Gesture spotting from IMU and EMG data for human-robot interaction

摘要

Gesture spotting is an important factor in the development of human-machine
\ninteraction modalities, which can be improved by reliable motion segmentation methods.
\nThis work uses a gesture segmentation method in order to distinguish dynamic from static
\nmotions, using IMU and EMG sensor modalities. The performance of the sensors
\nindividually as well as their combination was evaluated, with thresholds and window size
\nmanually defined for each sensor modality, through 60 sequences performed by 6 users. The
\nmethod which used the IMU alone obtained the best results in regards to the total
\nsegmentation error (11.88%), in comparison to the other two methods (EMG = 43.75% e
\nIMU+EMG= 12.92%). When considering gestures which only contain arm movement, the
\nbest error obtained was 1.11% by the IMU method (EMG = 58.89% e IMU+EMG= 7.22%).
\nHowever, when considering gestures which have only hand motion, the combination of the
\n2 sensors achieved the best performance, with an error of 10% (IMU = 30.83% e EMG=
\n17.5%). Results of the sensor fusion modality varied greatly depending on user, with
\nsegmentation errors varying between 1.25% and 26.25%, where users with more training
\nobtained better results. Application of different filtering method to the EMG data as a
\nsolution to the limb position resulted in an error for the combination of sensors of 9.17%,
\nwith all gestures performing similarly or better than the IMU method but with an increased
\nnumber of non-detected gestures.