6DOF Inertial IMU Head Gesture Detection: Performance Analysis Using Fourier Transform and Jerk-Based Feature Extraction
Ionuţ-Cristian Severin, Dan-Marius Dobrea · 2020
In this paper, we further develop the system used in head gesture recognition, based on an inertial sensor, previously developed and presented in another research paper. In this research, we evaluated the performance of the head gesture recognition system using Fourier Transform and Jerk-based features. The main focus of this paper was to develop and check an alternative method for increasing the head gesture classification rate. The proposed method converts the acquired time series to the frequency domain and, from here, selects only the necessary head gesture frequency components. After selection, a conversion to the time domain with the help of the Inverse Fourier Transform was performed. During the experiments, we discovered the fact that the head gesture pattern is located in the frequency band started at 0.3 Hz and up to 23Hz. The evaluation of the classification accuracy, in this paper, was made with the help of 9 predictive algorithms and 5 deep learning algorithms. The results show that the classification accuracy of the proposed method can reach a value equal to 99.48% when the predictive models are used and 87.23% when the deep neural networks were employed.