calculated based on the measured resulting voltage. This impedance is estimated to presents relevant data for hand gesture recognition. In this paper, the change of surface electrical impedance in the forearm muscle was collected during performing gestures representing the American Sign Language numbers. The measurements were carried out on the flexor muscle on the right forearm (dominant hand) on a group of 10 healthy subjects of the same age approximately. A four-pole method was considered with an inter-electrodes distance of 6 cm and a current injected amplitude equal to 0.1 mA in the frequency range between 1 kHz and 1 MHz. In order to reach high accuracy, features inspired by the medical use of sEIM for cancer and neuromuscular diseases diagnosis are extracted from the collected measurements in the first step followed by the implementation of an extreme learning machine (ELM) classifier. A five fold cross validation was applied in the training phase of the model and a prediction accuracy of 70,17% was noted during the testing phase.