Tehnički vjesnik, Vol. 29 No. 2, 2022.
Izvorni znanstveni članak
https://doi.org/10.17559/TV-20210820150837
Evaluation of Vibration Signals Measured by 3-Axis MEMS Accelerometer on Human Face using Wavelet Transform and Classifications
Okan Oral*
; Akdeniz University, Department of Mechatronics Engineering, Engineering Faculty Dumlupinar Boulevard 07058 Campus Antalya Turkey
Suleyman Bilgin
; Akdeniz University, Department of Electric and Electronics Engineering, Engineering Faculty Dumlupinar Boulevard 07058 Campus Antalya Turkey
Mehmet Umit Ak
; Akdeniz University, Department of Electric and Electronics Engineering, Engineering Faculty Dumlupinar Boulevard 07058 Campus Antalya Turkey
Sažetak
Various studies have been conducted to reveal and analyse tissues from humans with distinct properties. The interpretation of human facial tissues was the subject of a few of these investigations. The aim of this study was to look at the energy ratios of vibration signals recorded from the human face using a 3-axis Micro-Electro-Mechanic System accelerometer sensor. 9 various measurement points on the faces of the subjects used to receive the signals are then analysed using frequency characteristics. During the analysis process, wavelet transformation values are estimated and evaluated. Thus, these regions' frequency ranges can be calculated. In addition, critical properties extracted from signals of vibration using wavelet packet transformation analysis were used as inputs of classification methods. In the next step, multilayer perceptual neural networks (MLPNN) were evaluated. In addition, the support vector machine (SVM), decision tree (DT) and binary convolution neural networks (CNN) methods were evaluated, and the success rates were compared. Finally, it is seen that the energy ratios of the signals in the hard regions are low and the energy ratios of the signals in the soft regions are high. And it has been observed that higher accuracy rate is achieved with binary CNN than with other methods.
Ključne riječi
Binary CNN; Multi-layer perceptron neural networks; Vibration signals; Wavelet packet transform
Hrčak ID:
272466
URI
Datum izdavanja:
15.4.2022.
Posjeta: 1.392 *